AMD Announces Sampling Of Eight-Core ARM 'Seattle' Processor At Facebook's Open Compute Summit

AMD's Andrew Feldman announced today that the company is preparing to sample its new eight-core ARM SoC (codename: Seattle). Feldman gave a keynote presentation at the fifth annual Open Compute Summit. The Open Compute Project (OCP) is Facebook's effort to decentralize and unpack the datacenter, breaking the replication of resources and low volume, high-margin parts that have traditionally been Intel's bread-and-butter.

We've known for months that AMD was working on its own ARM server core, but announcing imminent sampling means that the company is confident of shipping the part in fairly short order. The eight-core chip isn't HSA compatible -- it's not clear what GPU IP is used to provide a video signal -- but AMD has chosen to dip its toe into the market with a chip it could bring to market quickly, rather than working to integrate a custom IP solution based on ARM's own designs.

AMD is claiming that the eight ARM cores offer 2-4x the compute performance of the Opteron X1250 -- which isn't terribly surprising considering that the X1250 is a four-core chip based on the Jaguar CPU, with a relatively low clock speed of 1.1 - 1.9GHz. We still don't know the target clock speeds for the Seattle cores, but the embedded roadmaps AMD has released show the ARM embedded part actually targeting a higher level of CPU performance (and a higher TDP) than the Jaguar core itself.

Looking ahead to the future, Feldman believes that ARM cores could account for up to 25% of the datacenter market by 2019, with a great deal of custom work being done between a major server owner like Google, Facebook, or Microsoft, and the vendors that supply those servers. The implication is that chips may be customized for small, highly specific workloads and relatively low production runs between companies.

Feldman predicts that x86 will still dominate the server market, but that more workloads will have shifted to highly efficient x86 processors. This is something of a moot point, I think -- if you look at Intel's latest Xeon hardware, the company has server chips based on "big-core" x86 designs that hit the 25W TDP mark. This is where the question comes into play of whether big or small cores are better for a given workload, but Intel also has Atom chips in the 7W - 8W range based on the latest Avoton SoC (that's Bay Trail's server variant).

Still, AMD clearly wants to lead this space -- Feldman boldly declared the company would be at the forefront of the ARM server revolution. If it manages to do so, it would give AMD a powerful anchor in an emerging segment in years to come, and revitalize the company's prospects in the overall server market.

Update:  AMD is claiming a SpecInt_Rate score of ~80 for the 8-core Seattle at a 25W TDP, compared to a SpecInt score of 28.1 for the quad-core X2150. An Intel Avoton (top SKU) has a SpecInt score of 106 and a TDP of 20W. How well SpecInt maps to workloads in the microserver industry is, of course, an open question, but Intel's 22nm process gives it a non-trivial performance/watt advantage in this test.