Bitcoin Forum

http://www.neweggbusiness.com/Product/Product.aspx?Item=N82E16814161399

code MTMKSpice10

$450 - $45 code = $405

That is not Newegg.com; it is Newegg Business, not for consumers, but if only it was...

Well that sucks ...  8)

Signs up for newegg business

Even though newegg business and the normal newegg are not exactly the same (and sometimes even has different prices), you can still make an account on newegg business and shop like normal.

Btw, with the 7970 being nearly the same performance as the 5870, is it even worth it, even at this sale price? Unless people are planning to buy it to flip and profit.

Can I get your setting? I'd love my 5870's to get 650MH/s

I think you mean the 5970.

In that case, the power efficiency of the 7970 will allow for an earlier break-even point if you pay a relatively normal price for electricity.

Of course if power is free, the 7970 is a much better overclocker and usually exceed overclocked 5970 speeds and so wins on that front too.

So you're saying a 7970 can hit over 850 MH/s?

I'm too lazy to pull out my 2nd card (water-cooled so I'd have to disassemble my loop), but there is one clocked at 1290 doing 825 MH/s on the wiki and I have a card that does 1350 that should scale up to ~860 if you do the math. Keep in mind that my card is not exactly an extreme clocker. There are many people with 1.35GHz+ clocks.

Also I haven't done the Asus BIOS flash which gives me 1.4V (instead of 1.3V with other BIOS), which should be good for another 50-100 MHz

discount code has run out  :(

I don't know. My 5970s do 730-790MH/s each, all on stock 1.05V. My 7970 reaches 1170MHz at stock voltage of 1.175V and I don't think it's a particularly bad card, just an average one. That gives it about 700MH/s, perhaps a bit more. Normally I run it undervolted to 1.1V @ 1100MHz to keep it cooler. I'm not sure bringing overvolting into discussion is a good idea, you can do that on 5970 too to get those 850+ MH/s numbers.

I'd say the 5970 is a bit faster mining card, but 7970 has the plus of being new and available, thus having warranty. Probably better resale value in the future too, as it's a better card for most games and featurewise. I suppose it's slightly more power-efficient as well, but not by much, 5970s consume <300W while mining as well.

I think you have a dud card. All 3 of my 7970s can do significantly better than 1170 on stock. I can get 1220, 1230, and 1250 on all my stock voltage 7970s.

The highest clocked 5970 on the wiki did 863.4 MH/s and it is likely with watercooling and overvolting. Since there are so many of those cards, I would assume that's a fairly high result amongst 5970s. However when I looked at the 7970s, most didn't bother seriously overvolting their hardware.


7970s consume < 200W while mining. Using 2/3rds of the power is not "slightly".

  My 7970 will mine for awhile at 1212mhz before crashing.  Max I ran continually at stock was 1175.  ASIC quality of 85%~  However,"stock" voltage varies from card to card.  Some 7970 have a higher default stock voltage.  so your cards reaching higher clocks at stock might simply be because your cards came with a slightly higher default stock voltage. 


Or you got some nice cards and we got duds.

lol watercooling, srsly? I can hit 860 MH/s+ with an overvolt/overclock on my 5970s, and I think my cards are average.  The problem with running them that hard is they guzzle power like crazy.

During the day I run mine a little undervolted/underclocked to save on power.

3x 5970
Rig idle: 228w
Rig mining: 541w
1.67 GH/s
3.08 MH/j

But that's with a Athlon II x4 960t processor.  I could probably trim out a few more watts by going with an AMD Sempron or something.

I'm curious to see how much more efficient your 7970s are.

2.6GH/s@850W from the wall. So we have the same MH/J. That is with a sempron and 4x 7970s.

You really expect your 40nm GPU to actually even come close to a 28 nm GPU in terms of power efficiency? I mean yes we gained a bunch of scheduler hardware, but we're actually 1.5 nodes away from 40 nm (since TSMC cancelled their 32 nm node).

For efficiency result, look no further than here: https://bitcointalk.org/index.php?topic=57410.msg683755#msg683755

That guy is pulling over 4 MH/J (1650 MH/400W) on air, more than 5 MH/j on water.

