NEC is recognized for its vector processor-run supercomputers, most notably the Earth Simulator. Typically, NEC’s vector processors have been aimed at numerical simulation and related workloads, but not too long ago NEC unveiled a platform that makes its latest SX-Aurora Tsubasa supercomputer-class processors usable for synthetic intelligence and machine understanding workloads.
“The vector processor, with advanced pipelining, is a engineering that proved alone prolonged ago,” wrote Robbert Emery, who is dependable for commercializing NEC Corporation’s highly developed technologies in HPC and AI/ML system remedies.
“Vector processing paired with middleware optimized for parallel pipelining is reducing the entry obstacles for new AI and ML applications, and is set to resolve the worries both today and in the potential that had been when only attainable by the hyperscale cloud providers.”
NEC’s SX-Aurora for AI
The SX-Aurora Tsubasa AI System supports both Python and TensorFlow development environments as nicely programming languages this kind of as C/C++ and Fortran.
NEC offers several variations of its most current SX-Aurora Tsubasa variations for desktops and servers that can tackle FHFL playing cards. The most superior Vector Motor Processor design is the Sort 20 that capabilities 10 cores functioning at 1.6GHz and paired with 48GB HBM2 memory. The card provides a peak effectiveness of 3.07 FP32 TFLOPS or 6.14 FP16 TFLOPS.
Though peak general performance numbers supplied by the SX-Aurora Tsubasa search somewhat pale when in contrast to people made available by the most current GPUs (which are also a class of vector processors), these types of as NVIDIA’s A100, NEC believes that its vector processors can nonetheless be aggressive, specially on datasets that require 48GB of onboard memory (as NVIDIA ‘only’ has 40GB).
As an additional edge, the NEC SX-Aurora Tsubasa card can run standard supercomputing workloads in a desktop workstation.
NEC does not publish costs of its SX-Aurora Tsubasa playing cards, but these who want to consider the products can get hold of the enterprise for quotes. In addition, it is probable to attempt the hardware in the cloud.
Resources: ITMedia, EnterpriseAI, NEC (via HPCwire)