PSA: Malta 7990s are NOT suitable for mining!

[dogie]

Hello all. After my experiences with 2x partner card (PC) non Malta 7990s, I wanted to raise my concerns about the new Malta (M) versions. I hope one of my first threads is useful to the community.

Mining:
Just for reference, at 1050 cores I get 1280 per card for ~400W from the wall.


Pricing:
So the PC cards come in at about $900, 7970s at $400. This means they are viable for limited space internal card builds and will possibly hold their value better as they'll stay in the usable performance bracket longer.

The new M cards come in at $1000. Although they have more games, its pushing it a bit. Not a problem, but something for each to evaluate.

Tldr: New M cards more expensive, but have more games.


Temperatures and cooling:
Partner cards are 3.3 slot and weigh nearly 3kg each. This means they do NOT work without at least a full slot gap. Most motherboards don't have a usable non-riserable slot in the first expansion slot, so you require an 8 slot case. Not a problem.

They take up 3 slots AND the typical gap between the bottom and next slot. Running without a free gap you can only mine on 3 cores. Regardless of the underclock, TDP limiting, undervolting or intensity, the 4th core always overheats.

With a gap the bottom card loads at +45-50 above ambient, and the top card +55-65 above ambient. [Note I don't have voltage control of the 2nd core on each card, hopefully this will arrive with newer drivers as I have control when there is only ONE card - this will improve temps slightly]. This is with a 1050 core, undervolted.

Now the new M cards have dual slot coolers, ouch. At stock clocks they were loading just below the 80s, horizontally mounted in an open air, single card system. This was on GPGPU computing. Due to the heatsink design, ZERO air is exhausted and its all left to be removed by the case.  My PC cards heat the case ambient up to 40+ even with 3x250mm, 5x240mm of fans removing heat. You will have to open air, just have to.

Tldr: Mining is doable on the 3 slot PC cards, but will be almost impossible to dual M cards within a case. The M cards are 2 slot and don't exhaust air at all.


Power delivery:
This is the major problem. My PC cards have 3x8pin for a max of 525W. These new M cards only have 2x8 pins for 375W max. Even with my moderate overclocks (focusing on power consumption rather than MAX POWA), I'm going over/hitting the power delivery limits of the PC cards at about ~400W from the wall per card [1050 core, mem 625 and undervolted]. AnandTech saw this in Furmark, the card was power throttled HEAVILY. Now imagine trying to scypt mine with decent clocks - its not going to happen.

Tldr: The power delivery on these cards LIMITS the performance of these cards and causes power throttling even on moderate computing load. Hence these cards are not appropriate for mining.

I'm more than happy with my 3 slot non Maltas, hopefully more stock is made. If I helped, small small small donations are appreciated and pay for good things

[1714797247]

1714797247

[computerparts]

You really should do more research so that you know what you're talking about when posting. There's a ton of misinformation you have so I'll only touch on one. You correlate Furmark and scrypt. First, there's no dual gpu card in existence that does not get throttled from running Furmark. Second, scrypt is more reliant on ram than it is on the core clock. Where as Furmark stresses the core more than anything.

[dogie]

You really should do more research so that you know what you're talking about when posting. There's a ton of misinformation you have so I'll only touch on one. You correlate Furmark and scrypt. First, there's no dual gpu card in existence that does not get throttled from running Furmark. Second, scrypt is more reliant on ram than it is on the core clock. Where as Furmark stresses the core more than anything.

Anand: FurMark ended up being harder on the 7990 than we expected. Both NVIDIA and AMD throttle it of course, but we’ve never seen it throttled so hard as on the 7990. The 7990 completely stalled at times and momentarily dropped to its medium power state (500MHz) while running FurMark. The results are still enlightening since we’re clearly hitting it peak power limits regardless, but it’s not a very good sustained load in this case."
http://anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-official/16

Just tried for arguments sake, furmark working fine maxing out on my two PC 7990s for the last 20 minutes. But anyway, this is not the source of my power consumption figures. I am comparing my cards, in hand, and their power consumption bitcoin and litecoin mining. They use way, way more than the PC cards can output and I'm not even pushing them hard [1050 core, 625 ram, undervolted core and ram]

Maybe you should do more research so that you know what you're talking about when posting, thanks.

[computerparts]

Tldr: The power delivery on these cards LIMITS the performance of these cards and causes power throttling even on moderate computing load. Hence these cards are not appropriate for mining.

You're joking right? Just because these cards hit their power limits on Furmark means they are not suitable for mining? Furmark is the most stressful test a gpu can ever face. SHA-256 does stress the gpu but it is strictly a computational load which is a big difference from the rendering load in Furmark. And scrypt comes nowhere near the stress Furmark is capable of.

Tldr: Mining is doable on the 3 slot PC cards, but will be almost impossible to dual M cards within a case. The M cards are 2 slot and don't exhaust air at all.

This is completely wrong. I'm not sure where you came up with the idea that it won't exhaust any air out of the back. The bracket is vented for a reason. But most serious miners are going to be running open air rack/crate mounted rigs anyway.

Power delivery:
This is the major problem. My PC cards have 3x8pin for a max of 525W. These new M cards only have 2x8 pins for 375W max. Even with my moderate overclocks (focusing on power consumption rather than MAX POWA), I'm going over/hitting the power delivery limits of the PC cards at about ~400W from the wall per card [1050 core, mem 625 and undervolted]. AnandTech saw this in Furmark, the card was power throttled HEAVILY. Now imagine trying to scypt mine with decent clocks - its not going to happen.

About the only thing you have right here are the pin inputs. You seem to be confusing TDP with power draw. TDP is heat measured in watts. They are not the same. Example: The Malta has a TDP of 375w but power draw at load can exceed 500w.

[dogie]

You're joking right? Just because these cards hit their power limits on Furmark means they are not suitable for mining? Furmark is the most stressful test a gpu can ever face. SHA-256 does stress the gpu but it is strictly a computational load which is a big difference from the rendering load in Furmark. And scrypt comes nowhere near the stress Furmark is capable of.

I'm actually basing my power consumption figures on those 7990s im mining on now - they are very very close to the furmark based consumption in those reviews - ie the cards are maxing. But then again my actual data is clearly superseded by your thin air evidence.

This is completely wrong. I'm not sure where you came up with the idea that it won't exhaust any air out of the back. The bracket is vented for a reason. But most serious miners are going to be running open air rack/crate mounted rigs anyway.

"For the 7990 however AMD has dropped the blowers entirely for a completely open air design. The tradeoff between the two being that while blowers are self-sustaining and ensure all hot air is expelled by the card – or at least half of the hot air in the case of half-blowers – open air coolers move more of the work to the chassis in exchange for generally lower noise levels.

At the same time this change does mean that the chassis/case used becomes more important than ever. Of the 375W of heat generated by the 7990, only a fraction of it will be kicked out via its vent; the rest will be dumped into the case"
http://www.anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-official/2

I think you can keep your magical fluid dynamics.

About the only thing you have right here are the pin inputs. You seem to be confusing TDP with power draw. TDP is heat measured in watts. They are not the same. Example: The Malta has a TDP of 375w but power draw at load can exceed 500w.

"On the power front their binning has enabled them to get a dual-GPU Tahiti card out at 375W. The 7990 is a 375W card...."
http://www.anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-official/16

"Maximum Power 375 W.
AMD buttons the Radeon HD 7990 up into a dual-slot card that only needs two eight-pin auxiliary power connectors to drive it. With that said, the card jams right up against the PCI-SIG’s electromechanical specification for one x16 slot and two eight-pin connectors: 375 W. "
http://www.tomshardware.co.uk/radeon-hd-7990-review-benchmark,review-32675-2.html

I'm very surprised for someone calling themselves 'computerparts' how uneducated you are, but at the same time the air of superiority you carry when you are plain wrong. In fact, the TDP is a great indicator of how much electricity the chip itself will use. Smaller effects keep the two being identical but they are very, very close.

Electrical energy is converted PURELY into heat energy - there is nothing else. If you think it consumes 500W and only generates 375W, where do you think that 125W of energy is going xD MAGIC.

TDP = maximum heat dissipation = maximum heat energy created/s = maximum electrical energy consumed/s = maximum power consumption. THAT ENGINEERING.

[computerparts]

I tried to educate you and this is what I get in return. So I see you not only need classes on thermodynamics but reading comprehension as well.

I'm actually basing my power consumption figures on those 7990s im mining on now - they are very very close to the furmark based consumption in those reviews - ie the cards are maxing. But then again my actual data is clearly superseded by your thin air evidence.

I never said anything about the cards not maxing. The fact is the gpu itself is rendering during furmark. It is NOT rendering during SHA-256 or scrypt. Do you even know what causes throttling? There is NOTHING that will make it throttle during computational loads.

At the same time this change does mean that the chassis/case used becomes more important than ever. Of the 375W of heat generated by the 7990, only a fraction of it will be kicked out via its vent; the rest will be dumped into the case"
http://www.anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-official/2

Again I never said anything about it not blowing air into the case. But it will exhaust air out of the back. It's not a big a deal as you're making it out to be.

"On the power front their binning has enabled them to get a dual-GPU Tahiti card out at 375W. The 7990 is a 375W card...."
http://www.anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-official/16

"Maximum Power 375 W.
AMD buttons the Radeon HD 7990 up into a dual-slot card that only needs two eight-pin auxiliary power connectors to drive it. With that said, the card jams right up against the PCI-SIG’s electromechanical specification for one x16 slot and two eight-pin connectors: 375 W. "
http://www.tomshardware.co.uk/radeon-hd-7990-review-benchmark,review-32675-2.html

First of all, don't ever quote anything from Tom's. Doing so will reduce your credibility. I knew you were going to use that quote from Anand. Shame on them for not clarifying. Look at the specs. See where it says Powertune/TDP? And what is that number? 375. They are not talking about the maximum power draw. They are talking about the efficiency as a result of the binning process. Do you know what binning is? In layman's terms, it's when chip manufacturers pick out the best yields and choose the chips with the tightest tolerances.

Electrical energy is converted PURELY into heat energy - there is nothing else. If you think it consumes 500W and only generates 375W, where do you think that 125W of energy is going xD MAGIC.

TDP = maximum heat dissipation = maximum heat energy created/s = maximum electrical energy consumed/s = maximum power consumption. THAT ENGINEERING.

I'm not even going to say how dumb you make yourself look here. If this were the case none of your electronics would work because according to you, all of the electrical energy a device is using is converted to heat. You're completely ignoring the law of conservation of energy. I'm glad the real world isn't based on your rules. If a device consumes 500w and only generates 375w of heat, that means it's using 125w of that energy to do what it's suppsed to do and the rest (375w) is wasted as heat.

[dogie]

I tried to educate you and this is what I get in return. So I see you not only need classes on thermodynamics but reading comprehension as well.

Masters in Engineering, thanks. How about you? Nope, didn't think so.

If a device consumes 500w and only generates 375w of heat, it means it's using that 125w to do what it's supposed to do and the rest (375w) is wasted in heat.

This is priceless. I'm actually going to put this quote up on my wall.

[Spendulus]

I tried to educate you and this is what I get in return. So I see you not only need classes on thermodynamics but reading comprehension as well.

Masters in Engineering, thanks. How about you? Nope, didn't think so.

If a device consumes 500w and only generates 375w of heat, it means it's using that 125w to do what it's supposed to do and the rest (375w) is wasted in heat.

This is priceless. I'm actually going to put this quote up on my wall.

A part of the wattage is converted into accelerating air from a standstill to design speed for cooling purposes.  I suspect this becomes quite substantial. 

