NEC is acknowledged for its vector processor-powered supercomputers, most notably the Earth Simulator. Normally, NEC’s vector processors have been aimed at numerical simulation and very similar workloads, but not long ago NEC unveiled a platform that would make its latest SX-Aurora Tsubasa supercomputer-class processors usable for artificial intelligence and equipment understanding workloads.
“The vector processor, with advanced pipelining, is a engineering that proved alone prolonged back,” wrote Robbert Emery, who is liable for commercializing NEC Corporation’s sophisticated systems in HPC and AI/ML system solutions.
“Vector processing paired with middleware optimized for parallel pipelining is reducing the entry limitations for new AI and ML programs, and is set to address the troubles both of those right now and in the long term that were at the time only attainable by the hyperscale cloud providers.”
NEC’s SX-Aurora for AI
The SX-Aurora Tsubasa AI Platform supports both Python and TensorFlow development environments as effectively programming languages this kind of as C/C++ and Fortran.
NEC offers various versions of its newest SX-Aurora Tsubasa versions for desktops and servers that can manage FHFL cards. The most state-of-the-art Vector Motor Processor model is the Type 20 that capabilities 10 cores operating at 1.6GHz and paired with 48GB HBM2 memory. The card provides a peak efficiency of 3.07 FP32 TFLOPS or 6.14 FP16 TFLOPS.
Even though peak overall performance figures available by the SX-Aurora Tsubasa look instead pale when compared to people available by the newest GPUs (which are also a course of vector processors), this kind of as NVIDIA’s A100, NEC thinks that its vector processors can nevertheless be competitive, primarily on datasets that have to have 48GB of onboard memory (as NVIDIA ‘only’ has 40GB).
As an included edge, the NEC SX-Aurora Tsubasa card can operate common supercomputing workloads in a desktop workstation.
NEC does not publish selling prices of its SX-Aurora Tsubasa cards, but those people who want to consider the item can speak to the company for prices. In addition, it is achievable to check out the hardware in the cloud.
Sources: ITMedia, EnterpriseAI, NEC (via HPCwire)