NVIDIA Takes GPU Computing to the Next Level

HPCwire ^ | September 29, 2009 | Michael Feldman, HPCwire Editor

[Ernest_at_the_Beach]

GPU Computing 2.0 is upon us. Today at the NVIDIA GPU Technology Conference in San Jose, Calif., company CEO Jen-Hsun Huang unveiled a seriously revamped graphics processor architecture representing the biggest step forward for general-purpose GPU computing since the introduction of CUDA in 2006. The stated goal behind the new architecture is two-fold: to significantly boost GPU computing performance and to expand the application range of the graphics processor.

The new architecture, codenamed "Fermi," incorporates a number of new features aimed at technical computing, including support for Error Correcting Code (ECC) memory and greatly enhanced double precision (DP) floating point performance. Those additions remove the two major limitations of current GPU architectures for the high performance computing realm, and position the new GPU as a true general-purpose floating point accelerator. Sumit Gupta, senior product manager for NVIDIA's Tesla GPU Computing Group, characterized the new architecture as "a dramatic step function for GPU computing." According to him, Fermi will be the basis of all NVIDIA's GPU offerings (Tesla, GeForce, Quadro, etc.) going forward, although the first products will not hit the streets until sometime next year.

Besides ECC and a big boost in floating point performance, Fermi also more than doubles the number of cores (from 240 to 512), adds L1 and L2 caches, supports the faster GDDR5 memory, and increases memory reach to one terabyte. NVIDIA has also tweaked the hardware to enable greater concurrency and utilization of chip resources. In a nutshell, NVIDIA is making its GPUs a lot more like CPUs, while expanding the floating point capabilities.

First up is the addition of ECC support, a topic we covered earlier this month, (not realizing that NVIDIA was just weeks away ... announcing ...

[Ernest_at_the_Beach]

fyi...

Earlier and related thread:

NVIDIA's Fermi: Architected for Tesla, 3 Billion Transistors in 2010

[ConservativeMind]

Apparently, Nvidia is only getting a 2% return on wafer production for there highest-end stuff.

http://www.semiaccurate.com/2009/09/15/nvidia-gt300-yeilds-under-2/

This is where Nvidia is heading with their enormous chip designs. A land of no sales beckons.

[NVDave]

I’ve been wondering the same thing for some time now. The gamer-boy market just isn’t that large compared to the whole PC market, and increasingly their chips serve purposes for a smaller and smaller niche. A 2% (or ANY single-digit yield) on a chipset that isn’t going to be a super-high-margin item with pretty robust sales is a money losing proposition. I’m sure they can get their margins up, but the sales rate? Not so sure on that, unless they can plaster these monster chipsets onto a mobo.

What I’d like out of a GPU more than anything these days is support for at least two 1900x1200 screens at full video rates. Four screens would be nice.

[mowowie]

I have a GTX260 that works flawless and FAST with everything i throw at it, for only 180 bucks last Dec.
would never buy like 2 or 3 for SLI or buy the latest and greatest, it’s a waste of money.
Besides there isn’t exactly a good number of great games being produced for the PC. Maybe this chip will change that.

This GPU sounds awesome though.

[Ernest_at_the_Beach]

Well I think AMD has answered that with their announcement...FR Threads:

AMD's Radeon HD 5870 graphics processor

And

AMD's ATI Eyefinity Technology Review

AND

AMD Eyefinity 24 Display Tech Demo On Linux

[Ernest_at_the_Beach]

See #3...they are having the usual trouble with new and big chips....

[NVDave]

Yea, they do. And my current PC is Athlon-based, so I’ve been a user of AMD products in the past.

I just don’t see AMD having a very bright financial future tho. That’s my biggest issue with AMD these days.

[Ernest_at_the_Beach]

Well looks like they are going to own the Video card market for awhile cause both Nvidia and Intel don't have product....to compete with the ATI 5000 series...

And they all seem to be going in different directions...with their architectures.

[mowowie]

Did i read somewhere that the current top of the line Nvidia’s are being bottlenecked by the current top of the line proccesors?

If computers can’t keep up with current video boards how would they ever with these new GPU’s?

[dangerdoc]

These new GPUs are coming with drivers to allow them to do functions that were previously performed by the CPU.

[mowowie]

Makes sense.

[DocRock]

Maybe something for the Folding@Home ping list?

[Ernest_at_the_Beach]

Related thread:

Nvidia Puts On Graphic Power Display With Fermi ( Natl Labs interested )