ARM platform appears in MLPerf benchmarks

by ·

Nvidia has been entangled in the acquisition of ARM for a long time, and the progress of the transaction is very slow. It is not easy to pass the review of the regulatory agencies of various countries. However, no matter what the result is, I believe that the cooperation between Nvidia and ARM will be closer to compete for the more profitable supercomputer and data center market.

According to Wccftech, Nvidia has been promoting the ARM architecture in benchmark tests for some time. The latest 1.1 benchmark test results released by MLPerf recently show that when equipped with an A100 GPU, the performance of the Arm-based server is very close to that of the x86-based similar products. Although the x86 architecture server still has higher peak performance, the ARM architecture server is superior to x86 products in power consumption and high-efficiency application scenarios.

For a long time, in the field of high-performance computing, x86 architecture CPUs have occupied an absolute market share. In this reasoning test, the ARM architecture server equipped with A100 GPU defeated the x86 server in the relative niche 3d-Unet workload, but it was still not as good as the other in the more common workloads such as ResNet 50.“Arm, as a founding member of MLCommons, is committed to the process of creating standards and benchmarks to better address challenges and inspire innovation in the accelerated computing industry,” said David Lecomber, a senior director of HPC and tools at Arm.

“The latest inference results demonstrate the readiness of Arm-based systems powered by Arm-based CPUs and NVIDIA GPUs for tackling a broad array of AI workloads in the data center,” he added.

In addition, the NVIDIA A100 GPU performed very well in the MLPerf benchmark test, which was 104 times faster than the CPU. MLPerf’s reasoning benchmark is based on the most popular artificial intelligence workloads and scenarios, including computer vision, medical imaging, natural language processing, recommendation systems, reinforcement learning, and other application scenarios. The platform equipped with Nvidia GPU has won first place in seven inference performance tests. This is also the first test conducted on an ARM-based system.

This benchmark test has 22 platforms, each provided by 7 vendors, most of which have been certified by NVIDIA.

Tags: