If you are researching the CPU industry and ask where the big money is, look at the server and data center market. Ever since the Opteron days, AMD's market share has been rounded to zero percent, and with its first generation EPYC processors using its new Zen microarchitecture, that number has skipped a small handful of points, but everyone has been waiting for breath for the second turn at ball. AMD's Roma platform solves the concerns of the first gene in Naples, plus this CPU family is designed to do many things: a new 7 nm CPU microarchitecture, offer up to 64 cores, offer 128 paths with PCIe 4.0, offer 8 memory channels and offer a unified memory architecture based on chiplets. Today marks the launch of Rome, and we have some of our own data to share about the results.
Review edited by Dr. Ian Cutress
Sixty-four cores. Each core with an improved Zen 2 core, offering ~ 1
Start for BIOS, check the node information.
[Note: That 1500 mV reading in the screenshot is the same reading we see on consumer Ryzen platforms; it seems to be the non-DVFS voltage as listed in the firmware, but isn’t actually observed]
It is clear that the raw specifications of our new Rome CPU are some of the most impressive on the market. The question then is whether this is the newest fastest server chip on the market – a claim that AMD is putting all its weight behind. If this is the latest fastest CPU on the market, then the question becomes "after how much?" And "how much does it cost?".
I've been covering server CPUs since the launch of Opteron in 2003, but this is nothing I've ever seen before: a competitive core and twice as many of them on a chip as the competition (Intel, Cavium, even IBM) ) can offer. To quote AMD's SVP from the Enterprise Division, Forrest Norrod:
"We designed this section to compete with Ice Lake and expected to make some progress on a threaded performance. We didn't expect to be reheated Skylake instead This is going to be one of the highlights of our career. "
Self-confidence is high at AMD at all times, and on paper it seems justified. The new Roma server processors have improved the IPC core, doubled the core number at the high end, and it uses a new production process (7 nm) technology in one turn. Usually, we see that a server company does one of those things at a time, not all three. It really is a great risk to take, and the potential to be exciting if everything falls into place.
To put this into perspective: promising up to 2x FP performance, 2x cores and a new process technology would sound so strange a few years ago. At the end of Opteron days, only 4-5 years ago, Intel's best CPUs were up to three times faster. At that time, there was little or no reason to buy a server with AMD Opterons. Two years ago, EPYC AMD got back into the server market, but even though performance per dollar ratio was much better than Intel's, it was not a complete victory. Not only was AMD still trailing in database performance and AVX / FP performance, but partners and OEMs were also reluctant to partner with the company without a proven product.
So now that AMD has proven its worth with Naples, and AMD promised more than twice the deployed designs of Rome with a very fast ramp to customers, we have to compare the old with the new. For the launch of the new hardware, AMD gave us a dual EPYC 7742 system from Quanta, with two 64-core CPUs.