Consider the case of a fan alone.  It generates some heat (minor) and useful work (Major).

I'd say both you guys had the amps in --> heat/work wrong.

LOL

[dogie]

I tried to educate you and this is what I get in return. So I see you not only need classes on thermodynamics but reading comprehension as well.

Masters in Engineering, thanks. How about you? Nope, didn't think so.

If a device consumes 500w and only generates 375w of heat, it means it's using that 125w to do what it's supposed to do and the rest (375w) is wasted in heat.

This is priceless. I'm actually going to put this quote up on my wall.

A part of the wattage is converted into accelerating air from a standstill to design speed for cooling purposes.  I suspect this becomes quite substantial. 

Consider the case of a fan alone.  It generates some heat (minor) and useful work (Major).

I'd say both you guys had the amps in --> heat/work wrong.

LOL

And where does the kinetic energy from the fan go? That's right, to heat. As soon as the air stops moving, its converted all its KE to more heat energy.

But no, a fan is not the same as a GPU.

[HellDiverUK]

So the short version of this is:

The 7990 can mine no problems, but can't run Furmark.

Good thread, OP.   

[melmo]

 OP is correct - computers in general (not just gpus) are almost 100% efficient in converting electricity to heat.

[dogie]

So the short version of this is:

The 7990 can mine no problems, but can't run Furmark.

Good thread, OP.   

Yes and no.

The 7990 from third parties [ie the 3 slot 525W version] CAN mine no problems as long as you have 1+ gap in-between.
The official 'Malta' 7990 in 2 slot and 375W CAN'T mine - too many heat and power limitations.

[BBQKorv]

So the short version of this is:

The 7990 can mine no problems, but can't run Furmark.

Good thread, OP.   

Yes and no.

The 7990 from third parties [ie the 3 slot 525W version] CAN mine no problems as long as you have 1+ gap in-between.
The official 'Malta' 7990 in 2 slot and 375W CAN'T mine - too many heat and power limitations.

I use those third party 3-slot 7990s for mining, and they do very well. I haven't tested Malta yet, but I can assume it will have issues mining. If you own a non-Malta 7990 just feel the thermals, you'll get trouble if that gets shrinked to 2 slots and to only 2 power connectors.

Furmark throttling is nothing in comparison to what scrypt mining or OC'd sha256 will do to the card. My mining temps and power draw is bigger in mining than in Furmark.

[Miner99er]

Why don't we just wait till a M***** F***** around here gets a Malta 7990, slaps CGminer on it, and see what the damn thing does?

Too much epenis waving in the air around here.

[dogie]

Why don't we just wait till a M***** F***** around here gets a Malta 7990, slaps CGminer on it, and see what the damn thing does?

Too much epenis waving in the air around here.

I'd rather wave it around and save someone $1000 of disappointment, time and dismay.

[computerparts]

I'm going to mine with 3 of them. OP has no clue what he's talking about.

[dogie]

I'm going to mine with 3 of them. OP has no clue what he's talking about.

xD PLEASE do, we will enjoy the spectacle of you trying to exhaust 1200W of heat from a case while getting 500MH a card due to throttling. Please post updates.

Self belief is only a positive attribute up to a point, at which it becomes a blinding detriment.

[computerparts]

I'm going to mine with 3 of them. OP has no clue what he's talking about.

xD PLEASE do, we will enjoy the spectacle of you trying to exhaust 1200W of heat from a case while getting 500MH a card due to throttling. Please post updates.

Self belief is only a positive attribute up to a point, at which it becomes a blinding detriment.

I never said I was going to stuff them in a case. But there is a case I can think of that has proven to handle that task so I'll glady take your challenge. Care to put some btc on it as well while we're at it?

[dogie]

I'm going to mine with 3 of them. OP has no clue what he's talking about.

xD PLEASE do, we will enjoy the spectacle of you trying to exhaust 1200W of heat from a case while getting 500MH a card due to throttling. Please post updates.

Self belief is only a positive attribute up to a point, at which it becomes a blinding detriment.

I never said I was going to stuff them in a case. But there is a case I can think of that has proven to handle that task so I'll glady take your challenge. Care to put some btc on it as well while we're at it?

No, because you're the only one who can verify the results and I don't trust you one bit. And oh look, you have a vested interest in one of the outcomes.

[Aggrophobia]

game package seems to be worth 400$

[Spendulus]

I tried to educate you and this is what I get in return. So I see you not only need classes on thermodynamics but reading comprehension as well.

Masters in Engineering, thanks. How about you? Nope, didn't think so.

If a device consumes 500w and only generates 375w of heat, it means it's using that 125w to do what it's supposed to do and the rest (375w) is wasted in heat.

This is priceless. I'm actually going to put this quote up on my wall.

A part of the wattage is converted into accelerating air from a standstill to design speed for cooling purposes.  I suspect this becomes quite substantial. 

Consider the case of a fan alone.  It generates some heat (minor) and useful work (Major).

I'd say both you guys had the amps in --> heat/work wrong.

LOL

And where does the kinetic energy from the fan go? That's right, to heat. As soon as the air stops moving, its converted all its KE to more heat energy.

But no, a fan is not the same as a GPU.

By that logic there is no "work", only "heat".  And that's simply not true.  The useful work performed here is in moving heat from one place to another through convection.  This is not complicated.

Next we'll be arguing that there's no useful work in the entire universe, only heat....

[computerparts]

I'm going to mine with 3 of them. OP has no clue what he's talking about.

xD PLEASE do, we will enjoy the spectacle of you trying to exhaust 1200W of heat from a case while getting 500MH a card due to throttling. Please post updates.

Self belief is only a positive attribute up to a point, at which it becomes a blinding detriment.

I never said I was going to stuff them in a case. But there is a case I can think of that has proven to handle that task so I'll glady take your challenge. Care to put some btc on it as well while we're at it?

No, because you're the only one who can verify the results and I don't trust you one bit. And oh look, you have a vested interest in one of the outcomes.

Lol in other words you have ZERO confidence in your claims. Put up or shut up.

[computerparts]

So the short version of this is:

The 7990 can mine no problems, but can't run Furmark.

Good thread, OP.   

Yes and no.

The 7990 from third parties [ie the 3 slot 525W version] CAN mine no problems as long as you have 1+ gap in-between.
The official 'Malta' 7990 in 2 slot and 375W CAN'T mine - too many heat and power limitations.

I use those third party 3-slot 7990s for mining, and they do very well. I haven't tested Malta yet, but I can assume it will have issues mining. If you own a non-Malta 7990 just feel the thermals, you'll get trouble if that gets shrinked to 2 slots and to only 2 power connectors.

Furmark throttling is nothing in comparison to what scrypt mining or OC'd sha256 will do to the card. My mining temps and power draw is bigger in mining than in Furmark.

The Malta would certainly have trouble IF it used the exact same chips used in the Powercolor version. But the Malta uses higher binned chips that are more efficient which is why they can give the Malta a 375w TDP

[waterboyserver]

Not taking sides on the discussion in this thread. If power and heat seems to be a problem with Malta, it would be interesting to have it liquid cooled - though EK is releasing a block, has anyone used Arctic Accelero Hybrid coolers for mining (despite its price)?  It stands to reason the pump/block section would fit well, and the tubing might be able to pass through the grills on the top of the Malta card. The 7970 version of the Arctic cooler is expensive, using an EK 79xx shim with the regular Arctic hybrid might work interestingly - almost cooled like the Ares II but two separate loops.

[BBQKorv]

So the short version of this is:

The 7990 can mine no problems, but can't run Furmark.

Good thread, OP.  

Yes and no.

The 7990 from third parties [ie the 3 slot 525W version] CAN mine no problems as long as you have 1+ gap in-between.
The official 'Malta' 7990 in 2 slot and 375W CAN'T mine - too many heat and power limitations.

I use those third party 3-slot 7990s for mining, and they do very well. I haven't tested Malta yet, but I can assume it will have issues mining. If you own a non-Malta 7990 just feel the thermals, you'll get trouble if that gets shrinked to 2 slots and to only 2 power connectors.

Furmark throttling is nothing in comparison to what scrypt mining or OC'd sha256 will do to the card. My mining temps and power draw is bigger in mining than in Furmark.

The Malta would certainly have trouble IF it used the exact same chips used in the Powercolor version. But the Malta uses higher binned chips that are more efficient which is why they can give the Malta a 375w TDP

If Malta does have even higher binned chips than those 3-slot models, then it will be a better product than the current 7990 offerings. I'm looking forward to see where this goes when we have the first mining benchmarks of Malta. I hope I was wrong in expecting Malta to have similar chips than current 3-slot models do have. When this is cleared, most likely I'll just get more of the better model, whichever it ends up being.

Density and power consumption are better than in 7970, and that counts for me.

[rograz]

Density and power consumption are better than in 7970, and that counts for me.

Good luck running 4 or more of these in 1 rig, power consumption is only better if you don't touch the vcore of the 7970s. In general the dual cards pull 5-10W extra due the bridge chip etc and if all other parameters are identical (vcore/clocks etc) in general 2x single cards will be cooler and pulls less power than a dual card. Also add the fact that the leakage increases quite a bit with the 7xxx architecture at higher temps and it becomes even worse.

Most of the time they cost more than 2x of a single card so tell me again why someone should buy these for mining

Back when people started mining large scale with GPUs back in 2011 it was a lot hard to find decently priced motherboards with 6x PCIe slots, that was one of the main reasons 5970 became so popular for mining. Also the usage and availability of extension cables was more limited back then. There was also the whole issue about the shorting the detection pins for 1x cables that wasn't really public knowledge until mid/late 2011. These days on the other hand you can find boards for less than 100 bucks with 6 slots and higher density than 6x7970/board is just completely pointless.

[dogie]

The Malta would certainly have trouble IF it used the exact same chips used in the Powercolor version. But the Malta uses higher binned chips that are more efficient which is why they can give the Malta a 375w TDP

Looking at the data, I don't believe they are. Stock clocks less than 1GHz chips, max overclocks seem similar to my PC boards and less than decent 7970s. Power consumption was 15W less than the PC, no where near enough to make up for the lack of power headroom.

Not taking sides on the discussion in this thread. If power and heat seems to be a problem with Malta, it would be interesting to have it liquid cooled - though EK is releasing a block, has anyone used Arctic Accelero Hybrid coolers for mining (despite its price)?  It stands to reason the pump/block section would fit well, and the tubing might be able to pass through the grills on the top of the Malta card. The 7970 version of the Arctic cooler is expensive, using an EK 79xx shim with the regular Arctic hybrid might work interestingly - almost cooled like the Ares II but two separate loops.

Did you mean 7990? But anyway, this WOULDN'T work.
Max. Cooling Capacity    320 Watts. You'll max out the cooler very quickly using 375W to 525W [disputed], then just start rocketing temps until it thermal throttle crashes. This gets away with cooling a Titan because its very close to actual max power consumption and it typically assumes you're not going to mine 24/7.

It also wouldn't fix the max power draw being an issue.

Density and power consumption are better than in 7970, and that counts for me.

Good luck running 4 or more of these in 1 rig, power consumption is only better if you don't touch the vcore of the 7970s. In general the dual cards pull 5-10W extra due the bridge chip etc and if all other parameters are identical (vcore/clocks etc) in general 2x single cards will be cooler and pulls less power than a dual card. Also add the fact that the leakage increases quite a bit with the 7xxx architecture at higher temps and it becomes even worse.

