Intel Deep Learning Chip Lake Crest is more powerful than GPU

Although consumer-grade chips are still one of the core businesses of Intel, NVIDIA, etc., in fact, the real "cash tree" also needs enterprise-level, industrial-grade products, such as acceleration cards, server computing solutions, etc. Hyun skill opportunity. Last year, Intel and NVIDIA were also eye-catching because of the “depth calculation” of certain indicators. Recently, Intel demonstrated the progress of its new core of the Lake Crest family for Deep Neural Network (DNN). Based on the Nervana platform, Intel claims to provide more powerful performance than GPU at the same transistor density.

It is reported that Intel has invested 350 million US dollars to develop the Nervana software and hardware integration platform for DNN since August 2016. Currently, it includes the Lake Crest family and the next generation Knights Crest. The former is an independent acceleration card scheduled to be released in the first half of this year. They should all be used in the form of ARRIA FPGAs to match Xeon processors.

Unlike the current Xeon Phi, Lake Crest is designed for AI loads. Intel said it uses the Flexpoint architecture, MCM multi-chip package, with 32GB HBM2 storage, internal interconnect speed is 20 times faster than PCIe, floating-point computing performance 8TFLOPS.

The roadmap shows that Intel should announce Lake Crest's debut when it releases the Knights Mill (Xeon Phi) and Xeon E5 V5 Skylake processors this year.

To say sorry, you may have to mention that because of the lock-up order, Intel's acceleration products cannot enter China.


Inclined Operating Control Desk

Inclined Operating Control Desk,Security Operation Console Desk,Tilted Operating Control Desk,Sloped Operating Control Desk

Huaian Qiangsheng Cabinet Co., Ltd. , https://www.qscontrolcabinet.com

Posted on