Intel continues to optimize its products around the AI



[20/08/2018] Normally this is the time of the year in which Intel would hold its Intel Developer Forum conference, which is full of announcements of new products. But with the fall of the stock exchange last year, the company instead held a full-day event that was broadcast live on the internet.

The company's Data Centric Innovation Summit formed the basis for a series of announcements about processors and memories focused on the data center and artificial intelligence in particular. Although Intel has no leader, it still has a significant momentum. Navin Shenoy, executive vice president and general manager of the Data Center Group, did the hard work.

News about Cascade Lake, the famous Xeon server chip

First there is the news about the Xeon Scalable processor, the new name of the chip for Xeon servers. The next-generation chip, code-named "Cascade Lake," will include a memory controller for the new Intel Optane DC persistent memory and an integrated artificial intelligence accelerator that, according to the company, will accelerate in-depth learning offload workloads 11 times compared with the current generation of Intel Xeon Scalable processors.

Cascade Lake also offers enhanced security features to address the Specter and Meltdown vulnerabilities, plus an expansion of the Intel, called Intel Deep Learning Boost, which extends the Intel AVX 512 and other instructions for the AI. Cascade Lake to start distribution later this year.

The next chip after that, at the end of 2019, is Cooper Lake. Intel did not go into detail, except to say that it would offer a general set of performance improvements, as well as improvements in the workload of AI training. Ice Lake is scheduled for 2020 and the only detail about this is that it will be made with a 10 nm production process versus the 14 nm of Cascade Lake and Cooper Lake. Intel has been struggling for years with its 10 nm chips and has been one of the biggest failures of the company in recent years.

Xeon was optimized for IA less than two years ago, said Shenoy, but has since improved its 5.4-fold inference performance on Skylake, the new architecture used on the Xeon Scalable platform.

Intel distributes the first Optane DC production units

Shenoy also offered updates on its continued Intel Optane DC memory, a new kind of memory and storage located somewhere between DDR and NAND flash memory in terms of speed and performance. And that is exactly where it is located, between the DRAM and the SSD's, which functions as an SSD cache and is capable of achieving up to eight times the performance of a DRAM-only scenario.

Intel said it has sent the first Optane production units to Google and that general availability is planned for 2019.


Source link

Leave a Reply