Most of the time they cost more than 2x of a single card so tell me again why someone should buy these for mining

Back when people started mining large scale with GPUs back in 2011 it was a lot hard to find decently priced motherboards with 6x PCIe slots, that was one of the main reasons 5970 became so popular for mining. Also the usage and availability of extension cables was more limited back then. There was also the whole issue about the shorting the detection pins for 1x cables that wasn't really public knowledge until mid/late 2011. These days on the other hand you can find boards for less than 100 bucks with 6 slots and higher density than 6x7970/board is just completely pointless.

Exactly, no way you'd be able to run 4, or there be much point. If you're running in a case, congratulations it alone with your $4k of cards is now on fire. If running out of a case, why did you buy such high density?

The ONLY case where it makes sense, is what I used it for. I wanted a mining rig, but I only have the space to store one reasonable case.

[rograz]

Maybe for rack mounting 3x7990 rigs, maybe. That would be the only reason I can think of getting them, and that's me grasping for straws.

[dogie]

By that logic there is no "work", only "heat".  And that's simply not true.  The useful work performed here is in moving heat from one place to another through convection.  This is not complicated.

Next we'll be arguing that there's no useful work in the entire universe, only heat....

In the use of a fan, electrical energy is converted to work which is very quickly and consistently turned to heat. Why do you think a fan makes a room hotter? Because it is converted electricity to heat. -anyway, this is fans.

In electronics there is zero work as the cycle happens so fast. Apart from the tiny portion of energy required to 'start' the cycle, all the conversion is from electricity to heat.

[dogie]

I'm going to mine with 3 of them. OP has no clue what he's talking about.

xD PLEASE do, we will enjoy the spectacle of you trying to exhaust 1200W of heat from a case while getting 500MH a card due to throttling. Please post updates.

Self belief is only a positive attribute up to a point, at which it becomes a blinding detriment.

I never said I was going to stuff them in a case. But there is a case I can think of that has proven to handle that task so I'll glady take your challenge. Care to put some btc on it as well while we're at it?

No, because you're the only one who can verify the results and I don't trust you one bit. And oh look, you have a vested interest in one of the outcomes.

Lol in other words you have ZERO confidence in your claims. Put up or shut up.

No, its like me offering you a bet as to how many cups I have on my desk right now. Regardless of the true answer, I can easily fake it and there is no way to prove otherwise. I'd have to be well, you, to accept that bet.

[computerparts]

The Malta would certainly have trouble IF it used the exact same chips used in the Powercolor version. But the Malta uses higher binned chips that are more efficient which is why they can give the Malta a 375w TDP

Looking at the data, I don't believe they are. Stock clocks less than 1GHz chips, max overclocks seem similar to my PC boards and less than decent 7970s. Power consumption was 15W less than the PC, no where near enough to make up for the lack of power headroom.

I'm not sure how you missed this in the Anand review you were touting so much about

But perhaps the most defining aspect of AMD’s 7990, and the thing that sets it apart from unofficial 7990s that came before it is the TDP. AMD’s 7990 has an official TDP of just 375W, which although common for official dual-GPU cards, is quite a bit lower than the TDPs of the unofficial 7990s. As the GPU manufacturer AMD has the ability to do finely grained binning that their partners cannot, so while Asus and PowerColor have essentially been putting together cards that really are two 7970s on a single card – right down to the TDP – official 7990s get the advantage of AMD’s binning process, significantly reducing power consumption.

Third paragraph below the spec comparison table http://www.anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-official

No, its like me offering you a bet as to how many cups I have on my desk right now. Regardless of the true answer, I can easily fake it and there is no way to prove otherwise. I'd have to be well, you, to accept that bet.

I'm not sure how I could possibly fake pictures of the setup and cgminer running. I'm sure it can be done but it is certainly beyond my ability to do so.

[dogie]

The Malta would certainly have trouble IF it used the exact same chips used in the Powercolor version. But the Malta uses higher binned chips that are more efficient which is why they can give the Malta a 375w TDP

Looking at the data, I don't believe they are. Stock clocks less than 1GHz chips, max overclocks seem similar to my PC boards and less than decent 7970s. Power consumption was 15W less than the PC, no where near enough to make up for the lack of power headroom.

I'm not sure how you missed this in the Anand review you were touting so much about

But perhaps the most defining aspect of AMD’s 7990, and the thing that sets it apart from unofficial 7990s that came before it is the TDP. AMD’s 7990 has an official TDP of just 375W, which although common for official dual-GPU cards, is quite a bit lower than the TDPs of the unofficial 7990s. As the GPU manufacturer AMD has the ability to do finely grained binning that their partners cannot, so while Asus and PowerColor have essentially been putting together cards that really are two 7970s on a single card – right down to the TDP – official 7990s get the advantage of AMD’s binning process, significantly reducing power consumption.

Third paragraph below the spec comparison table http://www.anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-official

No, its like me offering you a bet as to how many cups I have on my desk right now. Regardless of the true answer, I can easily fake it and there is no way to prove otherwise. I'd have to be well, you, to accept that bet.

I'm not sure how I could possibly fake pictures of the setup and cgminer running. I'm sure it can be done but it is certainly beyond my ability to do so.

I've read that, looking at specs tells us nothing other than one heat system can remove 375W and one 525W - that's not an indication of actual performance. If you look at the actual power consumption in Anand's review you'll see the cards aren't more than 15W apart.

The 'bet' is surrounding cooling and power solutions. How am I to know if you put it next to a 5C AC and go "oh look not overheating". No I don't and I don't trust you enough to even believe your name.

[computerparts]

The Malta would certainly have trouble IF it used the exact same chips used in the Powercolor version. But the Malta uses higher binned chips that are more efficient which is why they can give the Malta a 375w TDP

Looking at the data, I don't believe they are. Stock clocks less than 1GHz chips, max overclocks seem similar to my PC boards and less than decent 7970s. Power consumption was 15W less than the PC, no where near enough to make up for the lack of power headroom.

I'm not sure how you missed this in the Anand review you were touting so much about

But perhaps the most defining aspect of AMD’s 7990, and the thing that sets it apart from unofficial 7990s that came before it is the TDP. AMD’s 7990 has an official TDP of just 375W, which although common for official dual-GPU cards, is quite a bit lower than the TDPs of the unofficial 7990s. As the GPU manufacturer AMD has the ability to do finely grained binning that their partners cannot, so while Asus and PowerColor have essentially been putting together cards that really are two 7970s on a single card – right down to the TDP – official 7990s get the advantage of AMD’s binning process, significantly reducing power consumption.

Third paragraph below the spec comparison table http://www.anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-official

No, its like me offering you a bet as to how many cups I have on my desk right now. Regardless of the true answer, I can easily fake it and there is no way to prove otherwise. I'd have to be well, you, to accept that bet.

I'm not sure how I could possibly fake pictures of the setup and cgminer running. I'm sure it can be done but it is certainly beyond my ability to do so.

I've read that, looking at specs tells us nothing other than one heat system can remove 375W and one 525W - that's not an indication of actual performance. If you look at the actual power consumption in Anand's review you'll see the cards aren't more than 15W apart.

The 'bet' is surrounding cooling and power solutions. How am I to know if you put it next to a 5C AC and go "oh look not overheating". No I don't and I don't trust you enough to even believe your name.

Now you bring in ambient temps ha! You're full of excuses aren't you? I'll leave it at that.

[dogie]

The Malta would certainly have trouble IF it used the exact same chips used in the Powercolor version. But the Malta uses higher binned chips that are more efficient which is why they can give the Malta a 375w TDP

Looking at the data, I don't believe they are. Stock clocks less than 1GHz chips, max overclocks seem similar to my PC boards and less than decent 7970s. Power consumption was 15W less than the PC, no where near enough to make up for the lack of power headroom.

I'm not sure how you missed this in the Anand review you were touting so much about

But perhaps the most defining aspect of AMD’s 7990, and the thing that sets it apart from unofficial 7990s that came before it is the TDP. AMD’s 7990 has an official TDP of just 375W, which although common for official dual-GPU cards, is quite a bit lower than the TDPs of the unofficial 7990s. As the GPU manufacturer AMD has the ability to do finely grained binning that their partners cannot, so while Asus and PowerColor have essentially been putting together cards that really are two 7970s on a single card – right down to the TDP – official 7990s get the advantage of AMD’s binning process, significantly reducing power consumption.

Third paragraph below the spec comparison table http://www.anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-official

No, its like me offering you a bet as to how many cups I have on my desk right now. Regardless of the true answer, I can easily fake it and there is no way to prove otherwise. I'd have to be well, you, to accept that bet.

I'm not sure how I could possibly fake pictures of the setup and cgminer running. I'm sure it can be done but it is certainly beyond my ability to do so.

I've read that, looking at specs tells us nothing other than one heat system can remove 375W and one 525W - that's not an indication of actual performance. If you look at the actual power consumption in Anand's review you'll see the cards aren't more than 15W apart.

The 'bet' is surrounding cooling and power solutions. How am I to know if you put it next to a 5C AC and go "oh look not overheating". No I don't and I don't trust you enough to even believe your name.

Now you bring in ambient temps ha! You're full of excuses aren't you? I'll leave it at that.

.... Yes? As already discussed ambient is going to affect power usage and actual TDP of the cooling? Really, this is why you're what, 35, and still running a piddly little computer store trying to sell usb cables to diddy old grannies.

[FloridaBear]

Next we'll be arguing that there's no useful work in the entire universe, only heat....

Correct, ultimately.

[Dyaheon]

With regards to power throttling, there's a powertune +20% option. I believe that should let you mine without trouble, as it allows for 375W*1.2 = 450W. Furmark is indeed not a good benchmark, it is considered a 'power virus'. Even if power is a problem, slight undervolting should take care of that.

Temperatures and cooling may be a problem, but it's a matter of ambient temps and airflow. If you have an open air rig with extenders and plenty of space for the cards, I'm sure you'll be fine.

The question is more about whether it's a good idea to buy a $1000 card for mining now, instead of getting multiple 5xxx/79xx cards or an ASIC. Or just buying coins directly. The time of GPUs might be coming to an end.

[dogie]

With regards to power throttling, there's a powertune +20% option. I believe that should let you mine without trouble, as it allows for 375W*1.2 = 450W. Furmark is indeed not a good benchmark, it is considered a 'power virus'. Even if power is a problem, slight undervolting should take care of that.

Temperatures and cooling may be a problem, but it's a matter of ambient temps and airflow. If you have an open air rig with extenders and plenty of space for the cards, I'm sure you'll be fine.

The question is more about whether it's a good idea to buy a $1000 card for mining now, instead of getting multiple 5xxx/79xx cards or an ASIC. Or just buying coins directly. The time of GPUs might be coming to an end.

The limit is the cooling system and power DELIVERY in this case [same as you'd have on a high end SINGLE 7970 with a little more beef]. It doesn't matter if you tell the card its allowed more than designed as in this case its hard limited by the power delivery system. 75W PCIE and 2x150W PCIE 8 pins = 375W. Can't 'force' more due to the onboard hardware without seriously seriously risking card death.

[computerparts]

With regards to power throttling, there's a powertune +20% option. I believe that should let you mine without trouble, as it allows for 375W*1.2 = 450W. Furmark is indeed not a good benchmark, it is considered a 'power virus'. Even if power is a problem, slight undervolting should take care of that.

Temperatures and cooling may be a problem, but it's a matter of ambient temps and airflow. If you have an open air rig with extenders and plenty of space for the cards, I'm sure you'll be fine.

