Benchmarks II
NAMD is a molecular modelling benchmark that eats up cores for breakfast. The scores show how many days it takes to do a single nanosecond of simulation, where lower is better. Using v2.13b and analysing 327,506 atoms, the dual-socket AMD system is about 25 per cent faster than the Intel. That's not a great deal when one considers the core-and-thread benefit AMD enjoys, but nevertheless, it sets a new record for a 2P server.
As a point of context for those more familiar with desktop processors, like many of our readers, an 8C16T Intel Core i9-9900K takes 1.75 days for the same test, while a 12C24T Ryzen 9 3900X takes 1.5 days. This dual Epyc system is about 6x faster.
C-Ray is a ray-tracing application that fully exploits the floating-point prowess of a CPU. Running at 4K and 16 rays per pixel, it's impressive on the new Epyc processors, which turn in another record time.
Bear in mind that the Epyc 7601, which was the fastest 1P server processor until Epyc 7002 Series surfaced, takes about 24 seconds to complete this test. And the desktop Ryzen 9 3900X pulls in at just under a minute, going by submitted results.
OpenSSL is used to secure links between client and server. This standard test runs with RSA 4096-bit encryption. We're seeing nearly 2x the performance uplift here. A first-generation Epyc 7601 scores about 4,500 signs per second, showing the instrinsic muscle of a 2P system.
The Linux Kernel Build (4.18), as the name suggests, measures the time to compile the kernel. It wasn't so long ago that processors beat the 60-second barrier - you still need a very beefy chip to do so - but 2P server processors finish it with alacrity. Comparing generations, a 2P Epyc 7601 takes 24 seconds for the same task.