It will allow some single PCIE slot users to basically have 2 GPUs, and dual slot users to have more than 2 GPUs. Good news. My guess is that they will be slightly more energy efficient than the GTX580 cards (performance per Watt). Hopefully they will drive the price of other cards down too; the GTX580 is still far too rich for many, and if the GTX570 is only the 3rd fastest Fermi then those prices may see an early fall too.

AMD will be showcasing Bulldozer and Antilles at CeBit, which will be in Hannover from 1st to 5th March this year.

NVidia are now expected to release the GTX595 sometime after this; no doubt they want to test Antilles so that they can fine tune their GTX595 to outperform Antilles, at least in some promotional way.

So my guess is that a GTX595 will turn up by the middle of the year.

With high end Sandy Bridge CPU's, Bulldozer, Antilles and a GTX595 en route, this will be a big year for big CPU's and very big GPU's.

For games and regarding shaders it doesn't matter if you increase shader count or frequency. The task is "embarrassingly parallel", meaning even at just 1024x786 we've got 0.8 million pixels per frame and could to first approximation make use of just as many shaders in parallel (this number will go down by a factor of 20 or so if you consider pipeline depth, but still safe). And regarding frequency: we need frames at Hz region, whereas GPU frequencies are in the MHz region.

There are some parts of the GPU, however, whose number is not reduced upon cutting shaders. So these will work faster in a card with less higher clocked shaders, as they'll get higher frequencies too.

And regarding the GTX595 rumors: even a double GTX570 exceeds the 300 W power wall set by the PCIe specification by quite a bit. And they're already down to ~0.95 V, so there's not much room left at the bottom for TSMCs 40 nm process. I could imagine 2 full GF110 at no more than 0.90 V and frequencies below GTX570, making them about as fast as GTX570 SLI. It's going to be a powerful and interesting card, but don't expect miracles ;)

MrS
____________
Scanning for our furry friends since Jan 2002

Anticipated release date is the 22nd March, according to several reports - only 11 days.
It will be interesting to see what the performance is, as we have not seen a dual-Fermi of any kind. One things for sure, at the suggested entry price I won't be the first to fork out. Hopefully the other Fermi's will start to be more reasonably priced, I fancy a GTX570, but only when the price is right.

I can't see NVidia releasing a sub 300W GF110 dual-Fermi; the Radeon HD 6990 uses 350W (or 450W if you flick the performance switch).

Agreed - if they follow the HD6990 they can probably put out a decent dual GF110, i.e. without castrating it too much to stay within 300 W.

MrS
____________
Scanning for our furry friends since Jan 2002

This one is funny.

Perhaps a bit closer to the real thing.

LOL!

MrS
____________
Scanning for our furry friends since Jan 2002

You're about right on the cost - mortgage territory.

Yeah, with 2 GPU's, you would want two CPU cores/threads disabled if you use swan_sync (which you would do, of course).

Hexus suggest a very reasonable TDP of 365W. This might mean the 622MHz I read elsewhere (607) is real and might lay some doubt about if it will outperform a 6990. That said I still think it will be 50% faster than a single GTX580.

Well, if you'd need a mortgage to afford one.. you probably shouldn't ;)

Anyway, I fail to be impressed by these cards (rumored GTX590 and HD6990). In my opinion it's pushing single cards too far, especially for serious 24/7 crunching. I'd rather see more flexible use of more cards of the GTX580 caliber (card arrangement & cases, PCIe slots & their spacing).

MrS
____________
Scanning for our furry friends since Jan 2002

Most people that buy these cards will probably remove the heatsink and fans, and go straight to water cooling. If it's TDP is only 365W and the user has a PCIE2 slot then the system can supply up to 450W to the card – plenty of headroom to overclock to at least the GTX580 ref of 772MHz, probably more saying as these will be the sweetest of cores.
My concern would be the potential lack of capacitors on the card (to save on power). Hope NVidia did not scrimp in this area, but you never know. I doubt it, but if there is a switch such as with the 6990 then perhaps it could be a 365W/440W card.
I also read a suggestion that only 1000 of these would be made available at the outset. A very low number, suggesting a very high price tag, but I think in the long run many more will be released.