The question is more about whether it's a good idea to buy a $1000 card for mining now, instead of getting multiple 5xxx/79xx cards or an ASIC. Or just buying coins directly. The time of GPUs might be coming to an end.

The limit is the cooling system and power DELIVERY in this case [same as you'd have on a high end SINGLE 7970 with a little more beef]. It doesn't matter if you tell the card its allowed more than designed as in this case its hard limited by the power delivery system. 75W PCIE and 2x150W PCIE 8 pins = 375W. Can't 'force' more due to the onboard hardware without seriously seriously risking card death.

...

[dogie]

With regards to power throttling, there's a powertune +20% option. I believe that should let you mine without trouble, as it allows for 375W*1.2 = 450W. Furmark is indeed not a good benchmark, it is considered a 'power virus'. Even if power is a problem, slight undervolting should take care of that.

Temperatures and cooling may be a problem, but it's a matter of ambient temps and airflow. If you have an open air rig with extenders and plenty of space for the cards, I'm sure you'll be fine.

The question is more about whether it's a good idea to buy a $1000 card for mining now, instead of getting multiple 5xxx/79xx cards or an ASIC. Or just buying coins directly. The time of GPUs might be coming to an end.

The limit is the cooling system and power DELIVERY in this case [same as you'd have on a high end SINGLE 7970 with a little more beef]. It doesn't matter if you tell the card its allowed more than designed as in this case its hard limited by the power delivery system. 75W PCIE and 2x150W PCIE 8 pins = 375W. Can't 'force' more due to the onboard hardware without seriously seriously risking card death.

...

Still not caught on that power consumption is equal or very near to heat output?

[computerparts]

With regards to power throttling, there's a powertune +20% option. I believe that should let you mine without trouble, as it allows for 375W*1.2 = 450W. Furmark is indeed not a good benchmark, it is considered a 'power virus'. Even if power is a problem, slight undervolting should take care of that.

Temperatures and cooling may be a problem, but it's a matter of ambient temps and airflow. If you have an open air rig with extenders and plenty of space for the cards, I'm sure you'll be fine.

The question is more about whether it's a good idea to buy a $1000 card for mining now, instead of getting multiple 5xxx/79xx cards or an ASIC. Or just buying coins directly. The time of GPUs might be coming to an end.

The limit is the cooling system and power DELIVERY in this case [same as you'd have on a high end SINGLE 7970 with a little more beef]. It doesn't matter if you tell the card its allowed more than designed as in this case its hard limited by the power delivery system. 75W PCIE and 2x150W PCIE 8 pins = 375W. Can't 'force' more due to the onboard hardware without seriously seriously risking card death.

...

Still not caught on that power consumption is equal or very near to heat output?

Pci-e 2.0 is capable of delivering 150w not 75. About the power consumption yes you're right. I was thinking of something entirely different. Are you sitting there constantly refreshing the screen? I cannot fathom how you can reply so fast unless you have nothing better to do.

[dogie]

With regards to power throttling, there's a powertune +20% option. I believe that should let you mine without trouble, as it allows for 375W*1.2 = 450W. Furmark is indeed not a good benchmark, it is considered a 'power virus'. Even if power is a problem, slight undervolting should take care of that.

Temperatures and cooling may be a problem, but it's a matter of ambient temps and airflow. If you have an open air rig with extenders and plenty of space for the cards, I'm sure you'll be fine.

The question is more about whether it's a good idea to buy a $1000 card for mining now, instead of getting multiple 5xxx/79xx cards or an ASIC. Or just buying coins directly. The time of GPUs might be coming to an end.

The limit is the cooling system and power DELIVERY in this case [same as you'd have on a high end SINGLE 7970 with a little more beef]. It doesn't matter if you tell the card its allowed more than designed as in this case its hard limited by the power delivery system. 75W PCIE and 2x150W PCIE 8 pins = 375W. Can't 'force' more due to the onboard hardware without seriously seriously risking card death.

...

Still not caught on that power consumption is equal or very near to heat output?

Pci-e 2.0 is capable of delivering 150w not 75. About the power consumption yes you're right. I was thinking of something entirely different. Are you sitting there constantly refreshing the screen? I cannot fathom how you can reply so fast unless you have nothing better to do.

Look at the time stamps, same goes for you. I have 42 tabs open while im writing technical reports and they auto refresh every 20 minutes. Obviously you call 'tryhard' when you finally realise you were wrong, rather than telling me I 'need some lessons' or 'get your facts right'.

PCI-E 2.0 is capable of 75W, as I've said time and time again.



This is from electrical specification document by PCI-SIG for PCI-E 2.0: http://www.doc88.com/p-90922403627.html. When you've finished being demolished, feel free to leave this thread finally.

[Spendulus]

By that logic there is no "work", only "heat".  And that's simply not true.  The useful work performed here is in moving heat from one place to another through convection.  This is not complicated.

Next we'll be arguing that there's no useful work in the entire universe, only heat....

In the use of a fan, electrical energy is converted to work which is very quickly and consistently turned to heat. Why do you think a fan makes a room hotter? Because it is converted electricity to heat. -anyway, this is fans.

In electronics there is zero work as the cycle happens so fast. Apart from the tiny portion of energy required to 'start' the cycle, all the conversion is from electricity to heat.

Again, in case I was not clear. 

By your logic the Saturn V "did no work".

But there is a more profound error you make, which is in asserting that a fan, in moving heat from one place to another, "does no work".  That's wrong, of course it does work.   In fact, that's the very definition of work in a thermodynamic sense.

I replied to this conversation because it seemed interesting, but if you continue to stand on ignorance, this will get dull quickly.

[tacotime]

My 7970s at 901 MHz core / 1512 MHz RAM / 937 mV core pull about 170w AC (140w DC) per card while getting 630 KH/s.  This is without touching the power limiter.  I seriously doubt Malta will have a problem mining litecoins.  The 2x 8-pins themselves can give enough power to run the card at low clocks/volts, let alone the PCI-e slot.  The reference 7970 I have, by the way, runs at <60C load next a bunch of 6950s that run at 85C load.  Tahiti is insanely energy efficient when it comes to scrypt.

[Bitsaurus]

My 7970s at 901 MHz core / 1512 MHz RAM / 937 mV core pull about 170w AC (140w DC) per card while getting 630 KH/s.  This is without touching the power limiter.  I seriously doubt Malta will have a problem mining litecoins.  The 2x 8-pins themselves can give enough power to run the card at low clocks/volts, let alone the PCI-e slot.  The reference 7970 I have, by the way, runs at <60C load next a bunch of 6950s that run at 85C load.  Tahiti is insanely energy efficient when it comes to scrypt.

Eh aren't you mining in Siberia though?

Those temps look good but every single one of my cards goes over 80C when doing scrypt, even when undervolting the core.

[dogie]

By that logic there is no "work", only "heat".  And that's simply not true.  The useful work performed here is in moving heat from one place to another through convection.  This is not complicated.

Next we'll be arguing that there's no useful work in the entire universe, only heat....

In the use of a fan, electrical energy is converted to work which is very quickly and consistently turned to heat. Why do you think a fan makes a room hotter? Because it is converted electricity to heat. -anyway, this is fans.

In electronics there is zero work as the cycle happens so fast. Apart from the tiny portion of energy required to 'start' the cycle, all the conversion is from electricity to heat.

Again, in case I was not clear. 

By your logic the Saturn V "did no work".

But there is a more profound error you make, which is in asserting that a fan, in moving heat from one place to another, "does no work".  That's wrong, of course it does work.   In fact, that's the very definition of work in a thermodynamic sense.

I replied to this conversation because it seemed interesting, but if you continue to stand on ignorance, this will get dull quickly.

What I was trying to point out is there is a difference between temporary and permanent work. Compare lifting a weight on a ratchet and our fan.

While lifting the weight, work is done and stored in PE. When you stop putting in work, it stays and there is a clear and stored outcome of the work.

With the fan, as soon as we stop the fan, everything has returned to pre-fan conditions. The air is a fluid so regardless of how you move it, it will still settle to a [now hotter] equilibrium and your work was converted entirely to heat.

In the case of your rocket, it clearly did permanent work as maybe 5% of the energy consumed is stored as PE above the earth. The rest was consumed as heat.

[dogie]

My 7970s at 901 MHz core / 1512 MHz RAM / 937 mV core pull about 170w AC (140w DC) per card while getting 630 KH/s.  This is without touching the power limiter.  I seriously doubt Malta will have a problem mining litecoins.  The 2x 8-pins themselves can give enough power to run the card at low clocks/volts, let alone the PCI-e slot.  The reference 7970 I have, by the way, runs at <60C load next a bunch of 6950s that run at 85C load.  Tahiti is insanely energy efficient when it comes to scrypt.

I believe there is a difference between being FORCED to mine on low clocks and having the option to mine at low clocks. This is after all a $1k card.

[computerparts]

With regards to power throttling, there's a powertune +20% option. I believe that should let you mine without trouble, as it allows for 375W*1.2 = 450W. Furmark is indeed not a good benchmark, it is considered a 'power virus'. Even if power is a problem, slight undervolting should take care of that.

Temperatures and cooling may be a problem, but it's a matter of ambient temps and airflow. If you have an open air rig with extenders and plenty of space for the cards, I'm sure you'll be fine.

The question is more about whether it's a good idea to buy a $1000 card for mining now, instead of getting multiple 5xxx/79xx cards or an ASIC. Or just buying coins directly. The time of GPUs might be coming to an end.

The limit is the cooling system and power DELIVERY in this case [same as you'd have on a high end SINGLE 7970 with a little more beef]. It doesn't matter if you tell the card its allowed more than designed as in this case its hard limited by the power delivery system. 75W PCIE and 2x150W PCIE 8 pins = 375W. Can't 'force' more due to the onboard hardware without seriously seriously risking card death.

...

Still not caught on that power consumption is equal or very near to heat output?

Pci-e 2.0 is capable of delivering 150w not 75. About the power consumption yes you're right. I was thinking of something entirely different. Are you sitting there constantly refreshing the screen? I cannot fathom how you can reply so fast unless you have nothing better to do.

blah blah blah

Try looking for a more updated revision next time. Why in the world do you think some motherboards have molex connectors near the pci-e slots? The slot is capable of delivering 150w. There are even bios settings on some boards that allow you to change from 75w to 150w. It would be impossible to overclock some cards without having that extra power from the slot.

[dogie]

With regards to power throttling, there's a powertune +20% option. I believe that should let you mine without trouble, as it allows for 375W*1.2 = 450W. Furmark is indeed not a good benchmark, it is considered a 'power virus'. Even if power is a problem, slight undervolting should take care of that.

Temperatures and cooling may be a problem, but it's a matter of ambient temps and airflow. If you have an open air rig with extenders and plenty of space for the cards, I'm sure you'll be fine.

The question is more about whether it's a good idea to buy a $1000 card for mining now, instead of getting multiple 5xxx/79xx cards or an ASIC. Or just buying coins directly. The time of GPUs might be coming to an end.

The limit is the cooling system and power DELIVERY in this case [same as you'd have on a high end SINGLE 7970 with a little more beef]. It doesn't matter if you tell the card its allowed more than designed as in this case its hard limited by the power delivery system. 75W PCIE and 2x150W PCIE 8 pins = 375W. Can't 'force' more due to the onboard hardware without seriously seriously risking card death.

...

Still not caught on that power consumption is equal or very near to heat output?

