The AMD EPYC 9004 series, codenamed “Genoa” is nothing short of a game-changer. We use that often in the industry, but this is not a 15-25% generational improvement. The new AMD EPYC Genoa changes the very foundation of what it means to be a server. This is a 50-60% (or more) per-socket improvement, meaning we get a 3:2 or 2:1 consolidation just from a generation ago. If you are coming from 3-5 year-old Xeon Scalable (1st and 2nd Gen) servers to EPYC, the consolidation potential is even more immense, more like 4:1. This new series is about much more than just additional cores or a few new features. AMD EPYC Genoa is a game-changer, and we are going to go in-depth as to why in this article.
These are absolutely monster processors, and widen the already existing gap between AMD and Intel in the server space even more.
Article reads like an advert.
I’m guessing its “more cores, lower clock, smaller caches, same power consumption, worse single-thread performance, get your ass kicked by Amdahl’s law for anything that isn’t extremely embarrassingly parallel”.
No its “AMD has gone chiplet which means a hell of a lot more performance” as unlike with older designs they can take the best of the best chiplets (what we used to call “Gold Binned” back in the day) to make their server chips hella powerful while at the same time minimizing waste as the slower binned chips will still be more than fast enough for Threadripper or even Ryzen CPUs.
Say what you will about AMD but the last couple releases have been pretty impressive across the board, so much so that AMD’s biggest competition seems to be their previous releases more than it is Intel.
bassbeast,
Even if you build these high cores CPUs with top performing cores, I still think Brendan may have a valid point because typically the bottleneck for such massive chips is not the core performance so much as heat & power density, which is the main reason high core CPUs are clocked lower. Even so these extremely parallel CPUs have their place like the aforementioned “embarrassingly parallel” applications. Data centers in particular can benefit form massive server consolidation.
Sure, I’m very happy that AMD is doing well and I am looking forward to seeing how well these perform in both single core and multicore benchmarks when they come out.
These chips aren’t designed for you common or garden PC workload. These are for high end virtualised server environments. Installed in a vsphere host with a dozen or two virtualised servers, these things will be batting out of the park in performance terms. I suppose in general, virtualisation is kind of an embarrassingly parallel problem, in which these chips would be more than ideal for.
In the sense of supercomputing workloads and other high end tasks, a lot of those are embarrassingly parallel too.
“for anything that isn’t extremely embarrassingly parallel”
Not really single parrelized workloads. We’re talking about mostly virtualized and containerized workloads. Which describes basically every enterprise workload (lots of x86 VM) and every VPS and cloud provider, including probably 90% of hyperscaler workloads.
And of course you’re wrong about single threaded performance, obviously that increases each generation, frequently along with core count.
We and our sisters think this is a very good website.
Seo or Digital Marketing expert GO ME！
The only real edge Intel had was manufacturing and market dominance. They will need a lot more than Pat to undo the damage done by previous execs. I hope whoever that will succeed Lisa is from an engineering background just like her. Intel is a good case study of a company ran into the ground by business execs that only care about short term profitability.