AMD Unveils Epyc 7H12: A 280W 7nm Server Core to Rule the Roost

When AMD launched in August, it unveiled a top-to-bottom product stack that was set to challenge the best Intel had to offer on the Xeon side of its product family. The highest-end CPU previously announced, the Epyc 7742, is a 64- chip with a 2.25GHz base clock, 3.4GHz max, and 225W TDP for $6950.

Now, for its European launch, AMD has unveiled a new high-end 64-core part. This is a specialty chip, intended for liquid-cooled environments. AMD is partnering with Atos to deploy these principally in 1U solutions, but the company has said that the chip will be available to other customers as well.

The new Epyc 7H12 (I have no idea where the name came from) is a 280W part with 64-cores and the same 256MB of L3 cache. Clocks, however, bump up significantly from a 2.25GHz base to 2.6GHz, an increase of ~1.15 percent. Boost clock is a touch lower than the 7742, at 3.3GHz but these chips are unlikely to run at full boost clock all that often in any case. If you’re buying 64-core CPUs, it’s probably with the goal of loading them and making use of all those extra cores. The fact that the base clock is so much higher, however, suggests the all-core turbo is probably quite a bit higher as well.


AMD is projecting an increase in Linpack performance of roughly 1.11x for the 7H12 over and above the 7742. Pricing on the new chip wasn’t disclosed, but given that the 7742 starts at $6,950, we can assume the 7H12 is going to be priced at a fair bit more. Interestingly enough, AMD is “paying” for the base clock increase with a relatively modest TDP jump. Pushing the base clock up from 2.25GHz to 2.6GHz increases clock by 1.15x and increases TDP by 1.24x. TDP is not the same as power consumption and cannot be considered to be equivalent to power consumption, but it is not unusual for TDPs to rise much faster than the linear increase in clocks.


AMD made a number of announcements at the show emphasizing its own performance and partnerships with companies like Dell and TSMC, which will be deploying its own IT infrastructure on Epyc 7nm CPUs. It’s also partnering with Nokia’s cloud service to improve aspects of 5G performance and launching new hardware with OVHCloud. The company has broken over 100 benchmark records with Epyc according to what it claims are third-party results run by independent organizations.

We haven’t reviewed all of those claims ourselves, but a recent Linux comparison of Epyc versus Xeon performance illustrated that these chips are top performers against Intel in a huge range of tests and dominate comparisons in terms of performance-per-dollar. Intel has promised to retaliate with Cascade Lake chips that are a much stronger offer, but we don’t know more than that about these parts yet.

You might also like
Leave A Reply

Your email address will not be published.