I cannot compare on idle power since I run i7-3930K @ 4.7 GHz with serious volts all day doing continuous compilation on my work.

I really had no expectations, but the results were really surprising. If you read through the thread, you'd see I posted as well so I'm aware of Roadhog's findings. Where's he's doing 5.04 MH/j per card on water, I'm doing over 5.67 MH/j per card on air with my 5970 setup. 97w load while mining (135W idle at the plug, 232W while mining) 550 MH/s.  Same formula Roadhog used, so it eliminates variables like CPU and other accessories that might add to the system total.

Here's what my numbers look like

315W idle at desktop
Clocks 925/685
1100 MH/s (2 GPUs, I took one of my cards out), 550MH/s per card just like yours.
0.950 VCore
460W load.

460-315 = 145W

Divide by 2 for 2 cards = 72.5W

550 MH/s / 72.5W = 7.5 MH/J.

Edit: I wonder if Roadhog disabled ULPS that would explain the low efficiency numbers. If he didnt, the numbers are not comparable to 5970 since 5970 idles at 42W idle power while the 7970 is 15W for the first card and 3W for each additional card.

Anyways, I think a more fair metric would be "System idle power without graphics card installed". That would lead to better numbers. Then again at that point, you may end up with inaccuracies due to PSU efficiency since for a 1200W PSU, you're approaching the 10% load which is something usually 70-75% efficiency on an 80PLUS Gold unit.

Edit2: Maybe it is best if we compared via GPU-Z? (VDDC Current * VDDC) should give us correct power. Which in my case would be 550MH/s / 75W = 7.3 MH/J

315W at idle is curiously high.  In fact that idle number is getting close to what my gaming rig pulls when mining with dual 5970s (1100 MH/s @ ~360W ~382W). You're pulling 100W almost 80W more than I am at the exact same hash rate.

7970s should very closely approach 5970s for efficiency when downvolted, so I'm assuming there's still some room to streamline for power consumption on your system?

I have 4 hard drives, 2 SSDs, 7 fairly high speed 120mm fans, 4 CCFL UV tubes, and 2 pumps and an overvolted 3930K and 32GB of RAM to power. That's probably where all the extra power is going.


I digress. The advantage of 1.5 node gap in process should allow the 7970 to make a killing as far as efficiency goes. Could you post your GPU-Z VDDC and VDDC Current when mining on the 5970 @ 550 MH/s? Those 2 variables are the only things that can be comparable given our system's likely widely different configurations.

VDDC: 0.9000 V
VDDC Current: 32.0-34.0 A

28W*2 for 550 MH/s?

That's 19 9.5 MH/J, which is FPGA efficiency. Are you sure that's not idle?

That's what it says on GPU-Z.  Actually, switching between cards (I didn't notice I could do that), I see some of the GPUs are at 37.5A.

Edit: And yeah, that's at roughly 277 MH/s per core.

GPU-Z must be inaccurate then.

I did find this post by ArtForz:

https://bitcointalk.org/index.php?topic=60853.msg709254#msg709254


You seem to have a much lower undervoltage than he does, so maybe that's why your card matches 7970s in efficiency. ArtForz is showing 20% efficiency advantage to 7970.

Can't wait to see the 7990 efficiency numbers.

I wouldn't trust GPU-Z numbers and not really sure about the idle numbers, although 7970 does have zerocore and low idle consumption otherwise too. However if running multiple monitors or unofficial overclocking, those idle features may not work. That 7.5MH/J number would be pretty awesome, but I doubt you're getting anywhere near that...


Yours are watercooled though right? That could make a big difference due to lower temps/power consumption. Or were those numbers on stock cooling?

http://gpuz.techpowerup.com/12/04/25/d64.png
^This is the settings I'm using right now on my gaming/mining PC - just the one 7970 in there right now but my aim for the summer is to get three more and go quad-crossfire and watercooled - the thought of spending £300 just on the extra waterblocks I'll need is a tad intimidating though  ;D

Yup I dont trust the GPU-Z numbers anymore. I guess I'll just have to rip out all my cards the next time I redo my loop and measure the no-graphics card wattage. Anyways my total system consumption is 1100 MH/s at 460W which is kind of impressive considering the list of hardware I'm powering in my sig.