Ref specs:

512Shaders per GPU (1024 total)
Memory: 3072 MB DDR5 (1536MB per GPU)
Core Clock :607 MHz
Shader Clock: 1215 MHz
Memory Clock : 3414 MHz
Memory Interface : 768 bit (384 bit per GPU)
DirectCompute 5.0 and OpenCL support
Microsoft DirectX 11 support
NVIDIA PhysX -Ready
Quad SLI Ready
Three dual-link DVI + mini-displayport connectors
Power consumption: 365W

It's GF110 (Rev 1A), so it should work here straight out of the box (using the 267.71 driver). Of course ASUS already have one listed at 612 MHz, so expect some variation. GPUz image
Prices from £550 ($700~ish, 600 Euro)

Review by Ryan Smith at AnandTech.
Going by this Folding graph it should outperform a GTX580 by around 55% for crunching here or at Folding, and if OC'd might get to about 190% of a ref GTX580.

Perspective,

AMD cards do not work here, so any discussion of them is not yet relevant - hopefully some day, but not yet. Anyway, whoever the manufacturer I'm not forking out £500 for a GPU.

I very much doubt that NVidia will release a 28nm GPU in the 2nd quarter this year. They may have a prototype ready in the 3rd Q. but I can't see a 28nm release any time this year.

While I would expect a speed bump with a move to 28nm, I seriously doubt that any 28nm NVidia card will ever outperform a GTX590 by fifteen times.

Still, I'm in no hurry to replace my GTX470's with 500 series cards; the GTX470 is still the best GPU in terms of performance per purchase price, if you can still get one. Maybe next year.

28 nm cards should be a lot more attractive.. but don't expect miracles!

For Cayman ATI changed the fudamental shader architecture (-> different software optimization -> expensive!) for 10% more efficiency per transistor. Sure, more improvements will be made, but we've already reched a quite mature state.

And without architetural revolutions (which would mostly need software to catch up to be useful) there's brute force "more transistors, more shaders" left. They will gain some more headroom at 28 nm, but unless any power efficiency miracle happened (which didn't happen at TSMC 40 nm and didn't happen for Intel at 32 nm, just regular solid improvements), I don't see how they could suddenly use 2 to 3 times as many transistors. Sure, at 28 nm they can pack twice as many transistors into the same area as before, but the power savings from the new process alone will not be 2x, i.e. twice as many transistors will require a lot more power than the old design.

You can't have both, more transistors and significantly reduced power consumption. You have to choose one of them. And I think we all know how hard nVidia was pushing Fermi already.. so upping power consumption further is not really a promising option.

MrS
____________
Scanning for our furry friends since Jan 2002

A GPU with many small efficient cores could be the answer to significantly increase crunching and server farm performance per Watt, but such cards would not sell to gamers, or be of much use as entry level GPUs. Perhaps market changes will open gateways to such cards and the development of software such as OpenCL/OpenGL will drive either NVidia and/or ATI towards such cards, in a similar way to Intel being driven towards developing a range of small efficient CPU cores for server farms.

Well said, Retvari. I've just got to add one detail: in a new process node power the reduction in power consumption not only comes from reduced voltage. The smaller transistors usually also require less power for switching because, well, they're smaller (simply said). This adds to the voltage scaling but is difficult to predict (and TSMCs 40 nm was rather dissappointing in this regard).

@SK: well.. that's actually what a modern GPU is. You can view the individual "shader multiprocessors" in Fermis as individual "cores". Making them even more fine-granular (i.e. each one smaller) would actually cause more overhead for instruction decoding, distribution etc. Utilization may be better, but you'll get less shaders / computation power per transistor. That's why the SMs actually grew even bigger in CC 2.1.

MrS
____________
Scanning for our furry friends since Jan 2002

What I meant is having more cores, as in the GTX590 has two cores. A card with 4, 8 or 16 might well be a better performer in terms of computational power per Watt. I'm not talking about 3Billion transistors per core though, perhaps 1B or 1.5B at 28/22nm for 8 cores.

We, as a lab, are waiting to buy the new 28nm cards and PCI3 bus motherboards (for parallel runs). We are expecting a factor at least 2 in performance for kepler because the Fermi core design is now mature (see g80 to g200).

gdf

Well said again, Retvari. I think I could add some more, but don't see a need to do, especially since we'd be getting OT.

@SK: oh you mean distributing the shaders among several smaller chips (I don't think we should use the term "core" for GPUs, as it's too unspecific).

That could give you several benefits regarding power efficiency:
- You could physically spread the chips out, so you've got more space for cooling. This generally means higher performance cooling is possible. Reduced chip temperatures lead to higher power efficiency.

- You could spread the memory bandwidth demands among several narrower busses, which are themselves easier to implement and can provide more aggregate bandwidth than the current solutions, allowing for lower memory clocks.