Pci-e 2.0 is capable of delivering 150w not 75. About the power consumption yes you're right. I was thinking of something entirely different. Are you sitting there constantly refreshing the screen? I cannot fathom how you can reply so fast unless you have nothing better to do.

blah blah blah

Try looking for a more updated revision next time. Why in the world do you think some motherboards have molex connectors near the pci-e slots? The slot is capable of delivering 150w. There are even bios settings on some boards that allow you to change from 75w to 150w. It would be impossible to overclock some cards without having that extra power from the slot.

You asked for the PCI-E 2.0 specification, I gave it you. If you want to argue they made the specification wrong then go argue with them - your insolence is incredible.

[Spendulus]

By that logic there is no "work", only "heat".  And that's simply not true.  The useful work performed here is in moving heat from one place to another through convection.  This is not complicated.

Next we'll be arguing that there's no useful work in the entire universe, only heat....

In the use of a fan, electrical energy is converted to work which is very quickly and consistently turned to heat. Why do you think a fan makes a room hotter? Because it is converted electricity to heat. -anyway, this is fans.

In electronics there is zero work as the cycle happens so fast. Apart from the tiny portion of energy required to 'start' the cycle, all the conversion is from electricity to heat.

Again, in case I was not clear.  

By your logic the Saturn V "did no work".

But there is a more profound error you make, which is in asserting that a fan, in moving heat from one place to another, "does no work".  That's wrong, of course it does work.   In fact, that's the very definition of work in a thermodynamic sense.

I replied to this conversation because it seemed interesting, but if you continue to stand on ignorance, this will get dull quickly.

What I was trying to point out is there is a difference between temporary and permanent work. Compare lifting a weight on a ratchet and our fan.

While lifting the weight, work is done and stored in PE. When you stop putting in work, it stays and there is a clear and stored outcome of the work.

With the fan, as soon as we stop the fan, everything has returned to pre-fan conditions. The air is a fluid so regardless of how you move it, it will still settle to a [now hotter] equilibrium and your work was converted entirely to heat.

In the case of your rocket, it clearly did permanent work as maybe 5% of the energy consumed is stored as PE above the earth. The rest was consumed as heat.

Well, since you want to look at the rocket in that fashion, I also, will use your method.  By way of the virial theorum, the additional heat I added to the atmosphere by way of my fan has caused the gas envelope of the planet to become hotter, and thus it has expanded - admittedly, quite slightly.  We can show that of this expansion, 50% is a potential energy increase, and 50% is a kinetic energy increase.

You would like, it would seem, to have dictatorial control of the definition of "work".  But neither in engineering or physics is that plausible.  It is quite strictly defined.

[computerparts]

With regards to power throttling, there's a powertune +20% option. I believe that should let you mine without trouble, as it allows for 375W*1.2 = 450W. Furmark is indeed not a good benchmark, it is considered a 'power virus'. Even if power is a problem, slight undervolting should take care of that.

Temperatures and cooling may be a problem, but it's a matter of ambient temps and airflow. If you have an open air rig with extenders and plenty of space for the cards, I'm sure you'll be fine.

The question is more about whether it's a good idea to buy a $1000 card for mining now, instead of getting multiple 5xxx/79xx cards or an ASIC. Or just buying coins directly. The time of GPUs might be coming to an end.

The limit is the cooling system and power DELIVERY in this case [same as you'd have on a high end SINGLE 7970 with a little more beef]. It doesn't matter if you tell the card its allowed more than designed as in this case its hard limited by the power delivery system. 75W PCIE and 2x150W PCIE 8 pins = 375W. Can't 'force' more due to the onboard hardware without seriously seriously risking card death.

...

Still not caught on that power consumption is equal or very near to heat output?

Pci-e 2.0 is capable of delivering 150w not 75. About the power consumption yes you're right. I was thinking of something entirely different. Are you sitting there constantly refreshing the screen? I cannot fathom how you can reply so fast unless you have nothing better to do.

blah blah blah

Try looking for a more updated revision next time. Why in the world do you think some motherboards have molex connectors near the pci-e slots? The slot is capable of delivering 150w. There are even bios settings on some boards that allow you to change from 75w to 150w. It would be impossible to overclock some cards without having that extra power from the slot.

You asked for the PCI-E 2.0 specification, I gave it you. If you want to argue they made the specification wrong then go argue with them - your insolence is incredible.

I never asked for the pci-e 2.0 spec. I stated a fact and it remains fact that the pci-e 2.0 slot is capable of delivering 150 watts. Your ignorance must be bliss.

[dogie]

By that logic there is no "work", only "heat".  And that's simply not true.  The useful work performed here is in moving heat from one place to another through convection.  This is not complicated.

Next we'll be arguing that there's no useful work in the entire universe, only heat....

In the use of a fan, electrical energy is converted to work which is very quickly and consistently turned to heat. Why do you think a fan makes a room hotter? Because it is converted electricity to heat. -anyway, this is fans.

In electronics there is zero work as the cycle happens so fast. Apart from the tiny portion of energy required to 'start' the cycle, all the conversion is from electricity to heat.

Again, in case I was not clear. 

By your logic the Saturn V "did no work".

But there is a more profound error you make, which is in asserting that a fan, in moving heat from one place to another, "does no work".  That's wrong, of course it does work.   In fact, that's the very definition of work in a thermodynamic sense.

I replied to this conversation because it seemed interesting, but if you continue to stand on ignorance, this will get dull quickly.

What I was trying to point out is there is a difference between temporary and permanent work. Compare lifting a weight on a ratchet and our fan.

While lifting the weight, work is done and stored in PE. When you stop putting in work, it stays and there is a clear and stored outcome of the work.

With the fan, as soon as we stop the fan, everything has returned to pre-fan conditions. The air is a fluid so regardless of how you move it, it will still settle to a [now hotter] equilibrium and your work was converted entirely to heat.

In the case of your rocket, it clearly did permanent work as maybe 5% of the energy consumed is stored as PE above the earth. The rest was consumed as heat.

Well, since you want to look at the rocket in that fashion, I also, will use your method.  By way of the virial theorum, the additional heat I added to the atmosphere by way of my fan has caused the gas envelope of the planet to become hotter, and thus it has expanded - admittedly, quite slightly.  We can show that of this expansion, 50% is a potential energy increase, and 50% is a kinetic energy increase.

When you look at it in more detail it doesn't work that way. The air will give up its 'additional' heat by convection and interfluid diffusivity to EVERYTHING else, to the point the 'gas envelope' is unchanged.

[dogie]

I never asked for the pci-e 2.0 spec. I stated a fact and it remains fact that the pci-e 2.0 slot is capable of delivering 150 watts. Your ignorance must be bliss.

Pci-e 2.0 is capable of delivering 150w not 75.

What you've suggested is the standard is capable of delivering 150W. So if I took any PCI-E 2.0 motherboard off the shelf, I could power a 150W graphics card from it. You can't, the standard isn't and you're wrong.

You can't say the slot can supply 150W if it can't. Why do you think the Malta 7990 is 375W? 2x150W PCI-E 8 pin connectors = 300W.

375W - 300W does not equal 150W.

[computerparts]

I never asked for the pci-e 2.0 spec. I stated a fact and it remains fact that the pci-e 2.0 slot is capable of delivering 150 watts. Your ignorance must be bliss.

Pci-e 2.0 is capable of delivering 150w not 75.

What you've suggested is the standard is capable of delivering 150W. So if I took any PCI-E 2.0 motherboard off the shelf, I could power a 150W graphics card from it. You can't, the standard isn't and you're wrong.

You can't say the slot can supply 150W if it can't. Why do you think the Malta 7990 is 375W? 2x150W PCI-E 8 pin connectors = 300W.

375W - 300W does not equal 150W.

Here we go again, you're forgetting TDP is not max capable power draw. The specified TDP is only what the stock heatsink is required to dissipate. What do you think happens when you overvolt and overclock? Same goes for cpus. An i5-3570k has a tdp of 77 watts but overclock it and power consumption can easily double that number.

[dogie]

I never asked for the pci-e 2.0 spec. I stated a fact and it remains fact that the pci-e 2.0 slot is capable of delivering 150 watts. Your ignorance must be bliss.

Pci-e 2.0 is capable of delivering 150w not 75.

What you've suggested is the standard is capable of delivering 150W. So if I took any PCI-E 2.0 motherboard off the shelf, I could power a 150W graphics card from it. You can't, the standard isn't and you're wrong.

You can't say the slot can supply 150W if it can't. Why do you think the Malta 7990 is 375W? 2x150W PCI-E 8 pin connectors = 300W.

375W - 300W does not equal 150W.

Here we go again, you're forgetting TDP is not max capable power draw. The specified TDP is only what the stock heatsink is required to dissipate. What do you think happens when you overvolt and overclock? Same goes for cpus. An i5-3570k has a tdp of 77 watts but overclock it and power consumption can easily double that number.

No shit its not, but those 2 numbers happen to line up in this particular case.

[Spendulus]

By that logic there is no "work", only "heat".  And that's simply not true.  The useful work performed here is in moving heat from one place to another through convection.  This is not complicated.

Next we'll be arguing that there's no useful work in the entire universe, only heat....

In the use of a fan, electrical energy is converted to work which is very quickly and consistently turned to heat. Why do you think a fan makes a room hotter? Because it is converted electricity to heat. -anyway, this is fans.

In electronics there is zero work as the cycle happens so fast. Apart from the tiny portion of energy required to 'start' the cycle, all the conversion is from electricity to heat.

Again, in case I was not clear. 

By your logic the Saturn V "did no work".

But there is a more profound error you make, which is in asserting that a fan, in moving heat from one place to another, "does no work".  That's wrong, of course it does work.   In fact, that's the very definition of work in a thermodynamic sense.

I replied to this conversation because it seemed interesting, but if you continue to stand on ignorance, this will get dull quickly.

What I was trying to point out is there is a difference between temporary and permanent work. Compare lifting a weight on a ratchet and our fan.

While lifting the weight, work is done and stored in PE. When you stop putting in work, it stays and there is a clear and stored outcome of the work.

With the fan, as soon as we stop the fan, everything has returned to pre-fan conditions. The air is a fluid so regardless of how you move it, it will still settle to a [now hotter] equilibrium and your work was converted entirely to heat.

In the case of your rocket, it clearly did permanent work as maybe 5% of the energy consumed is stored as PE above the earth. The rest was consumed as heat.

Well, since you want to look at the rocket in that fashion, I also, will use your method.  By way of the virial theorum, the additional heat I added to the atmosphere by way of my fan has caused the gas envelope of the planet to become hotter, and thus it has expanded - admittedly, quite slightly.  We can show that of this expansion, 50% is a potential energy increase, and 50% is a kinetic energy increase.

When you look at it in more detail it doesn't work that way. The air will give up its 'additional' heat by convection and interfluid diffusivity to EVERYTHING else, to the point the 'gas envelope' is unchanged.

Actually, it does work that way.  Reference 'virial theorum'.  I need say no more.  The hot air of which you spoke, ROSE UP.  It did not impart it's energy to the earth or the water, but even were that the case, in turn those reservoirs over time would release their excess energy, as dictated by the upper stratospheric radiative balance.  Heat in and out from the sphere.  Still, the virial theorum rules.

[dogie]

Actually, it does work that way.  Reference 'virial theorum'.  I need say no more.  The hot air of which you spoke, ROSE UP.  It did not impart it's energy to the earth or the water, but even were that the case, in turn those reservoirs over time would release their excess energy, as dictated by the upper stratospheric radiative balance.  Heat in and out from the sphere.  Still, the virial theorum rules.

Well you're right it would eventually be radiated as with all heat, but do we exactly care what happens to the heat once its been created?

Its still elec -> ke -> heat.