Yeah they are WC.

I looked at making an account.  I don't have a business.  I don't see how you could set one up without lying through your teeth.  It requires things like business day and federal tax id# (EIN).  Making that up is asking for trouble IMHO.

It looks like at higher frequencies the 7970 will win out in efficiency, but a 5970 putting out 550 MH/s is only running at about 500 MHz and 0.9 volts, so dynamic power usage must be at a sweet spot.  Static power use of a semiconductor due to current leakage rises exponentially as the thickness of the insulators decrease, so as dies shrink, static power use can account for over 50% of a chips overall power usage.

I guess you can't write off the 5970s as long as you don't mind underclocking and undervolting.

Anyways, that's my analysis but take it with a grain of salt - I am an electrical engineer, but not a good one ;)

i did grab 2 before the discount expired and they arrived today.  my first 7970's and any HIS card.  i am a bit leery of the state of the packaging (outer lame box tamper seal was not secure and the plastic 'coating' around the outer plastic shroud was not fully adhered) but will likely fire them up and see what happens........

I ordered 12 of them. I will check mine out to see if there are any problems.

well, i decided to install one in my gaming rig and it does work - so that is good.  the highest stable clocks (with stock voltage) i can run on it are 1100/950 and it is pulling 656Mh/s with cgminer.  

i don't know enough about these cards right now to judge if that is within expectations or not, but i am suspicious that these could have been returns they just stuck back in inventory.

Not to say that your analysis is worth less than mine, but I have taken several graduate-level courses in semiconductor device physics and very-large-scale-integrated circuit design.


Just so that we are clear with our definitions:
Static power is the power consumed by a CMOS transistor when it is "static" (ie not going from 1 to 0 and 0 to 1). Dynamic power is the power used by a CMOS transistor when switching from 0 to 1 or 1 to 0.

There is no "sweet spot" in dynamic power usage when you look at it from a pure computational efficiency stand-point. Lower voltage (so less current leaks when you switch states) and lower frequency (less capacitive effect) is always more efficient (at least, until you reach the threshold voltage, then the device just stops working).

While it is true that static power consumption maybe lower on the 40 nm node, the lower threshold voltage on the 28nm node means that the transistor spends less time in the dynamic power usage. Engineers optimize static power usage and dynamic power usage using a variety of techniques, which is beyond the scope of this discussion. However, for a properly optimized process (ie. something a company would put out), the amount of energy (ie the sum of the static energy and dynamic energy) to do one unit work of computation always decreases as you shrink in process node (otherwise, there is little reason to shrink process nodes given modern ASIC's thermal-limitation).

I got what I would call a "bad" batch of Black Edition XFX 7970s. They were only able to do 1100/950 stable as well. The normal black editions, for me, were much better. For ~$99 less I think you got a decent deal.

From TSMC's website (http://www.tsmc.com/english/dedicatedFoundry/technology/28nm.htm):

The 28nm high performance (HP) process is the first option to use high-k metal gate process technology. Featuring superior speed and performance, the 28HP process targets CPU, GPU, FPGA, PC, networking, and consumer electronics applications. The 28HP process supports a 45 percent speed improvement over the 40G process at the same leakage/gate.

7970: 4.5B transistors
5970: 4.3B transistors

So 7970 should be 45% faster (minus whatever overhead the scheduling hardware imparts on the architecture) at the same power levels.

Of course. 

7000 series changed architectures and it is less favorable to mining.  Granted 7970 brute forces its way thought with higher clock & higher # of shaders but measured at the wall I haven't seen anyone "blow away" 5970s.

...

At least not yet.

4x5970
3.2 GH/s
1050W @ the wall.
~3.05 MH/J @ the wall.

The 7990 might change that.

You must have missed this then. Unless you consider 50W per card not a significant power savings.

I think the biggest problem with his post is that I have not been able to replicate his numbers fully on any of the 7970s I have bought. Without model information the post by itself, while helpful, does not show the whole picture.

It's not a dog. In fact SIMD is just VLIW with a scheduler slapped on top.

Talking about brute force, 3200 cores is not considered brute force? That's a >50% increase in the number of cores. Whereas the 7970 is clocked only 28% faster.


