Touring the Intel AI Playground – Inside the Intel Developer Cloud

We had the opportunity to do something that is just, well… cool. In August, just as STH was preparing to move from Austin to Scottsdale, I had the opportunity to head up to Oregon and tour something I had been asking about for at least a year. My questions to Intel were: What is the Intel Developer Cloud? Where does it run out of? Is it only a few systems set up that you get short-term SSH access to? All those and more were answered when I visited the Intel Developer Cloud.

As a quick note, we are going to say this is sponsored since we had to fly up to Oregon to do this piece, and also, this is not common access. It took well over a year to go from the idea to getting the approvals to doing the tour. As with everything on STH, our team produces this content independently, but we just wanted to call this out.

Touring the Intel AI Playground – Inside the Intel Developer Cloud

As one would imagine, we also have a video for this article since it was so cool to see.

For those who have not seen this yet, the Intel Developer Cloud is the company’s place to try out systems in a cloud environment with various technologies, including Intel Xeon, Xeon Max, GPU Flex Series, GPU Max Series, and (formerly Habana) Gaudi. Something that I did not know before the tour was that Intel has service tiers ranging from a more limited free test drive tier to paid plans for developers and teams. There is another option for enterprises that need larger scale deployments as a more customized program.

One can create an account, and through Intel’s various developer account types and presumably some sales logic, different platforms become available to you to try.

At that point, SSH credentials are deployed alongside instances running on hardware, and access is granted to develop on platforms. Some plans have early access to hardware, support, and additional toolkits. Here is an example of starting an instance with a 4th Gen Intel Xeon Scalable (Sapphire Rapids) Platinum 8480+ and four PCIe GPUs (Max 1000):

That was only part of the equation, however. What I wanted to know, and pushed Intel to let me get access to, was the hardware that this is actually running on. After many approvals, we got access to another part of a data center suite in Oregon that I had been to previously. We are only allowed to show a smaller fraction of this suite (one of many suites in the DC), but I grabbed this photo for some scale of the floor size. Check out the lights overhead as they extend well beyond the cage we are in, and it is a fence behind me, not even the corner of the floor.

What I can show is a few photos from my December 2022 visit to the facility from perhaps the opposite corner from the above photo. There, Intel has seemingly countless systems running large scale testing for things like reliability at scale for cloud providers.

At the time, it was exciting to see how many of these systems were running Sapphire Rapids that would not be launched for about a month after this visit.

Intel also has areas here that go far beyond what looks like standard servers. There are things like these development systems where Intel can cause voltage drops on different parts of the platform and more to test what would happen if a component failed, for example.

Hello to the Lantronix Spider we reviewed in the fun photo above.

For some sense of scale, this was just one aisle in the facility with systems set up for this type of testing.

The bottom line is that Intel has its own “farm” of development systems just down from its Jones Farm campus and Oregon fabs. Whether it is for the Intel Developer Cloud today, testing at scale for reliability for cloud customers, or doing platform development work,  there is a lot here. Knowing that we are going to show off just a tiny portion, of this suite in this single facility should give some sense of the Intel development scale.

With that, let us get to our tour.