Processors

Intel updates information on cascade lake, snow ridge and ice lake to 10nm for its datacenter processors

Table of contents:

Anonim

At this CES 2019 Intel has come up with new information on its 14nm architecture for the Cascade Lake naming datacenter as well as 10nm Snow Rigde and Ice Lake. We have had no official words from the electronic giant since the celebration of its Datacenter Summit last August. In addition, he discussed new support for AI and NNP security.

Source: Anandthech

It has been over a year and a half since the first Intel Xeon processors appeared under the 14nm architecture called Skylake-SP. And it is time for the manufacturer to provide us with news regarding the successors of these brutal processors, even if they are not 10 nm.

The information provided by Intel is that the new range of 14nm processors under the name Cascade Lake are ready to be purchased as production by large companies that are dedicated to cloud processing, such as AWS, Google, Azure or Baidus.. These customers are part of Intel's internal testing program to test their new 14nm processors, we are not making the leap to 10nm yet. Surely these large companies have been testing the first silicon during these months before until we finally have knowledge of it.

These processing units are not yet standardized for retail, as in the past Intel Datacenter Summit made it clear that they were custom-made units for very specific tasks, called “ off-route units ”.

In any case, one of the great innovations that Cascade Lake is going to bring is the support with Intel Optane Persistent Memory where we will increase the capacity of RAM memory per socket in figures of several TB (terabytes). Hardware security patches will also be provided for Specter v2. Intel's strategy is very clear, create units that implement security directly in hardware so that companies are almost forced to acquire these patched 14nm units en masse to obtain this level of security without the need for an already insecure, software.

Cascade Lake is not expected to be ready for retail until at least mid-2019.

10nm Snow Rigde architecture for connectivity and networking in datacenter

The second main news that Intel has given regarding its 10nm architecture has to do with 5G implementations and the news that are to come with the 802.11ax protocol. Snow Rigde's goal is to provide wireless access to virtualization stations, data centers and AI.

He has not elaborated on this new architecture, although based on the specifications of the Xeon processors, these new 10nm saddles are expected to contain multiple Sunny Cove cores, large memory addressing capabilities and are exclusively focused on network.

10nm Ice Lake Xeon, fact or fiction?

Source: Anandtech

We ended the section of news related to the Intel Xeon with Shenoy's words that dropped that the company is actively working on the Ice Lake Xeon architecture, to implement Xeon processors at 10nm. At the last event, Architecture Intel, a processor called Ice Lake Xeon was shown. Calling this piece of chip Ice Lake Xeon means that Intel already has tangible proof that they have functional, high-performance processors for the datacenter, what we don't know is when we're supposed to see a physical test of one's performance. from them.

New information on the Nervana NNP family of processors for Artificial Intelligence

The NNP or Nervana Neural Network Processor family of processors is under investigation for the manufacture of processors specially designed for large-scale Deep Learning and Artificial Intelligence jobs. The machine learning process consists of two phases, a learning phase (training), in which the machine processes enormous amounts of existing data to create patterns and relationships. Which in turn leads to the second phase of inference, in which this learned information is used to create new information by themselves to foresee events that have not yet occurred. In short, create an artificial intelligence network.

The fact is that Intel is working on silicons with high processing capacity called NNP-L, to cover the first stage of learning. But it has also started a line of investigation for lower-powered processors, called NNP-I, to exclusively dedicate them to the second phase of intelligence.

The company's goal for this same 2019 is, we suppose, to make these processors work together to create a scalable artificial intelligence system where several of these processors work together. We build on the structure that Nvidia has in its AI systems, with several Tesla V100 GPUs for the learning phase and as many Tesla P4 for the inference phase. We will be attentive to the next news that the manufacturer gives in this interesting (and dangerous) field of providing intelligence to machines.

How much will we see words come true?

Precisely there is the kit of the question. Intel has already spoken at various events about this new architecture with various winter names that even John Snow himself would question. So we want to see tangible evidence, performance graphs, and physical chips that implement 10nm architecture. We are already looking forward to experiencing this new performance step.

In the previous news about Ice Lake, we saw how Intel almost officially introduced the name Ice Lake-U and the beginning of 10nm processors for mobile equipment. Deploying patches under the name Cascade Lake is all well and good, but time doesn't tick in their favor with an AMD unleashed in creativity with its stunning Ryzen. Now it's up to you to say When do you think we will see all this news from Intel come true, will it be worth the wait, or will the AMD game win?

Anandtech font

Processors

Editor's choice

Back to top button