GAME ON: NVIDIA brings GPUs to 64-bit ARM servers

ARM's march into data centres and the world's most demanding applications has taken a major stride forward with GPU king NVIDIA throwing its weight behind ARM server-makers by releasing a version of its CUDA parallel programming platform that works with the architecture. HPC types have, in recent years, increasingly turned to …

[bazza]

That barrier is now gone and NVIDIA says three vendors have products ready to roll that bring GPU-assisisted co-processing to market. The three are: .....

and

Separately but at the same event, AppliedMicro also announced that its ARM-based X-Gene “Server on a Chip” is now in a state of “readiness” and that “ … development kits [are] available immediately, and production [will be] silicon available imminently.”

Hmm, if that little collection of news doesn't get Intel quaking in their boots then I don't know what will.

Thing is there's not a lot Intel can do. They could buy (for example) AppliedMicro and shut it down, but that would merely encourage all the others. It would mean that an ARM based server is viable in Intel's eyes. Alternatively they could buy them and keep ARM going, but that would say the same thing too about ARM.

And where would their "x86 can do anything including low power" stance be then? Intel aren't going to change their development direction so far as I can see, but it is surely a risky strategy. What if ARM really does turn out to be a better server chip than Intel?

Thing is, if Intel did make ARMs they'd be the best in the world. Intel are very good at silicon manufacturing, and it would give them a tremendous advantage in the ARM market.

[Peter2]

"What if ARM really does turn out to be a better server chip than Intel?"

Then history suggests that they will be virtually unavailable for 3 years+ due to no significant sized OEM supplying them, until intel has some form of answer to them.

[Charlie Clark]

The HPC market is still pretty small. In the data centre Intel still has the advantage of backwards compatibility and fear of the potential costs of recompiling all the relevant software for the relevant ARM chips and vendors going out of business (Calxeda). For the foreseeable future it needn't worry too much but once there is an industry standard for the boot process and drivers on ARM things might start to move quite quickly. Of course, the sales guys will be busy doing nice, long-term deals with server makers.

[ToddR]

ARM64 plus AMDGPU would be cool

Does nVIDIA have any plans to implement something similar to AMD's hUMA? Or shall we see an AMDGPU/ARM hybrid?

Either would be cool