[Spendulus]

Actually, it does work that way.  Reference 'virial theorum'.  I need say no more.  The hot air of which you spoke, ROSE UP.  It did not impart it's energy to the earth or the water, but even were that the case, in turn those reservoirs over time would release their excess energy, as dictated by the upper stratospheric radiative balance.  Heat in and out from the sphere.  Still, the virial theorum rules.

Well you're right it would eventually be radiated as with all heat, but do we exactly care what happens to the heat once its been created?

Its still elec -> ke -> heat.

No, it would be elec -> 1/2 ke + 1/2 pe

And then you have arguably some secondary effect (or not) on the entire planet's radiative balance at the top of the stratosphere, that having moved up and having a slightly larger surface area for those effects.

The point remains, you can't just redefine work and claim all is entropy.  The definition of thermodynamic work is as it is.  Deal with it.

[2Kool4Skewl]

I'd also be interested in seeing what these ref 7990s can hash.  But, I think they'll be oc limited by the power connectors on the cards and the heat.  The board itself will conduct/retain heat between the two chips.  At least with two 7970 cards you can physically separate the dies.

[dogie]

I'd also be interested in seeing what these ref 7990s can hash.  But, I think they'll be oc limited by the power connectors on the cards and the heat.  The board itself will conduct/retain heat between the two chips.  At least with two 7970 cards you can physically separate the dies.

I can't get above 1100 24/7 stable without going above stock voltage on my PC boards. Doubt the Maltas will be any better. = 1250-1300 a board.

[Miner99er]

When Mining SHA256, Memory clocks are reduced and the GPU cores are Ran like a bat out of hell...

When Mining Scrypt, the Clocks on both the CPU and memory are ran at a ratio of 0.57 to 1.

Either way, you're not pulling the max power draw the card is capable of.

Taco time stated 170w per 7970 while mining on unbinned Tahiti chips.

So 340 for this card, and add 15w for the crossbridge, 365w.

Close but under the limit.

[dogie]

When Mining SHA256, Memory clocks are reduced and the GPU cores are Ran like a bat out of hell...

When Mining Scrypt, the Clocks on both the CPU and memory are ran at a ratio of 0.57 to 1.

Either way, you're not pulling the max power draw the card is capable of.

Taco time stated 170w per 7970 while mining on unbinned Tahiti chips.

So 340 for this card, and add 15w for the crossbridge, 365w.

Close but under the limit.

...why are you trying to guess numbers when I have actual power consumption? 400 a card with minimum memory clocks and voltages, slightly undervolted cores. Thats on SHA256

[Miner99er]

When Mining SHA256, Memory clocks are reduced and the GPU cores are Ran like a bat out of hell...

When Mining Scrypt, the Clocks on both the CPU and memory are ran at a ratio of 0.57 to 1.

Either way, you're not pulling the max power draw the card is capable of.

Taco time stated 170w per 7970 while mining on unbinned Tahiti chips.

So 340 for this card, and add 15w for the crossbridge, 365w.

Close but under the limit.

...why are you trying to guess numbers when I have actual power consumption? 400 a card with minimum memory clocks and voltages, slightly undervolted cores. Thats on SHA256

Because you don't have a Malta 7990.

[dogie]

When Mining SHA256, Memory clocks are reduced and the GPU cores are Ran like a bat out of hell...

When Mining Scrypt, the Clocks on both the CPU and memory are ran at a ratio of 0.57 to 1.

Either way, you're not pulling the max power draw the card is capable of.

Taco time stated 170w per 7970 while mining on unbinned Tahiti chips.

So 340 for this card, and add 15w for the crossbridge, 365w.

Close but under the limit.

...why are you trying to guess numbers when I have actual power consumption? 400 a card with minimum memory clocks and voltages, slightly undervolted cores. Thats on SHA256

Because you don't have a Malta 7990.

-_- the cores are identical?

[goxed]

http://www.newegg.com/Product/Product.aspx?Item=N82E16814202036

[relm9]

http://www.newegg.com/Product/Product.aspx?Item=N82E16814202036

$50 above MSRP? Way to go Newegg...

[dogie]

http://www.newegg.com/Product/Product.aspx?Item=N82E16814202036

$50 above MSRP? Way to go Newegg...

I think its probably the distributer charging a pretty penny for the first batch, where demand is guaranteed to outstrip demand.

[Dyaheon]

-_- the cores are identical?

Not exactly. The new Maltas have the lowest consuming cores available, AMD is doing some serious binning with them.

[cdog]

Apparently they are terrible for gaming because of some random screen hitching which is visible but not detectable by FRAPs, hasnt been fixed with latest drivers, so until thats fixed the resale value for these is borked, useless for gaming, and I cant see why you would buy one for mining as 7950s still offer MUCH better value.

[dogie]

-_- the cores are identical?

Not exactly. The new Maltas have the lowest consuming cores available, AMD is doing some serious binning with them.

Which made ~5% difference according to the reviews.

Apparently they are terrible for gaming because of some random screen hitching which is visible but not detectable by FRAPs, hasnt been fixed with latest drivers, so until thats fixed the resale value for these is borked, useless for gaming, and I cant see why you would buy one for mining as 7950s still offer MUCH better value.

It was fixed with the prototype drivers, which may have been 13.5 pre beta.

[phillipsjk]

Here we go again, you're forgetting TDP is not max capable power draw. The specified TDP is only what the stock heatsink is required to dissipate. What do you think happens when you overvolt and overclock? Same goes for cpus. An i5-3570k has a tdp of 77 watts but overclock it and power consumption can easily double that number.

Somebody is wrong on the Internet.

For electronics, TDP is the same as power draw, to the point that it is sometimes easier to measure the change in temperature than the actual power draw. When you overclock or over-volt a chip, you change the TDP (hopefully not in a way you can't compensate for: such as using a lower ambient temperature).

Molex connectors next to PCIe slots are not part of the slot. Those connectors are likely there becuase the slot connector can not supply more than 75W safely, as stated by the spec quoted by dogie. But what do I know? I wash dishes for a living.

PS: RE: your claim 375Watt TDP equals ~500Watts of power draw. I have 2 possible theories of where you got your information:
1. Online Power supply calculators grossly overestimate the required power. They either want to sell you a large PSU, or assume the PSU you purchase will be sub-standard.
2. Volt-Amps (at the wall) is NOT the same as Watts. For non Power-factor Corrected power supplies, you "waste" about 1/3 of the current (assuming power factor of .67, if my math is right). This wasted current causes resitive heating in the wiring, but otherwise does not draw any power. This happens when the current peaks do not line up with the voltage peaks in the power cycle. It means you get current flow at the zero voltage crossing. (Power=(current)(voltage), by example  (5A)(0V)=0W)

[dogie]

Here we go again, you're forgetting TDP is not max capable power draw. The specified TDP is only what the stock heatsink is required to dissipate. What do you think happens when you overvolt and overclock? Same goes for cpus. An i5-3570k has a tdp of 77 watts but overclock it and power consumption can easily double that number.

Somebody is wrong on the Internet.

For electronics, TDP is the same as power draw, to the point that it is sometimes easier to measure the change in temperature than the actual power draw. When you overclock or over-volt a chip, you change the TDP (hopefully not in a way you can't compensate for: such as using a lower ambient temperature).

Molex connectors next to PCIe slots are not part of the slot. Those connectors are likely there becuase the slot connector can not supply more than 75W safely, as stated by the spec quoted by dogie. But what do I know? I wash dishes for a living.

PS: RE: your claim 375Watt TDP equals ~500Watts of power draw. I have 2 possible theories of where you got your information:
1. Online Power supply calculators grossly overestimate the required power. They either want to sell you a large PSU, or assume the PSU you purchase will be sub-standard.
2. Volt-Amps (at the wall) is NOT the same as Watts. For non Power-factor Corrected power supplies, you "waste" about 1/3 of the current (assuming power factor of .67, if my math is right). This wasted current causes resitive heating in the wiring, but otherwise does not draw any power. This happens when the current peaks do not line up with the voltage peaks in the power cycle. It means you get current flow at the zero voltage crossing. (Power=(current)(voltage), by example  (5A)(0V)=0W)

^_^ Thank you, finally someone with some sense!

[rograz]

It was fixed with the prototype drivers, which may have been 13.5 pre beta.

It was improved in a lot of titles but far from all, and it's still a lot worse than SLI. Crossfire is for all intent and purposes still quite useless (why bother when 2 gtx660 TI gives a better experience than 2x7970s). Best case atm you are looking at 50% "perceived" higher FPS with crossfire vs one card while the benched FPS scaling is in the 90%+ range. AMD have at least acknowledged the problem for some time and in some popular titles they have almost eliminated the problem as long as you use vsync or cap your fps.

The pre-beta drivers simply put the 7990 on the same level as 2 7970s in crossfire, they didn't solve the micro stutter problem completely, yet. AMD spent to long trying to get crossfire scaling as high as possible while Nvidia seems to have focused more on actually making SLI usable, it wasn't until a few large tech sites started bashing AMD for the poor experience with crossfire last year that they started looking at it.

[dogie]

It was fixed with the prototype drivers, which may have been 13.5 pre beta.

The pre-beta drivers simply put the 7990 on the same level as 2 7970s in crossfire

A 7990 is 2 7990s

[Dyaheon]

So anyone have one of these new 7990s yet? Kinda interested in some mining results.

Sadly the price seems really steep, quite a bit more than the older, beefier 7990s with triple PCI-e power connectors and triple slot coolers. Not going to be worth it if they're mediocre mining cards.

[greaterninja]

So anyone have one of these new 7990s yet? Kinda interested in some mining results.

Sadly the price seems really steep, quite a bit more than the older, beefier 7990s with triple PCI-e power connectors and triple slot coolers. Not going to be worth it if they're mediocre mining cards.

I am getting 2 x gigabyte 7990s tomorrow, I'll test one out and let u guys know.   So far the 7790, 7870 xt and 7950 cards seem to be the best cards for the $.

[bobsmoke]

with a dual gpu 7990 club 3d if you have enough cooling you should get around 1.2Ghz....of course with a lot of noise due to the fans that it brings.

[dogie]

Continuing greaterninja's thread because he cried and locked it: https://109.201.133.65/index.php?topic=200628
He left a lot of misnomers which should not be continued by the community.

As I recall you were saying the card would use over 500 watts, in reality it uses ~375 watts at max load and it still performs as well as the 1st generation 7990 but uses 2/3 the space of the first generation 7990.  

1) No I didnt, I said it will be power limited. If you're maxing out at STOCK CLOCKS and slightly undervolted, on bitcoins, have fun scypting.
2) Same performance? Nope, my cards are 1300+, yours are 1150.
3) 2/3 the space? Great, now try and use the card.

You also said it would cost $2000 dollars when it costs $1000-$1100.

For two cards *facepalm*

...the only thing you were maybe right about is the cooling and that is still for dispute as you use many more fans than I use to ventilate the card.   So all in all you weren't right about much.  You more so blurted out a bunch of nonsense and acted like a huge forum troll...good try at trying to be right.

Many more fans? I am using TWO of these cards, TWO. So yes, I do have 2 more fans than you because im generating another 400W+ of heat in the same space.

tldr: I warned you, you spent $3.3k on 3 cards and are now trying to sell them because you realized I was right. Enjoy your financial loss because you were so arrogant.

[greaterninja]

Continuing greaterninja's thread because he cried and locked it: https://109.201.133.65/index.php?topic=200628
He left a lot of misnomers which should not be continued by the community.

