by Anton Shilov
08/25/2008 | 09:29 AM
Just ahead of its Nvision event that kicks off on Monday, Nvidia Corp. decided to once again share its view on the future of computer industry and the products of its competitors, ATI, graphics product group of Advanced Micro Devices, and Intel Corp., who yet has to reveal its discrete graphics processing units (GPUs).
Since the merge between AMD and ATI Technologies in late 2006, the newly formed company has been working on development of central processing units with built-in GPU core. In fact, Intel is also developing its processors that feature graphics processing capabilities. The new chips will hardly offer truly high performance, but will reduce the amount of energy computers consume, their sizes and eventually pricing. But Nvidia believes that such products, which AMD calls Fusion, do not make any sense.
“Joining both components on the same die doesn’t buy you that much. It’s not like there’s a real bottleneck there. And every square millimeter you add to the die is a very expensive millimeter. It’s an incremental expense, not a linear function. It’s cheaper to separate them,” said John Montrym, the chief architect of Nvidia G200, the GPU that powers GeForce GTX 200-series GPUs, reports PC Pro web-site.
For years entry-level personal computers relied on core-logic sets with built-in graphics core, as such a chipset is cheaper than discrete core-logic and graphics card. Nvidia also has a lineup of such chipsets, but there is a problem for the company: once AMD and Intel launch their CPU-GPU products, hardly a significant amount of customers would need those chipsets with integrated graphics, which is a reason why Nvidia reportedly plans to halt its core-logic business (even though Nvidia denies it has such intentions).
“The class of buyer who buys that type of graphics isn’t really buying graphics. They don’t care about it. You’re working out what you can really afford to put on a CPU, and you’re selling it to a customer who doesn’t care. The economics don’t make sense,” said Andy Keane, general manager of Nvidia’s GPU computing group.
It is not completely clear why Nvidia still sells integrated chipsets compatible with AMD’s and Intel’s CPUs to people, who do not care about performance of graphics cores inside their PCs.
Intel plans to enter the market of discrete graphics processors sometime in 2009, but while it has released certain details of its chip, it has never disclosed any specifications, which allows its competitors to claim that Larrabee will never become a success.
“There’s an incredible amount about Larrabee that’s undefined. You can’t just say ‘it’s x86 so it’s going to solve the massively parallel computing problem’,” said Mr. Keane.
Intel hopes that x86-compatible processing engines inside its Larrabee GPU will allow it to offer better programmability compared to existing ATI Radeon or Nvidia GeForce products while still retaining compatibility with DirectX and OpenGL application programming interfaces. One of the problem is that Intel’s Larrabee substitutes certain traditional graphics processors’ blocks, such as render back ends (RBEs) with software implementation, something, which may greatly reduce actual performance of the solution.
“Intel is not a stupid company. They’ve put a lot of people behind this, so clearly they believe it’s viable. But the products on our roadmap are competitive to this thing as they’ve painted it. The reality is going to fall short of the optimistic way they’ve painted it,” said Mr. Montrym.
In addition, Mr. Montrym quoted technology analyst Peter Glaskowsky as saying that Intel’s Larrabee would demonstrate performance on par with ATI or Nvidia solution from 2006.
“As Peter Glaskowsky said, the ‘large’ Larrabee in 2010 will have roughly the same performance as a 2006 GPU from Nvidia or ATI,” Mr. Montrym claimed.
Mr. Glaskowsky denied that he made such a claim and said that Nvidia mis-quoted him.
“What I actually described as equating to ‘the performance of a 2006-vintage... graphics chip’ was a performance standard defined by Intel itself – running the game F.E.A.R. at 60 fps in 1600 x 1200-pixel resolution with four-sample antialiasing,” said Mr. Glaskowsky.
The problem with Intel Larrabee performance is that Intel has not defined performance, but offered estimations of performance and scalability of its future graphics architecture. Still, despite of not having any details about actual speed of Larrabee, Nvidia called the chip due in 2009 or 2010 as one having “roughly the same performance as a 2006 GPU from Nvidia or ATI”.
Nvidia admits that it underestimated performance of ATI Radeon HD 4850 and 4870 graphics cards and now faces tough times on the market. However, the company will work harder from now on.
“We underestimated ATI with respect to their product. We’ve looked very closely at this, and we know there are certain things we can do better. There will be improvements to things from all angles: there are some easy fixes in the software domain that will soon be forthcoming. Believe me, it’s a very prime focus of ours,” said Mr. Montrym.
But the company claims that thanks to its compute unified device architecture (CUDA) compiler as well as PhysX application programming interface, owners of the GeForce 8, 9 and GTX 200-series graphics cards will still get benefits. He did not indicate how can gamers take advantage of CUDA, which is aimed mostly at non-consumer applications.
“ATI did not spend on things like PhysX and CUDA. But we believe that people value things beyond graphics. If you compare only on graphics, that's a relative disadvantage to us, but the notion of what you measure a GPU on will change and evolve. We are forward-looking. And sometimes, when someone is forward-looking, they get a little bit ahead of the game. And that's kind of where we are,” said Mr. Montrym.