GPGPU vs. CPU will be the war of 2008-9

GENERAL-PURPOSE COMPUTATION ON GPUS (GPGPU) will be the buzzword of 2008 and pretty much everybody in the industry knows it.

The reason Intel is entering the discrete graphic market with Larrabee is not to show people pretty pictures, but rather to be able to offer a professional series of cards, especially one for the upcoming GPGPU market - which is set to change the landscape of many industries. When you take a look into Graphzilla's bottom financial line, you will get the importance of Quadro series for Nvidia.

We have repeatedly heard Intel downplaying the importance of Floating-Point units but those guys were only doing their job. In terms of raw performance, a GPU of today eats even Cell CPUs, yet alone Clovertown or similar multi-core CPUs.

The reason for the downplay is simple: the next-generation of professional cards from both Nvidia and AMD are such floating-point monsters that no CPU will be able to compete with the G80 and R600 cards. AMD is touting the phrase Stream Computing, while Nvidia is talking about GPU Computing - but the bottom line is that both are the same: using GPU for CPU-style computation.

We should point out that the GPU is still far from being a replacement for the CPU. GPU chips can do insanely well in many maths-intensive, applications, but ones that do not rely on random factor. For any prediction needed, CPUs have massive amounts of cache that can store predictable and non-predictable instructions. GPU does not have room to miss a cycle.

To get back on the subject, in this quarter both AMD and Nvidia will prepare their GPGPU cards, based on R600 and G80 chips respectively. These cards are maxing out possible amounts of memory that can be fitted on the PCB, not too much on display capabilities - around one DVI will be present on the boards, but do not expect too many monitors to be connected to these boards.

We have already posted details about the R600 board here. And we understand Nvidia's GPU Computing products, under the supervision of Andy Keane, head of Computing Business Unit, will release its super computing boards probably around Computex timeframe, in early June.

It is still too early to talk about specifics of the final product, beside the fact that it revolves around the G80 chip and has maximum amount of memory, but Graphzilla is working simultaneously on its software and hardware part - CUDA. The CUDA marchitecture will get its usable framework, and optimised apps will be able to utilise the GPU as an application accelerator.

Like we said up front, over the next two years, the GPGPU buzzword will be bouncing off the walls. It remains to be seen what impact the hype will have. µ