What's the latest on 7xxx series from radeon?

[CanaryInTheMine]

It may be a joke, but on deepbit, someone has a mining team: 'AMD-7995-Prototype' with a user 'Quad-Core-GPU' hashing at around 4GHashes....  Hmmmm....

[1713416858]

1713416858

[1713416858]

1713416858

[johnj]

It may be a joke, but on deepbit, someone has a mining team: 'AMD-7995-Prototype' with a user 'Quad-Core-GPU' hashing at around 4GHashes....  Hmmmm....

This is why I'm not so sure everyone buying up those 5970's and 6990's the last few days at $500+ knew that these were (in theory) just right around the corner...

[CanaryInTheMine]

It may be a joke, but on deepbit, someone has a mining team: 'AMD-7995-Prototype' with a user 'Quad-Core-GPU' hashing at around 4GHashes....  Hmmmm....

This is why I'm not so sure everyone buying up those 5970's and 6990's the last few days at $500+ knew that these were (in theory) just right around the corner...

around the corner is a long time wrt to mining...

AND I bet the price will be insane, at least initially...

[mike678]

It may be a joke, but on deepbit, someone has a mining team: 'AMD-7995-Prototype' with a user 'Quad-Core-GPU' hashing at around 4GHashes....  Hmmmm....

This is why I'm not so sure everyone buying up those 5970's and 6990's the last few days at $500+ knew that these were (in theory) just right around the corner...

I remember hearing most of the top end 7xxx cards wont even come out till q1 so there's still a long time to go before the high end cards get here.

[johnj]

[Mousepotato]

It may be a joke, but on deepbit, someone has a mining team: 'AMD-7995-Prototype' with a user 'Quad-Core-GPU' hashing at around 4GHashes....  Hmmmm....

This is why I'm not so sure everyone buying up those 5970's and 6990's the last few days at $500+ knew that these were (in theory) just right around the corner...

I heard that the top-end HD 7XXX series won't even be available until next year.

[Mousepotato]

It may be a joke, but on deepbit, someone has a mining team: 'AMD-7995-Prototype' with a user 'Quad-Core-GPU' hashing at around 4GHashes....  Hmmmm....

This is why I'm not so sure everyone buying up those 5970's and 6990's the last few days at $500+ knew that these were (in theory) just right around the corner...

I remember hearing most of the top end 7xxx cards wont even come out till q1 so there's still a long time to go before the high end cards get here.

I should read through the entire thread before replying

[mike678]

I should read through the entire thread before replying

Its all goooood

[mrb]

This is why I'm not so sure everyone buying up those 5970's and 6990's the last few days at $500+ knew that these were (in theory) just right around the corner...

"Right around the corner" == "multiple months". Any large-scale miner (10+ Gh/s) constantly scaling up and up simply cannot afford to sit idle on his mining profits for this length of time.

Also, people saying that the dual-GPU 7000 series doing ~1.5Gh/s will "change the mining game" are exaggerating. This figure is correct but this is only twice better than a 5970s or 6990s, or business as usual in the GPU industry. (The last manufacturing nodes 55nm -> 40nm -> 28nm roughly doubled the number of ALUs per square millimeter each time, hence doubled the mining speed for the same power envelope).

[johnj]

This is why I'm not so sure everyone buying up those 5970's and 6990's the last few days at $500+ knew that these were (in theory) just right around the corner...

"Right around the corner" == "multiple months". Any large-scale miner (10+ Gh/s) constantly scaling up and up simply cannot afford to sit idle on his mining profits for this length of time.

Also, people saying that the dual-GPU 7000 series doing ~1.5Gh/s will "change the mining game" are exaggerating. This figure is correct but this is only twice better than a 5970s or 6990s, or business as usual in the GPU industry. (The last manufacturing nodes 55nm -> 40nm -> 28nm roughly doubled the number of ALUs per square millimeter each time, hence doubled the mining speed for the same power envelope).

Yeah, I hadn't thought about the large-scale guys who gotta keep on the upnup. *glances at pitiful 5770*

Correct me if I'm wrong though, but i thought the 5x series was superior to the 6x series.  Bitcoin hasn't seen a huge jump in over a 'generation' of hardware.  And when the 7x does hit it'll divide the miners into the 'haves' and 'have-nots', similar to the divide between cpu/gpu mining when that first hit.  But i'm still gettin my bitcoin history straight, i could be off.

[indio007]

http://www.rage3d.com/interviews/amdchats/eric_demers_july_2011/index.php?p=4

AMD's Graphics Core Next Gen architecture is scheduled to appear in products due in Q4, 2011, as both Eric Demers and Rick Bergman confirmed for us. It is likely they will use TSMC's 28nm process, and have a dual architecture split like Northern Islands.

[mrb]

Correct me if I'm wrong though, but i thought the 5x series was superior to the 6x series.  Bitcoin hasn't seen a huge jump in over a 'generation' of hardware.  And when the 7x does hit it'll divide the miners into the 'haves' and 'have-nots', similar to the divide between cpu/gpu mining when that first hit.  But i'm still gettin my bitcoin history straight, i could be off.

