Nvidia: High performance ARM servers 'one or two years' away
While there's a certain allure to using ARM processors for server applications because of their low power consumption, Nvidia believes the chips won't be ready for high performance applications for a couple of years.
Sumit Gupta, general manager of Nvidia's Tesla accelerated computing unit, sees the x86 architecture as being the king of the server room, for the time being at any rate.
"Performance of these ARM cores is still not where it needs to be for servers. It is getting there; the new ARM64 is going to get it part of the way," said Gupta, as reported by The Inquirer. "One day I think ARM will at least get to similar performance levels as X86 performance. The belief is that over the next one or two years these ARM SoCs will be good enough for cloud applications and web serving. I think it will take some more time to be good enough for accelerated computing".
According to Gupta, Nvidia has been experimenting with combining its ARM-based Tegra 3 SoC -- System-on-a-Chip -- processor with an Nvidia GPU, and Gupta was "shocked by the number of those kits that have been sold".
Nvidia's investment in GPGPU -- or "general-purpose computing on graphics processing units" -- which allows the GPU to take over work previously carried out by the CPU, seems to be paying off.
"The interest in this ARM plus GPU is far larger than even I expected. If the GPU can do dynamic parallelism, it becomes more independent than how powerful CPUs do you need? I believe the first thing that will happen is that people will start using lower performing Xeons. Then, at some point, when these Atom based processors become available they might use that, and when ARM64 is available they'll use that".
Given that Nvidia's primary business is GPUs, the company would like nothing more than to see them become the defining factor in high-performance servers. It would give the company a significant foothold in the server market, and possibly the edge it has been looking for over other players, especially Intel.