- You could bin the chips more carefully (less variation in 200 mm² than in 500 mm²) and drive them at optimal working points (voltage & clock).

- Yield would improve (which has no direct influence on power efficiency).

I do see some problems, though:
- You already said such cards would be worse at gaming due to non-ideal scaling in SLI / Crossfire. Well, screw gaming.. but there's more to it. The same principle limits performance at HPC applications. For BOINC we'd be fine with that, since each chip could run its own code without any need for communication with the other ones. However, such GPUs are not built for this market alone, but rather for the entire HPC crowd. And here you've got many problems which you want to solve as fast as possible. Think of a single time step at GPU-Grid - you can not run this across different GPUs efficiently because the communication would be too slow. Far slower than on-chip. Even if we could get the bandwidth between chips (e.g. using optical chip-to-chip communication) this would still add latency, which would make the multi-chip solution slower than the fat chip (everything else being equal). High-bandwidth long-range communication can also quickly become a power hog, reducing efficiency.

- In a multi-chip solution you still want to produce only a single type of chip (lithography mask sets are really expensive). Therefore you'd duplicate units, which you need only once for a single card. You'd need more transistors to get the same job done, which usually reduces power efficiency.

- If your individual shader multiprocessors in a Fermi execute different code (or just different stages of the same code), the requirements regarding cache-size and -bandwidth and memory bandwidth may differ. In the fat chip you get automatic load balancing accross all shared ressources. In the multi-chip solution you loose this advantage, potentially reducing power efficiency.

All in all I think it's probably not worth it, as long as the fat chips can be handled. Anyway, worth thinking about :)

MrS
____________
Scanning for our furry friends since Jan 2002

Easy: TSMC 28 nm is not ready yet :p

MrS
____________
Scanning for our furry friends since Jan 2002

Zoltan, could you post up the details here - thanks

In theory the GTX590 (ref clocks) should be about 50% more productive than the GTX580. I noticed that the GTX580 is about 50% faster than the GTX470, which is about 27% faster than a GTX465. Anyway, we are now in the position that the top CC2.0 Fermi has about 290% the performance of the lowest at GPUGrid.
Should the GTX590 clock to the same freq as a ref GTX580 (772MHz) then a GTX590 could do about 369% the work of a GTX465.

In terms of maturity the GTX500 series cards are definitely more energy efficient than their GTX400 counterparts; in terms of performance per Watt a GT240/GT340 could outperform a GTX470, but this is not the case with the CC2.0 GTX500 series cards. I still think that more refinements are possible and these would be augmented with a move to 28nm. I just hope they do not take the architectural frailties of CC2.1 to 28nm.

The ACEMD client is having trouble to keep the Fermi GPUs busy.

I mean the Fermi GPU utilization is pretty much CPU speed dependant, and WU type dependant. If these factors won't change for the better in the future, then the ACEMD client will need 8GHz+ CPU cores to feed the new GPUs.

GPU utilization is 96-98% if using only the GPU and there will be no problem at least for a couple of generations. GPU utilization is low if you don't use SWAN_SYNC or for some workunits which uses a little of CPU. This last case will be less and less important as the new application moved even this tiny bit on GPU.

gdf

It is the case that GPU's are progressing faster than CPU's, so any app that depends partially on a CPU will increasingly hinder the GPU performance, in the long run. However there is a bit more to it that brute force speed. Take for example your 2.8GHz Pent D - this 3 generation old processor is not actually as quick as a 2.4GHz IC2D (E6600). In turn a Wolfdale is about 10 to 15% faster than an equally clocked Conroe. Move on to an i3 and you see another increase in performance, and then again when you move to Sandy Bridge. These 32nm SB processors clocks well and 5GHz is not too hard to reach. By the time we move to 22nm GPU's, it will be old hat for CPU's and 5GHz will be common for high end CPUs. By then I would expect to see many app improvements, but even if there was not it would not be all doom and gloom, so long as you upgrade you CPU when you are upgrading your GPU.

Yes, the FP execution units in 1st generation Core i CPUs are the same as in Core 2 CPUs. But due to numerous tweaks actual hardware utilization is better (depending on load, of course).

The higher voltage using the i3 is intersting. Could be increased GPU temperatures due to the higher GPU utilization, which reduces its maximum frequency at a given voltage slightly, in turn requiring more voltage for the same clock.
Or the PSU could be worse (e.g. stronger ripple).

MrS
____________
Scanning for our furry friends since Jan 2002