Both the 5000 and 6000 series are 40nm, there is no big difference between them. The 5000 series cards are somewhat favored because they are less expensive, but that's all.

We already witnessed a technology gap between the 4000 series (55nm) and 5000/6000 (40nm), dividing people between the haves and have-nots. The 7000 series will simply create yet another similar gap.

[jjiimm_64]

Both the 5000 and 6000 series are 40nm, there is no big difference between them. The 5000 series cards are somewhat favored because they are less expensive, but that's all.

I would have to disagree with you on this point.  the 5 series runs cooler, clock em higher, and they have more stream processors.

[Pipesnake]

I would have to disagree with you on this point.  the 5 series runs cooler, clock em higher, and they have more stream processors.

Agreed

5970 >> 6990

[mrb]

Pay attention to my wording, guys.

I said 5000 vs 6000, not 5970 vs 6990.

You cannot generically state that the 5000 series is "cooler". The 5830 is terribly inefficient (Mh/J-wise) when compared to the 6990 for example. In fact, power efficiency overlaps a lot between the two series. And yes the 5970 is my personal favorite too.

[wndrbr3d]

I hope people realize that the transition to 28nm means they can fab the same GPU on less silicon. A die shrink doesn't mean they're going to fab a GPU with the same surface area as a 40nm GPU but with 85% more ALU's, because that wouldn't make sense at all.

Odds are they're going to use the die shrink to crank out a higher number of GPU's to meet (hopefully high) demand. With the move away from VLIW, it's anyone's guess as to how the on-die real estate will get divided up.

[NetTecture]

I hope people realize that the transition to 28nm means they can fab the same GPU on less silicon. A die shrink doesn't mean they're going to fab a GPU with the same surface area as a 40nm GPU but with 85% more ALU's, because that wouldn't make sense at all.

Odds are they're going to use the die shrink to crank out a higher number of GPU's to meet (hopefully high) demand. With the move away from VLIW, it's anyone's guess as to how the on-die real estate will get divided up.

I hope you realize what you say is utterly sutpid.

ATI has no problem makin as many 6990 as they want - the problem they have is that this does not happen "at will" and they simply scheduled too little capacity on those chip lines. You can bet you ass that they will scheule larger percentage for high end cards of the 7xxxx cards are better than the 6xxx series for mining (which they will be) to accomodate the increased miner usage.

But, again, once scheduled this is not really the problem -as can be seen by the amounts of 6970 available on distribution channel.

What is more criticla is that they are in a struggle with nvidia that is to a large degree decided by speed. So, getting a 28nm card out, beforre nvidia, they will make sure they kill there. They likely will make chips that are significantly faster than the 6xxx series (and no, 85% is not the number... we talk of 185% more density), also ramp up the speed (which is a side effect of smaller structures) and come out with something 2-3 times as fast as the last generation (the question is whether this will also be integer performance). THEN they will also use a little less power - and start killling NVIDIA as competition as long and as strong as they can.

Doing what you says would mean they totally loose the competitive advantage of moving to 28nm before Nvidia. it is so utterly stupid it would result in the board of AMD getting letters from some lawyers from their shareholders for destroying corporate value. This is not a struggle for yield and production costs for a large part, it is simply a struggle for performance.

This is, btw., the same thing that always happend with the exception of the 5xxx -> 6xxx transition which was a really bad thing - basically the 6xxx series was supposed to come IIRC in 32nm, but they could not get the production process reliable in time, so they went with the same "old" process and got stuck in 40nm like the 5xxx series. Bad timing. This time they are early - earlier than Nvidia - and I bet (as in: buying AMD stock vs. Selling Nvidia) that they will use it for some really nice marketing spin.

[wndrbr3d]

I hope you realize what you say is utterly sutpid.

... blah blah blah.. tl;dr;

My original point is that you won't see double transistor count after a die shrink. Sure, manufacturers want to maintain an edge over their competition by rearchitecting the core or making other enhancements, but it's just naive to think that a core shrink from 40nm to 28nm will yield double performance.

Thanks for your super smart insight though 

[NetTecture]

I hope you realize what you say is utterly sutpid.

... blah blah blah.. tl;dr;

My original point is that you won't see double transistor count after a die shrink. Sure, manufacturers want to maintain an edge over their competition by rearchitecting the core or making other enhancements, but it's just naive to think that a core shrink from 40nm to 28nm will yield double performance.

Thanks for your super smart insight though 

It is not. Do not forget that there are two items at the same time:

* More transisteos per surface
* Higher clock speeds possible.

Find that bad argumentation? Look at what happens to the CPU front - same thing.

Even with only 50% more transistors they can more than double performance by ramping up clock speed.

[mrb]

My original point is that you won't see double transistor count after a die shrink. Sure, manufacturers want to maintain an edge over their competition by rearchitecting the core or making other enhancements, but it's just naive to think that a core shrink from 40nm to 28nm will yield double performance.

Huh? AMD did precisely that when going from 55nm (HD 4870 = 800 ALUs) to 40nm (HD 5870 = 1600 ALUs).

They will do the same from 40nm to 28nm.

In fact it will be a bit easier this time to double performance because 55**2/40**2 = 1.89 but 40**2/28**2 = 2.04.