Stanford DAWNBench v1 Tests Reveal Impressive Results For Intel Xeon CPU

Intel has been having a rough time with AMD Ryzen and Spectre security exploits, however, Intel Xeon CPU is proving to be quite impressive as Intel Xeon has beaten the Stanford DAWNBench v1 Inference tests beating Google’s TPU v2 and NVIDIA’s various GPUs.

The tests were conducted between Intel Xeon CPU, Google’s TPU v2 and various Nvidia GPUs including the Tesla V100. While Google’s TPU v2 was the one to beat the ImageNet but, it was Intel Xeon that took the crown due to its cost efficiency. The Xeon processor was able to process 10,000 images for just $0.02 with a latency if 9,96ms.

Speaking of Intel, the company has halted the production of H310 chipset. According to the report, sources have revealed that the reason why Intel has suspended the production of H310 chipset is the 14nm process.

The increased demand for the 14nm process has caused the suspension of the production of Intel H310 chipset. Furthermore, the production halt will remain until July 2018 and the H310 chipset will become available in the market in July.

Speculation has emerged claiming Intel suspended the supply of H310 because of a manufacturing process change from 14nm to 22nm for the chipset after a delay in advancing to 10nm had resulted in tight capacity for 14nm.

Furthermore, Intel reportedly will reveal its Discrete GPUs at CES 2019. According to the report, Intel has completed the first phase of the development of its discrete GPUs and are now preparing for a big reveal and launch of the Intel Discrete GPU.

Also, the upcoming Intel Cascade Lake architecture supports up to 224 Physical Cores for a single system. According to leaked slides from an Intel ‘Saudi Conference”, Intel Cascade Lake architecture will feature 28 Cores on a single chip which is way low compared are to AMD EPYC which features 64 cores on a single chip.

However, Intel is prepared to counter AMD EPYC by offering dual, quad and even octa-socket support for the upcoming Intel Cascade Lake which will bring a maximum of 224 physical cores on a single system.

Source: Stanford