Revealed the first data regarding the performance of the new AMD EPYC Rome, pointing to an improvement of up to 29% in IPC compared to EPYC Naples.
A few days ago the AMD EPYC Rome processors were presented, large processors based on the Zen 2 architecture intended for data centers and similar systems. These new processors are characterized by reaching up to 64 cores and 128 cores through 'chiplets' based on 7nm lithography that are interconnected through a DIE of I / O ports based on 14nm lithography that manages memory and PCIe 4.0 lanes, correcting bottlenecks and improving performance at high workloads, a problem in current EPYCs.
Preliminary data on the performance of AMD EPYC Rome processors
The company reports that the increase in the IPC is 29% with respect to the Zen architecture processors (architecture used in the Naples EPYC) in 'combined and integer floating point benchmarks', an interesting improvement that we will see if it is true, no like the benchmarks of the Vega 20 @ 7nm. AMD has highlighted that it has achieved an increase in the IPC of 52% when compared to Piledrive, the architecture that gave rise to the Bulldozers.
This is the best result obtained in the laboratory, but the IPC is a variable value and afterwards not all systems use FPU work. FLOPs are somewhat more important in data centers, so it is possible that AMD focused on the development and promotion of FLOPs in the new Rome.
We must emphasize that this information is far from what we will end up seeing for the desktop processor market. The processors destined to such specific segments as Data Centers must have other characteristics and support instructions that may not be used in the gaming segment. What is true is that the Rome paint quite well for the professional segment and will be more than a toothache for Intel.
Source: NC
