In a lot of ways, Nvidia is the belle of the GPU/accelerator ball these days. (Make your reservations early for the upcoming "GPU Fancy Dress Cotillion" later on this year; tuxedo t-shirts encouraged.) Intel withdrew Larrabee, IBM isn't pushing Cell, FPGAs aren't gaining a lot of traction yet, and AMD is late to the party with …
The CUDA shaker?
What really matters for most HPC isn't CUDA, it isn't even OpenCL, it's what sits on top of those - the higher level numerical libraries such as BLAS, LAPACK and PETSc.
Real hardware muscle is fine, but it must provide the means to solve huge sparse matrices - as are abound in fluid dynamics and tectonic acoustics modelling - in a way that requires very little altering of source code. If I'm an expert in turbulence closure schemes for finite volume CFD (for example), it should be clear that I probably won't have the time - or unused grey matter! - to learn OpenCL or CUDA. I hedge my bets on well-established techniques which rely on software libraries that aren't going to disappear any time soon.
So where am I going with this? AMD is missing the boat on this. If the hardware has the capability, one way to completely floor Nvidia in the HPC arena would be to release GPU acceleration patches for the open source numerical libraries that almost every scientist and his white-coated dog use.
Overnight, I can bet you scientists looking to buy clusters will be looking for AMD kit.
What about act 3?
So, as I see it, the question is, for HPC that require more than their xeon/opteron cpus can currently handle, do you go the nvidia or AMD route?
Or do you keep things simple, wait for a year or two (if you can), and intel comes out with its knights corner multiprocessor, negating the need to do additional coding with CUDA or open CL?
If you think that you're not going to have to recompile/rewrite a large portion of your x86/x64 code to run on intel's MIC architecture, you better think again. Not to mention that it's not out yet, the first generation is supposedly only going to be for software development, and we have no real idea how it will perform.
- Mounties always get their man: Heartbleed 'hacker', 19, CUFFED
- Batten down the hatches, Ubuntu 14.04 LTS due in TWO DAYS
- Samsung Galaxy S5 fingerprint scanner hacked in just 4 DAYS
- Feast your PUNY eyes on highest resolution phone display EVER
- AMD demos 'Berlin' Opteron, world's first heterogeneous system architecture server chip