They're 7970s? Pretty much all of them are reference? What information were you looking for?

You guys have convinced me to drain my loop and do no-card/single-card tests tonight.

Where do you see 50W less per card at the same hashrate?


The 5970 & 7970 is roughly the same performance at high clock.  The 7970 has slightly higher efficiency at a modest undervolt but not this game changing nonsense you keep talking about.  Only at significantly lower clock/voltage the 7970 pulls ahead.  Of course at that significant of an undervolt the 7970 doesn't make much economical sense given FPGA still beat it in MH/W and underclocked it gives up a lot of MH/$.

The last couple of posts in this thread have been talking about mining with efficiency (ie. You pay ~10c/kwh).

I already told you what I was looking for. Model, maybe the specific tools used to measure the outputs? Pictures? What tools used to set the speeds? How long of a mine to determine stability?

Why are you being so hostel? I own 20 7970s. I have 12 more on the way and I will let you know what I find with them. Not one of them has been able to replicate that posts findings for long term stability mining purposes. However, all of my cards are XFX Reference Design Blacks. What can possibly get hurt by sharing more information?

Post was changed. But anyways, you get $0 resale value with FPGA so your MH/$ value argument is invalid. Very few people mine a graphics card until it is worth nothing.

I'm not sure why Art was running his at 950mv since all of my 5970s will do 899mv/610/150, ~232W (at the wall), 554 MH/s.

1) It doesn't make economical sense to undervolt (at this time) given the price of Bitcoins.  I make more at full clock full voltage than I do undervolting.

2) 5970 can go much higher than 600 Mhz @ 0.95V.  I have test 680 Mhz @ 0.95V and 750 Mhz @ 1.0v.  But as I said above it actuall results in LESS net revenue (after electrical cost) to undervolt.


4x5970 rig
820/300 @ 1.05V
3.2 GH/s
1050 W (at the wall)
Daily revenue gross = $10.50 (current price/difficulty)
Electrical cost = -$2.50 (@ $0.10 per kWh)
Net Revenue = ~$8.00

4x5970 rig
680/300 @ 0.95V
2.6 GH/s
840W (at the wall)
Daily revenue gross = $8.50 (current price/difficulty)
Electrical cost = -$2.00 (@ $0.10 per kWh)
Net Revenue = ~$6.50

So yes undervolting does reduce electrical costs but it loses more in gross revenue.
Save $0.50 in electrical card per rig per day but lose $2.00 in gross revenue per rig per day.

Even if we assume 30% overhead due to AC cost it is still less profitable to undervolt.

1) I agree here if you were a dedicated miner who just vents their air outside, but undervolting makes sense who pay electrical and cooling costs (ie. their cards are in a gaming machine). At that point your net revenue may come pretty down close to 0 after cooling costs unless you undervolt (and dont need to turn on your A/C as a result)

2) It's not valid to compare clock frequencies/voltage between different architectures. What IS valid is comparing power usage at the same hashing performance. So what power usage can you achieve at 550MH/s? Feel free to suggest any other point of reference up to ~700MH/s. Seems most 7970s can't get above that even though 5970s might.

That is nonsense.  Gross revenue is currently ~4x electrical cost.  Even with 30% overhead for cooling it is >2x.  Undervolting doesn't improve profits.  Undervolting is useful in large farms where you simply can't handle the heat.  I updated post above to show that.

30%? There's a lot of data-centers that would kill for that kind of cooling efficiency.

Typical A/C operates with a Coefficient Of Performance of 2. Which means it takes 1 W of energy to move 2 W of heat out.

Given that, I've requoted your post and edited it to show that undervolting results in more net revenue.

1) SEER 13 (min legal standard since efficiency standard) is a COP of 3.4.  Units as high as SEER 20 ( COP of 5.2)  even AC units sold in the 1980s had to be at least SEER 9 (~2.4 COP).

2) Anecdotally I know at least my home system required ~30% extra for AC consumption.

3) Still lets assume COP of only 2 (1970s era technology).  You added the AC cost to the first but not the second.

