NVIDIA: The future of supercomputers is ARM, not x86

In the future, HPC is most confident in the power consumption ratio of the ARM architecture. NVIDIA believes that the future of HPC computing is tending toward ARM rather than x86. The company is currently developing the world’s first with the Barcelona Supercomputing Center (BSC) in Spain. ARM architecture HPC testing machine, using NVIDIA Tegra 3 and CUDA GPU combination.

In an interview with EETime, NVIDIA HPC product manager Sumit Gupta stated that the only strength of the x86 system is that it allows Windows-like operating systems to run faster, responding more quickly to unknown tasks such as clicking on a mouse or using a keyboard. However, ARM's energy consumption can not be replaced by this advantage.

Gupta added that this is why a large part of the x86 CPU die is used to perform branch prediction and speculative execution. In HPC, these two points are no longer important.

Gupta said: "For everyday computing applications that are used every day, the x86 processor is very good, but it is not suitable for HPC." At the same time, he also said that the future of HPC must be more energy-intensive CPU + GPU hybrid acceleration computing platform Advantage.

At present, the goal of NVIDIA and Barcelona Supercomputing Center is to bring the ExaScale calculation while reducing the energy consumption to 15 to 1/30 of the current level.