As I recall you were saying the card would use over 500 watts, in reality it uses ~375 watts at max load and it still performs as well as the 1st generation 7990 but uses 2/3 the space of the first generation 7990.  

1) No I didnt, I said it will be power limited. If you're maxing out at STOCK CLOCKS and slightly undervolted, on bitcoins, have fun scypting.
2) Same performance? Nope, my cards are 1300+, yours are 1150.
3) 2/3 the space? Great, now try and use the card.

You also said it would cost $2000 dollars when it costs $1000-$1100.

For two cards *facepalm*

...the only thing you were maybe right about is the cooling and that is still for dispute as you use many more fans than I use to ventilate the card.   So all in all you weren't right about much.  You more so blurted out a bunch of nonsense and acted like a huge forum troll...good try at trying to be right.

Many more fans? I am using TWO of these cards, TWO. So yes, I do have 2 more fans than you because im generating another 400W+ of heat in the same space.

tldr: I warned you, you spent $3.3k on 3 cards and are now trying to sell them because you realized I was right. Enjoy your financial loss because you were so arrogant.

Nope.  You said 1 card costs $2000, and 1 card will use over 500 watts to get the same hash power or it will hash a lot less.  The new malta uses less space and performs more efficiently for the power. The malta still hashes 1150 MH/s - 1250 mh/s easily.  Now you might be right about the temperature, however everything else you were wrong on.   Finally, I am selling each card for a significant profit, certainly not because of financial loss.     Congrats at making such a big deal out of what you really don't know or have.

[computerparts]

 

[dogie]

Continuing greaterninja's thread because he cried and locked it: https://109.201.133.65/index.php?topic=200628
He left a lot of misnomers which should not be continued by the community.

As I recall you were saying the card would use over 500 watts, in reality it uses ~375 watts at max load and it still performs as well as the 1st generation 7990 but uses 2/3 the space of the first generation 7990. 

1) No I didnt, I said it will be power limited. If you're maxing out at STOCK CLOCKS and slightly undervolted, on bitcoins, have fun scypting.
2) Same performance? Nope, my cards are 1300+, yours are 1150.
3) 2/3 the space? Great, now try and use the card.

You also said it would cost $2000 dollars when it costs $1000-$1100.

For two cards *facepalm*

...the only thing you were maybe right about is the cooling and that is still for dispute as you use many more fans than I use to ventilate the card.   So all in all you weren't right about much.  You more so blurted out a bunch of nonsense and acted like a huge forum troll...good try at trying to be right.

Many more fans? I am using TWO of these cards, TWO. So yes, I do have 2 more fans than you because im generating another 400W+ of heat in the same space.

tldr: I warned you, you spent $3.3k on 3 cards and are now trying to sell them because you realized I was right. Enjoy your financial loss because you were so arrogant.

Nope.  You said 1 card costs $2000, and 1 card will use over 500 watts to get the same hash power or it will hash a lot less.  The new malta uses less space and performs more efficiently for the power. The malta still hashes 1150 MH/s - 1250 mh/s easily.  Now you might be right about the temperature, however everything else you were wrong on.   Finally, I am selling each card for a significant profit, certainly not because of financial loss.     Congrats at making such a big deal out of what you really don't know or have.

WHY WOULD I SAY ONE CARD COSTS $2000, are you kidding me. I was always comparing dual cards. I'd ask you to quote me but I know you'll just come up with some excuse again and then lock the thread again.... OH WAIT ITS MY THREAD.

[phillyj]

Nope.  You said 1 card costs $2000, and 1 card will use over 500 watts to get the same hash power or it will hash a lot less.  

He said, she said... I hate taking sides but just post the quote where dogie said 1 card cost $2k. I looked at the first page of this thread and did not see that.

dogie probably overreacted. What does "suitable" mean anyway? It's all relative to the buyer. It's not a Nvidia vs AMD mining comparison. It's more of a power-draw vs price vs hashrate suitability discussion.

Edit: changed 1k to 2k typo

[greaterninja]

Now the new M cards have dual slot coolers, ouch. At stock clocks they were loading just below the 80s, horizontally mounted in an open air, single card system. This was on GPGPU computing. Due to the heatsink design, ZERO air is exhausted and its all left to be removed by the case.  My PC cards heat the case ambient up to 40+ even with 3x250mm, 5x240mm of fans removing heat. You will have to open air, just have to.

Tldr: Mining is doable on the 3 slot PC cards, but will be almost impossible to dual M cards within a case. The M cards are 2 slot and don't exhaust air at all.

WRONG.  mining is quite doable on the dual slot reference cards.  See next quote remarks.

Power delivery:
This is the major problem. My PC cards have 3x8pin for a max of 525W. These new M cards only have 2x8 pins for 375W max. Even with my moderate overclocks (focusing on power consumption rather than MAX POWA), I'm going over/hitting the power delivery limits of the PC cards at about ~400W from the wall per card [1050 core, mem 625 and undervolted]. AnandTech saw this in Furmark, the card was power throttled HEAVILY. Now imagine trying to scypt mine with decent clocks - its not going to happen.

WRONG. ...the M  7990s seems best at scrypt mining and for bitcoin mining they still perform well at 1150-1250mh/s   I believe they can even be pushed to 1300+mh/s  All at ~375 Watts

Tldr: The power delivery on these cards LIMITS the performance of these cards and causes power throttling even on moderate computing load. Hence these cards are not appropriate for mining.

WRONG.  this quote is where dogie is assuming stuff again, it was a good assumption, however 375w is more than enough to power both cores on the new Malta 7990.  AMD has had about 1 year to redesign power consumption of the new 7990.

WHY WOULD I SAY ONE CARD COSTS $2000, are you kidding me. I was always comparing dual cards. I'd ask you to quote me but I know you'll just come up with some excuse again and then lock the thread again.... OH WAIT ITS MY THREAD.

 
Dogie did say it costs $2000, at the time I didn't notice how bad he was trolling so I did not capture the quote.   it seems he had gone back and edited his post because everyone was thinking he was a moron.

So good job dogie, you were really right about nothing.  Next time get both pieces of the hardware, then do a unbiased test of both pieces of hardware in the same environment.   Spouting numbers and assumptions in every 7990 thread solely based on your self-qualified "masters of engineering" experience and not having the hardware has made you look like a jackass.  Sup Troll!?!?!?!

[computerparts]

I gave up arguing with him when I saw I was clearly wasting my time. Maybe now he has a clue as to the significance of higher binned chips.

[dogie]

Now the new M cards have dual slot coolers, ouch. At stock clocks they were loading just below the 80s, horizontally mounted in an open air, single card system. This was on GPGPU computing. Due to the heatsink design, ZERO air is exhausted and its all left to be removed by the case.  My PC cards heat the case ambient up to 40+ even with 3x250mm, 5x240mm of fans removing heat. You will have to open air, just have to.

Tldr: Mining is doable on the 3 slot PC cards, but will be almost impossible to dual M cards within a case. The M cards are 2 slot and don't exhaust air at all.

WRONG.  mining is quite doable on the dual slot reference cards.  See next quote remarks.

Power delivery:
This is the major problem. My PC cards have 3x8pin for a max of 525W. These new M cards only have 2x8 pins for 375W max. Even with my moderate overclocks (focusing on power consumption rather than MAX POWA), I'm going over/hitting the power delivery limits of the PC cards at about ~400W from the wall per card [1050 core, mem 625 and undervolted]. AnandTech saw this in Furmark, the card was power throttled HEAVILY. Now imagine trying to scypt mine with decent clocks - its not going to happen.

WRONG. ...the M  7990s seems best at scrypt mining and for bitcoin mining they still perform well at 1150-1250mh/s   I believe they can even be pushed to 1300+mh/s  All at ~375 Watts

Tldr: The power delivery on these cards LIMITS the performance of these cards and causes power throttling even on moderate computing load. Hence these cards are not appropriate for mining.

WRONG.  this quote is where dogie is assuming stuff again, it was a good assumption, however 375w is more than enough to power both cores on the new Malta 7990.  AMD has had about 1 year to redesign power consumption of the new 7990.

WHY WOULD I SAY ONE CARD COSTS $2000, are you kidding me. I was always comparing dual cards. I'd ask you to quote me but I know you'll just come up with some excuse again and then lock the thread again.... OH WAIT ITS MY THREAD.

 
Dogie did say it costs $2000, at the time I didn't notice how bad he was trolling so I did not capture the quote.   it seems he had gone back and edited his post because everyone was thinking he was a moron.

So good job dogie, you were really right about nothing.  Next time get both pieces of the hardware, then do a unbiased test of both pieces of hardware in the same environment.   Spouting numbers and assumptions in every 7990 thread solely based on your self-qualified "masters of engineering" experience and not having the hardware has made you look like a jackass.  Sup Troll!?!?!?!

So much bullshit in here its not even worth it any more. And I edited my post? That's convenient isn't it, I edited a post that didn't exist

[ecliptic]

They seriously make 2x 8 pin 7990 cards?  That's really cutting it close.

I've got a 3x 8 pin powercooler 7990 on (unpowered) risers running fine @ ~1.08MH/sec, haven't fined tuned it much.  2x would be very ugly

[dogie]

They seriously make 2x 8 pin 7990 cards?  That's really cutting it close.

I've got a 3x 8 pin powercooler 7990 on (unpowered) risers running fine @ ~1.08MH/sec, haven't fined tuned it much.  2x would be very ugly

+1. We both have non malta'ed 7990s and can see the truth.

[notaek]

They seriously make 2x 8 pin 7990 cards?  That's really cutting it close.

I've got a 3x 8 pin powercooler 7990 on (unpowered) risers running fine @ ~1.08MH/sec, haven't fined tuned it much.  2x would be very ugly

Haha the funny thing is that the "very ugly" 2x8pin 7990 malta gets higher hash rates than yours does.  

[relm9]

I picked up one of these for my gaming rig yesterday with no real intention of mining on it (live in AZ and it's way too hot here during the summer) but I tried anyway to get some results.

Ran for 2 hours. With scrypt mining, straight out of the box using the same config I used for 7950s I was getting 1.3 MH/s, this is at 1000 MHz core and 1500 memory, all defaults - 0% PowerTune. Completely stable. I don't know if the pool I use is running better or what but I was getting less than 0.05% rejects on shares which is impressive, my 7950s often hovered around 2%.

Temps were somewhat high, around 86-87 C though bare in mind the room temp was 81 F and I was running it inside of a case.

Interestingly, if I knocked PowerTune down to -20% I still was able to get 1.05 MH/s from the card! This surprised me. With PowerTune at -20% temps hovered around 78C.

I'd imagine these figures would be a lot more reasonable at a lower room temp.

Price-wise, the card isn't a great value, if you sel' the games you're looking at $850 out of pocket, where as you could probably get 2x7970 for around $700 after selling the games, but it is an option if you are low on slots. Also resale value may hold better if AMD's previous dual GPU cards are any indication.

Didn't get to play around with the voltage to see how low I could go, seems MSI Afterburner doesn't support it on these yet?

Pic:

[greaterninja]

mine is voltage unlocked, you have to have the ati.dll to do extreme overclocking...its in another thread.

the dll is atipdlxx.dll  and you ahve to put it in   C:\Program Files (x86)\MSI Afterburner

and you have to modify MSIAfterburner.cfg in that very same folder.   delete everything in the profiles sub-directory.

[relm9]

mine is voltage unlocked, you have to have the ati.dll to do extreme overclocking...its in another thread.

the dll is atipdlxx.dll  and you ahve to put it in   C:\Program Files (x86)\MSI Afterburner

and you have to modify MSIAfterburner.cfg in that very same folder.   delete everything in the profiles sub-directory.

