Alibaba and AMD collaborate on GPUs in the cloud

This is a really big deal

AMD FirePro logoAMD may be late to the cloud GPU compute game but an Alibaba win shows their progress. SemiAccurate thinks this win, while substantial, is probably the first of many.

The release is simple enough, Alibaba is now using AMD professional GPUs for their cloud GPGPU offerings. That’s it, not a big deal right? Actually it is for two reasons, the scale of Alibaba and AMDs groundwork, together they have some interesting implications.

First is the sheer scale of Alibaba, they are one of the top few cloud players in the world so the volumes from even a small win can be pretty immense. For comparison think about Microsoft and their cloud offerings, when this relative pipsqueak in the field did a FPGA ‘trial’ it was over a thousand units. When it rolled out that grew by several orders of magnitude. Alibaba is much bigger than Microsoft in cloud.

The other thing about Alibaba that most westerners don’t realize is that they are on the cutting edge of several research areas most importantly AI. Search, response time, predictive behaviors, and the rest of the supply chain and fulfillment process optimization has the company on the bleeding edge of most applicable technologies. Unlike the common perception in the west, they know what they are doing and then some. If they pick a product, there is a damn good reason for it.

That brings us to AMD and their groundwork. If you look back to the announcements around Firepro/Pro/whatever and memory sizes, virtualization, and storage/SSG, it is clear that AMD has been quietly putting the groundwork in and innovating. No one else has several key features that they have, nor do they have the pricing tiers. WX, Pro, and vanilla. The middle one is important because if you don’t need specific software certifications or hand-holding support, you can cut your GPU compute MSRP in half.

This isn’t to say that Alibaba doesn’t’ need support but they probably don’t need a certified tech to swap out failed cards in their datacenter with a four-hour SLA. They definitely don’t need certified CAD drivers either, their run their own software so the standard professional driver should be more than enough. Lastly as Intel has shown, megadatacenter prices don’t have any relationship to MSRP. That said AMD has laid the groundwork to halve the price vs Nvidia while still making vastly larger margins than before.

So in the end, we have a cutting edge AI and cloud company endorsing AMD GPUs for compute purposes. They wouldn’t have touched AMD unless the products lived up to the promises on both hardware and software, this tier doesn’t play around. If there were any discrepancies we wouldn’t have gotten anywhere near test deployments much less public announcements and availability. Now AMD can go around saying, “If it is good enough for the best, you probably should take a look too”. This is a big deal.S|A

The following two tabs change content below.

Charlie Demerjian

Roving engine of chaos and snide remarks at SemiAccurate
Charlie Demerjian is the founder of Stone Arch Networking Services and SemiAccurate.com. SemiAccurate.com is a technology news site; addressing hardware design, software selection, customization, securing and maintenance, with over one million views per month. He is a technologist and analyst specializing in semiconductors, system and network architecture. As head writer of SemiAccurate.com, he regularly advises writers, analysts, and industry executives on technical matters and long lead industry trends. Charlie is also available through Guidepoint and Mosaic. FullyAccurate