Apples to Apples would be increasing both electrical costs by 50%.  So $3.75 for full voltage rig and $3.00 for undervolted rig.  So it does save $0.75 (instead of $0.50) but still loses $2.00 in revenue.

If you have any other ridiculous claims which need refuting it will have to wait till tomorrow. :)

The lower voltage really helps the 5970 keep the power usage down since switching power use rises exponentially with the voltage.  I really want to see what a 7990 with cherry-picked chips can do...

+1.  I may ditch my 5970s when the 7990s finally come out.

1) SEER-13 is only applicable to central air, and also doesn't take into account system-wide inefficiencies (caused by duct works, leaks, underinsulated window panes, etc etc). The overall COP of the ENTIRE system is clearly under 2.

2) That's great, not everyone has super-efficient cooling as you do. BTW if you run anything else in your house (Air-conditioner? Microwave? Electric water heater?) of course your overhead wont be 50%. You're not comparing the overhead of air-conditioning to cool ALL your appliances, some of which do not generate heat in the form of infrared radiation and thus is not exercising your air-conditioner.

3) I was talking about having it clocked so high you need to turn on A/C where the 2nd case doesnt need A/C.

Loop drained and cards tested:

No-load (remoted in with RDP to make sure system was idle and then logged out and left to idle for 2 minutes): 253W
1-card : 356W (delta 103W)
2-card load: 460W (delta 104W)

550MH/s / 104W = 5.29 MH/J card efficiency.

0.95 VCore.
Diamond Reference 7970s.

Now all that's said and done, please post some no-card, 1-card, 2 card numbers for 5970s @ 550MH/s. Please dont use idle system power with cards plugged in, because it will skew the results since you will be measuring the efficiency of idle to full-load power, not the 0 to full-load power.

I have been mining at said settings for 3 days and there are no signs of errors. All measurements are taken with Kill-A-Watt and verified with the power meter built into my APC BX1500G.

Sapphire 5970s 899mv, 610/167MHz.  All watt readings are at the plug via Kill-a-Watt meter.

No-load: 73W (not sure if I'm doing this right, but all I did was remove my video cards and power the system up and let it sit for a couple minutes)
1-card: 122W idle
1-card load: 221W @ 558 MH/s
2-card: 171w idle
2-card load: 279W @ 558 MH/s (only 1 card mining)
2-card load: 380W @ 1116 MH/s (both cards mining)

So.. whatever that works out to per card in MH/j.

What temps were you guys running when you tok those measurements?  Azn seems to be water cooling, which apparently helps with efficiency quite a bit.

I just want to see an apples to apples comparison.

That looks like it should be right.



For your first card it looks like 221-73 = 148W. So for that card it is 3.77 MH/J.

For your second card it looks like 380-221 = 159W. So for that card it is 3.51 MH/J.


About 55C load. But when the cards are undervolted water doesn't help that much with power. I only saved 50W per card when I maxed out the voltage to 1.3V and ran 1350 MHz clocks.

Oh wrd.  So that is roughly the same as a 7970 on air I believe.  I can't wait to see what the 7990s will do!

My 7970s undervolted do about 570Mhash and 4.94Mhash/Watt at the wall  ;)
They do even better than that if I drop the volts all the way to <0.9V, but I lose another 50mhash/s or so in the process and my electrical rates aren't too bad as is.

7990s I'd be offended if they did worse than 5.5Mhash/Watt :P

>1 GH/s @ >5 MH/W will be interested.  The other good news is NVidia is being aggressive on performance & pricing.  AMD probably would have released it at $850-$900 had 890 flopped.  I don't think they can support that price anymore.  <$800?  Well a guy can hope.

Didn't the 6990 launch at $699?  I'm hoping the 7990 is in that ballpark when it comes out.  And $799 isn't in that ballpark :)

Yeah, not quite sure why some of these ppl seem more concerned with efficiency than profitability.  *scratch*   But, then, I guess you also have people loading up with 7970's that will decrease in value like nobody's business, so..

Couple months ago, I never undervolted.   Now I do so during the day only, because of heat issues (considering different solutions as to where I wouldn't have to do this).  Then, I'm speaking from a perspective of anyone that gets electricity for, oh, I don't know, probably 15-20c or so or less.  I pay ~8-9c per kwh.