Ah thanks, I'll try this

[greaterninja]

mine is voltage unlocked, you have to have the ati.dll to do extreme overclocking...its in another thread.

the dll is atipdlxx.dll  and you ahve to put it in   C:\Program Files (x86)\MSI Afterburner

and you have to modify MSIAfterburner.cfg in that very same folder.   delete everything in the profiles sub-directory.

Ah thanks, I'll try this

at stock clock 1000mhz gpu and 1500mhz ram, 0% volt i hit 1370+ khash/sec, 20k+ thread concurrency, 20 aggression.


i found it best to turn aggression or thread concurrency down a bit; the card runs nice and cool ~1200khash/sec

Right now its hard to say if I want to mine BTC or LTC...I've figured out how to mine BTC in the 70C range doing easily 1200+mh/s;  undervolted -20%.

[Miner99er]

So three of these could fit on an 890FXA-GD70 with space between each so they don't cook themselves. Sounds like a done deal, and much easier to setup/configure/route cables with 6 individual cards.

not that I'm done with buying Single GPU cards, I'm not having density issues yet.

[BBQKorv]

Very nice results. This card seems very impressive in comparison to the 3-slot 7990s. It's nearly unbelievable they managed to get the cooling to work that well on it.

Do you have any VRM noises at any clocks or volts?

[greaterninja]

Very nice results. This card seems very impressive in comparison to the 3-slot 7990s. It's nearly unbelievable they managed to get the cooling to work that well on it.

Do you have any VRM noises at any clocks or volts?

Yep dogie was a complete moron for not physically comparing both cards.  He rather created a thread based on speculation of the old 7990 hardware that is 1+ year old and now he looks like a jackass.  Now hes mad to the point that he is trolling people's threads like a degenerate.   (degens gonna degen.)

As of this post the Sapphire 7990 Malta seems to be the most stable card as far as temps, clock, power usage, etc.

I can get 1260mhash/sec stable; even up to 1300+mh/sec while bitcoin mining; using msi afterburner; stock cooling only.
ROI on this card is 5-6.2 months.

I'm sure I can sell it for a pretty penny after mining a bit of btc as well.

With it all said the 7870 tahiti LE, 7950, 7970, 5870 5970 cards seem to be the best for ROI; this card does OK in that realm.

[notaek]

Do the fans look that pink in person, or just on camera?

[relm9]

Do the fans look that pink in person, or just on camera?

Not as bright, more of a light-red. Camera flash oversaturated it.

I haven't heard any VRM noises or coil whine, the card is whisper quiet until >50 % fan speed, only really gets loud at 70 % or above and even then it's not bad compared to other cards I've owned.

[ISAWHIM]

The "extra" pins are both ground... (-12v). They are there for "legal reasons", no other real reason. 6-pin = 3x+12, 3x-12 [8-pin = 3x+12, 5x-12v] The "theory", is that if you don't have a 6+2 pin connection, then your PSU MAY not be designed to handle the power. The reality is, connectors has nothing to do with "available power". Some PSU's have 25A, some have only 14A, some have up to 75A available on those same pins, or more, or less.

You also have to "know" what your computer draws, before adding the card to the system. When the CPU and RAM is "loaded" by 80%, as that is what will give you the "base-line" to subtract from your "wall wattage" of a "running and loaded" video-card. (You can't just "not use the card", because if it is plugged-in, it is consuming power, and that is NOT the baseline. If you subtract that value from the loaded GPU tests, you will wrongly be measuring the GPU consumption.)

Also, your wall-wattage is the consumption of the whole system, including all inefficiencies. Unless you have a "platinum 95%" running on 240v, then assume you are showing 20-30% more power use at the wall, than the system is actually using inside. Most PSU's are about 75-85%, at best, running on a 120v power line. Even platinum-95's.

But I digress... saying it is NOT good, "in your crappy setup", should be elaborated to state...

"In my crappy closed-case, poorly vented, poorly exhausted, under-powered, and horrible unmanaged card-setup settings... The card was not good for mining."

Yes, the design of the card requires special care, as you mentioned. It is power-hungry... but you never even stated a power/hash rate, nor did you mention any actual adjustments to "control power", you just poured-in what your PSU fed it. You failed to mention PSU model, rail-power, or your system-consumption... Thus, I assume you have a shoddy PSU and severely way over-powered CPU that is sucking up a good 100w when it runs. (My whole system uses 34W max, running at 100% load, but mining, it runs below 20% and 18W total.)

As for the card... did you even TRY mining? Did you even try scrypt? Or are you just guessing, based off some useless stats that someone spouted-off?

I can run 3x 7970 cards, easily off 550w @ 1.5Ghps, and 650w @ 1.9GHs, and 720w @ 2.1Ghs (My 750w PSU doesn't even crank-up the fan. But my GPU's run about 50%-80% fan-speed, respectively.) Of those wattages, 18w is the system, and about 20-25% is the losses from the PSU itself.

You say 1 runs at 450w... that is like two 7970's running at 225w each... so you are obviously underpowered, and it is having to slow-down to keep a "constant voltage", or has to slow-down due to your crappy venting of heat. You do realize that 2x 7970's would consume 4-slots, and the 7990 consumes only 2.5 (Which is a GAIN of 1.5 slots.)

Running 4x 7990's as opposed to 8x 7970's (Which is impossible), is better in more ways than you imagine. However, the price needs to drop to $800, to make it even remotely viable to use for mining. $1000-$1300 for a 7990 is NOT reasonable for getting the power of 2x of a $400 7970 card. That, I will agree with...

P.S. They all come with 4 free games now. If you don't get them, it is because someone took the cards for themselves, and sold them.

[ssateneth]

"The "extra" pins are both ground... (-12v)."

Wrong. Ground is 0v. -12v is used in old RS232 connectors, AKA Serial ports.

I didn't read whatever else you typed, because I assumed it would be just as wrong too.

[dogie]

Very nice results. This card seems very impressive in comparison to the 3-slot 7990s. It's nearly unbelievable they managed to get the cooling to work that well on it.

Do you have any VRM noises at any clocks or volts?

Yep dogie was a complete moron for not physically comparing both cards.  He rather created a thread based on speculation of the old 7990 hardware that is 1+ year old and now he looks like a jackass.  Now hes mad to the point that he is trolling people's threads like a degenerate.   (degens gonna degen.)

As of this post the Sapphire 7990 Malta seems to be the most stable card as far as temps, clock, power usage, etc.

I can get 1260mhash/sec stable; even up to 1300+mh/sec while bitcoin mining; using msi afterburner; stock cooling only.
ROI on this card is 5-6.2 months.

I'm sure I can sell it for a pretty penny after mining a bit of btc as well.

With it all said the 7870 tahiti LE, 7950, 7970, 5870 5970 cards seem to be the best for ROI; this card does OK in that realm.

What are you talking about? No one took you up on your shit blade off because... it was shit.

[greaterninja]

Very nice results. This card seems very impressive in comparison to the 3-slot 7990s. It's nearly unbelievable they managed to get the cooling to work that well on it.

Do you have any VRM noises at any clocks or volts?

Yep dogie was a complete moron for not physically comparing both cards.  He rather created a thread based on speculation of the old 7990 hardware that is 1+ year old and now he looks like a jackass.  Now hes mad to the point that he is trolling people's threads like a degenerate.   (degens gonna degen.)

As of this post the Sapphire 7990 Malta seems to be the most stable card as far as temps, clock, power usage, etc.

I can get 1260mhash/sec stable; even up to 1300+mh/sec while bitcoin mining; using msi afterburner; stock cooling only.
ROI on this card is 5-6.2 months.

I'm sure I can sell it for a pretty penny after mining a bit of btc as well.

With it all said the 7870 tahiti LE, 7950, 7970, 5870 5970 cards seem to be the best for ROI; this card does OK in that realm.

What are you talking about? No one took you up on your shit blade off because... it was shit.

Lol, the thing you do not get in your head is you keep failing when you keep assuming stuff.   The 7990 Malta thread that you created = Fail because you never had the hardware.  You still do not have the 7990 Malta hardware.  Imagine how much of a dumbass people think you are now that they are posting real  7990 Malta results that put your assumptions to shame?   Trolling on my asicminer group buy threads = Fail.  I have ~750 BTC of interest.  Why do you even try?

The thing i don't get is why you go through such depths to say nothing useful on people's threads.  In fact you would rather troll, be negative and contribute nothing productive or useful to the community.
If you do this in real life you will not be very successful.  

dogie gonna degen and degens gonna degen.

[dogie]

Very nice results. This card seems very impressive in comparison to the 3-slot 7990s. It's nearly unbelievable they managed to get the cooling to work that well on it.

Do you have any VRM noises at any clocks or volts?

Yep dogie was a complete moron for not physically comparing both cards.  He rather created a thread based on speculation of the old 7990 hardware that is 1+ year old and now he looks like a jackass.  Now hes mad to the point that he is trolling people's threads like a degenerate.   (degens gonna degen.)

As of this post the Sapphire 7990 Malta seems to be the most stable card as far as temps, clock, power usage, etc.

I can get 1260mhash/sec stable; even up to 1300+mh/sec while bitcoin mining; using msi afterburner; stock cooling only.
ROI on this card is 5-6.2 months.

I'm sure I can sell it for a pretty penny after mining a bit of btc as well.

With it all said the 7870 tahiti LE, 7950, 7970, 5870 5970 cards seem to be the best for ROI; this card does OK in that realm.

What are you talking about? No one took you up on your shit blade off because... it was shit.

Lol, the thing you do not get in your head is you keep failing when you keep assuming stuff.   The 7990 Malta thread that you created = Fail because you never had the hardware.  You still do not have the 7990 Malta hardware.  Imagine how much of a dumbass people think you are now that they are posting real  7990 Malta results that put your assumptions to shame?   Trolling on my asicminer group buy threads = Fail.  I have ~750 BTC of interest.  Why do you even try?

The thing i don't get is why you go through such depths to say nothing useful on people's threads.  In fact you would rather troll, be negative and contribute nothing productive or useful to the community.
If you do this in real life you will not be very successful. 

dogie gonna degen and degens gonna degen.

Crying going to cry more. I have interest from the president for me to run the world, doesn't mean its happening.

Oh and assumptions... you keep saying your Maltas are better than the 3 slots. How many 3 slots do you have? None. So using your logic = fail because you never had the hardware.

[rahrahrah]

We have 1 system with 2 x 7990 in (Gigabyte GV-R799D5-6GD-B) and we're having an absolute nightmare with cooling. Open rig at room temp, additional fans, and a guy with a degree with thermodynamics and fluid engineering helping to maximise airflow.

The second GPU on each card seems to run hotter than the first. The second GPU on the card further inside the (open) case cooks; it climbs to over 90 quite quickly every time, and we've seen a peak of 101. At these high temps the system crashes. Even running cgminer and disabling that GPU doesn't help (the temp still goes up!). We've tried various cgminer settings from I12 to I13, with -g1 to -g2 and threadconcurrency of 8192, 14336 and higher. Nothing is workable at the moment, though we'll try a lower intensity later today. Dropping from I13 to I12 though does result in a loss of around 400kH/s which is quite a lot over the course of a full day/week/month...

Next step is to try going to try replacing the thermal paste with something higher rated.

It's very frustrating because I personally have another 6 cards and 3 rigs to build... or not (and send it back). The only help with the additional rigs is that they'd be in a garage outside with an air temp of ~15 lower (throughout winter at least).