VizWorld.com

I’ve just received an interesting rumor from a reputable source about some of the goings-on inside NVidia.  It seems that, due to lack of a market, they will be trimming their drivers to only support 8 GPUs per system image instead of 16.

If you’re not aware, NVidia’s drivers right now allow you to run 16 GPU’s on a single computer simultaneously.  The only real reason you’ld want to do this is for massive GPGPU work, and in the HPC arena there are some interesting uses.  Companies like ScaleMP allow you to cluster multiple computers into a “single system image”, effectively turning a small cluster into one giant SMP machine, coagulating all of the resources together. Companies like SGI with their UltraViolet offering do it as well, but in hardware, allowing you to take a giant Rack of equipment and run it as one giant computer, or lots of individual computers, depending on the application.  Right now, users can slap 16 GPU’s in an SGI UltraViolet machine and access all 16 of them from their 1 OS install, reaping the benefits of simple pthreads and shared memory (rather than something more complex like MPI).  In fact, several of the larger labs have been pushing NVidia to raise the limit so that they can add more GPU’s to their machines.  Now, NVidia is going the other direction.

This is a tricky subject.  Yes, the market for 16+ GPU’s in a Single System image is small, but it’s a market with very deep pockets and a market that is at the forefront of what you can do with GPGPU computing.  While they’ll probably save money on reduced driver development , they’re going to burn a bit of their credibility with the HPC community if this really happens.

Also, they might wind up alienating a few of their vendor partners.  Folks like NextIO are preparing big PCI Express Expansion Chassis’ that will allow up to 24 GPU’s to be connected to a single PC.  I’m sure they were hoping the driver limit would go up rather than down as well.

I’ve got a call out to NVidia for any more details.. I’ll update when I hear more.

Update 8:14pm: Lots of information from multiple sources.

First off, seems I might have been mistaken in some of my comments above.  Currently, only 8 GPUs are supported in most platforms but vendors have been promising 12 and eventually 16 in the near future. So the “Reduction” to 8 is really just a freeze at the current capabilities.

Then, an official response from Nvidia (which isn’t really surprising given the above correction):

NVIDIA has never offered support for 16 GPUs within a single image, so nothing has been ‘trimmed’.

As NVIDIA has demonstrated in the past, we pay close attention to the needs of our customers. If we see significant market demand for this functionality, then we will re-evaluate the need to support it.

So, it’s disappointing to see things stopping at 8 GPUs per image, although the usage of such systems is pretty small.  Hopefully, tho, more systems will come online (either in super-workstations, UltraViolets, or PCI Expansion chassis) and they’ll re-evaluate this.

Hardware nvidia, rumor

Over at VRZone, they’ve dug up a screenshot of GPU-Z and 3DMark Vantage Extreme that claim to be showing the results of the ATI Radeon HD6870.  If it’s true, it’s a significant boost that puts it ahead of the NVidia GTX480 (which comes in under 10,000).

The core clock is same as the HD 5870 – 850 MHz. This suggests the performance boost comes from more functional units (perhaps 1920 SP) or improved performance per clock (using some of Northern Islands’ units) or a combination of both. The GDDR5 speed is boosted by a whopping 33% to 1.6 GHz, or a whopping 6.4 GHz effective. The same 256-bit memory interface is retained, but the ultra fast memory results in a massive memory bandwidth of 204.8 GB/s – well over the GTX 480′s 177.4 GB/s. Of course, one of the possibilities for such a high memory clock speed, as well as the impressive benchmark, could be that the card is benchmarked overclocked.

Now, take this with a grain of salt as these could easily have been photoshopped.  However, most of the details make sense.

via Alleged ATI Radeon HD 6870 3DMark Vantage Benchmark leaked by VR-Zone.com.

Hardware amd, ati, rumor

Over at KitGuru they’re caught wind of some “background buzz” that AMD may finally be killing off the ATI brandname, rolling the video products under the AMD banner.

The world is full of weird paradoxes , surprises and contradictions. Apple PCs have always been famous for graphics and yet they don’t support DirectX or, generally, ship with anything faster than a single, middle of the range graphic card. Intel and graphics are two words that most people don’t readily put together, yet Intel is the biggest supplier of 3D graphics processors in the world. In a world where cinematic gets wheeled out every 18 months to describe the next generation GPU without a hint of irony, should we be taken aback by the loss of a brand like ATI?

The ATI brand has quite a long history, but has been a bit confusing since they were acquired by AMD.  Even AMD’s own press releases seem to get a bit confused sometimes about “AMD Video Card” and “ATI Video Cards”.  Bringing it all under one name would cut the costs associated with maintaining 2 high-profile brands simultanously, but remove any pretense of niceties between AMD and NVidia, and ATI and Intel.

via ATI logo finally gone? | KitGuru.

Hardware amd, ati, rumor

Another day, another Fermi rumor.  This time, Toms Hardware brings us news posted over at VR-Zone of the Fermi-based GeForce GTX480 and GTX470 cards:

GeForce GTX 480 : 512 SP, 384-bit, 295W TDP, US$499

GeForce GTX 470 : 448 SP, 320-bit, 225W TDP, US$349

Internal benchmarks reveal that GeForce GTX 470 is some 5-10% faster than Radeon HD 5850 and similiar [sic] for GeForce GTX 480 over the Radeon HD 5870.

Also, the VR-Zone forums are playing host to a picture from Turkish Website DonanimHaber, purportedly of the first Nvidia-made GeForce GTX480.

Are they real?  We’ll know in 7 days.

