While the Intel Developer Forum conference might have met its end, that doesn’t mean that Intel itself is no longer ushering in new products. The outgoing conference was replaced with the Data Centric Innovation Summit, and Intel used that as the backdrop against which it made quite a few important product announcements. Those announcements show that the chipmaker continues to focus its efforts on the development of products primarily aimed at the data center and AI sectors, particularly with products like Intel Optane DC persistent memory and its new Xeon chips.
The Rebranded Xeon Chip
One item of interest for data centers and others who require state of the art server technology is the next generation Xeon server, which currently goes by the codename “Cascade Lake”. This is a step forward for the company in many ways, not the least of which is because the new Xeon chips will feature Intel Optane DC persistent memory, as well as a baked-in accelerator designed for artificial intelligence.
According to Intel, the accelerator will be able to speed up deep learning inference workloads by 11 times what the current processors can offer. Additionally, the accelerator will extend the AVX 412 and other instructions related to artificial intelligence.
There are a few additional points of interest with the new Xeon chips. One of those is that they offer new security solutions for Meltdown and Spectre vulnerabilities. Note that the new Xeon chips will eventually ship sometime in late 2018.
Cooper Lake
Intel did not stop with just the redesigned Xeon chipset – so-called Cascade Lake. The next chipset is already in the pipeline and it will debut close to the end of 2019. Dubbed Cooper Lake, this one is supposed to offer even more improvements over what Cascade Lake brought to the table, although Intel was skimpy with the details. You can bet on yet more focus on AI, though.
Ice Lake
Intel sketched out plans for yet another chipset, this one to debut in 2020 – Ice Lake. Like Cooper Lake and Cascade Lake, this one will feature baked-in AI-supportive technologies. However, the only concrete detail that Intel released about this future chip generation is that it will be based on a 10nm manufacturing process. Both Cascade Lake and Cooper Lake are based on a 14nm process.
Reaching that milestone has been one of the biggest gripes from customers – Intel has consistently failed to meet this size reduction, despite it being a stated goal for several years at this point. Note that, according to Navin Shenoy, Intel’s current executive vice president and general manager of the Data Center Group, the company has managed to optimize their chipsets for AI with a 5.4x increase in inference.
Optane DC Persistent Memory
For those looking for memory that performs between DDR and NAND flash when it comes to power, performance, and speed, the time is soon. Intel’s Optane DC persistent memory will reach the market later in 2018, although it will be 2019 before it is available to general buyers. Note that this is a new class of memory that acts as an SSD cache and is able to increase performance by up to eight times in comparison with standard DRAM-based systems.
Nervana AI Processor
The focus during Intel’s conference wasn’t solely around new storage and the Xeon chipset. In fact, Intel announced the release of the company’s first public-facing AI chip. Dubbed Nervana NNP L-1000, it will eventually come to the market sometime in 2019. While a developer-only NNP L-1000 was released, this will be the first time the chip has been available for general purchasers. It is supposed to offer up to a four-fold improvement in training performance over the chip’s first generation.
In the end, despite having no one at the helm, Intel continues to push forward, to innovate, and build exciting new technology.