Update 10am: VRZone has revised the figures like so:

GeForce GTX 480 : 480 SP, 700/1401/1848MHz core/shader/mem, 384-bit, 1536MB, 250W TDP, US$499

GeForce GTX 470 : 448 SP, 607/1215/1674MHz core/shader/mem, 320-bit, 1280MB, 225W TDP, US$349

* The intended GF100 has 512 SP clocked at 725/1450/1050MHz with 295W TDP. It should still be released in the future but just not now. For this launch, GTX 480 has 480 SP with clocks lowered to 700/1401/1848MHz at 250W TDP.

I hate to say it, but the Fermi is starting to look a bit disappointing from a Graphics angle.  I’m sure it will rock at GPGPU (mostly thanks to the double-precision support), but for graphics it’s not looking any better than ATI’s offerings (even lacking a bit, if you bring in the 5890).

via More GeForce GTX 480, GTX 470 Details Leaked.

Hardware fermi, nvidia, rumor

Do you remember the story we posted, ORNL Looks to NVidia GT300 for next Super, back on September 30th? Well, according to SemiAccurate.com, the project was killed since Fermi consumes too much power. However, Legit Reviews contacted several people at NVidia and ORNL who all say that the rumor is false.

Read the original rumor at: Oak Ridge cans Nvidia based Fermi supercomputer

Read the rebuttal at: False Rumor – Oak Ridge Cancels NVIDIA Fermi Based Supercomputer!

Graphics fermi, gpgpu, nvidia, ornl, rumor

Now, there’s not really much to confirm this but a slide deck has leaked out showing some (purported) benchmark results for the upcoming GeForce GTX360 and 380, the first Fermi based video cards from NVidia.

Do not ask us how or where these benchmarks source/originate from as we do not know. For all we care they are fake. However, a user in our forums posted screenshots of benchmarks from what seems to be a NVIDIA Slide Deck presentation. I can confirm that the layout and style is exactly what NVIDIA uses for it's presentations.

The slides reveal internal NVIDIA benchmarks of Far Cry 2, Resident Evil 5 and STALKER Clear Sky. Now this remains speculation, but if the numbers are right, then the future is looking bright and sunny for NVIDIA — well if they get their pricing and TDP right that is.

Update 12/15: These are fake.

via GeForce GTX 360 and 380 Benchmarks.

Hardware benchmark, fermi, nvidia, rumor

Now this is a completely unfounded rumor, but a fun one to think about: Robert Cringely has posted a thought on his “I, Cringely” blog that Intel might simply buy NVidia:

Intel this week cancelled Larrabee, its proposed graphics processing unit (GPU) that was intended to compete with both nVIDIA and ATi (now a part of AMD). The moment AMD bought ATi Intel had to decide whether to build or buy its own GPU to stay in contention. They decided to build, or at least said they had. It’s hard to say how viable Larrabee ever was but at some point it turned from a weapon against nVIDIA to a barrier to Intel buying nVIDIA. So Larrabee had to go, because without that chip Intel presents a much less imposing target for the Department of Justice and Federal Trade Commission which might oppose a merger on anti-trust grounds.

It is an interesting thought.. Buying NVidia would settle several of Intel’s problems:

• Instant power in the HPC space, with the acquisition of the only currently viable HPC GPGPU solution (Tesla)
• Settle the Atom/Ion lawsuits and investigations
• Make them feature-equal with the AMD/ATI combination
• Cover up the Larrabee debacle

It’s an interesting thought, but I’m not sure Intel has the cash reserves to make such an acquisition.  NVidia currently has a market cap of $8.63Billion, as opposed to Intel’s $109.45Billion, so I suppose Intel could leverage some assets to make it a reality.  What do you think?

via I, Cringely » Blog Archive » Intel Will Buy nVIDIA – Cringely on technology.

Hardware intel, nvidia, rumor

It’s still a rumor, but stories are that Nvidia is taking pre-orders on the upcoming GF100, the Fermi chip we’ve discussed before, with an uncommitted ship-date.

It is also said that the company is now taking pre-orders for the GF100 GPU, without any guarantees on the final shipping date for and to it’s partners. Designed on the 40nm fab process, the NVIDIA GF100 graphics processing unit is expected to end-up faster than AMD’s Radeon HD 5870 flagship card, although we have yet to receive specific numbers, features and information.

They are targeting HPC and workstation markets first, as they have the most use for the special features of the Fermi card.  As usual, this is all rumor right now so take it with a grain of salt.

via GeForce Fermi GF100 GPU Launches Early December.

Hardware fermi, nvidia, rumor

A rumor from Asia states that AMD’s newest DX11 chip, to be used in the RV870, could come with a 384-bit  memory interface.

This could indeed be possible, concerning that the RV770 is already quite dependent on bandwidth and the RV870 is expected to have twice the computing power. Also the recent information that the chip would be bigger than 300 mm², affirms the theory of 384 Bit memory interface, because already the dimensions of the RV770 would have made it impossible there.

This would put it on-par with NVidia’s G80 chipset, and could double the memory bandwidth of the existing HD4870 card, increasing it to 240GB/s.

via AMD’s RV870 with 384 Bit memory interface? – News – Hardware-Infos.

Hardware amd, rumor

Still nothing official on the SGI & PSC falling-out we reported on last week.  From SGI’s PR firm gave us the following statement:

SGI remains committed to the HPC community and in fact still has a business relationship with the Pittsburg Supercomputing Center. This means that all of the business dealings are considered internal and confidential, and SGI cannot publicly comment on status of any deal at this time. I am sure you understand.

So what does this mean?  Well, it’s possible that SGI & PSC are still negotiating, either lowering the final performance (sub 1-Petaflop) or raising the amount they are willing to pay.  Either way, we hope to hear more from them.

Hardware psc, rumor, sgi