Intel Xeon 6766E/6780E Sierra Forest vs. Ampere Altra Performance & Power Efficiency

Written by Michael Larabel in Processors on 5 June 2024 at 11:22 AM EDT. Page 4 of 6. 10 Comments.
miniFE benchmark with settings of Problem Size: Small. EPYC 9684X was the fastest.
miniFE benchmark with settings of Problem Size: Small. EPYC 9684X was the fastest.
miniFE benchmark with settings of Problem Size: Small. EPYC 9684X was the fastest.

While HPC workloads aren't the typical focus of the Intel Xeon 6700E/6900E processors, for some cases like unstructured finite element analysis with miniFE had the Xeon 6766E/6780E performing very well while the Ampere Altra Max M128-30 was performing similar to the 64-core EPYC 8534P/8534PN processors. On a performance-per-Watt basis the Xeon 6766E/6780E were at the front alongside the EPYC 9754 Bergamo processor and well ahead of the Ampere Altra Max for power efficiency.

QuantLib benchmark with settings of Configuration: Multi-Threaded. EPYC 9754 was the fastest.
QuantLib benchmark with settings of Configuration: Multi-Threaded. EPYC 9754 was the fastest.
QuantLib benchmark with settings of Configuration: Multi-Threaded. EPYC 9754 was the fastest.

There was a surprisingly good showing for Ampere Altra Max with the QuantLib quantitative finance open-source software and even managed to secure a rare first place finish for power efficiency compared to these latest generation x86_64 processors. But QuantLib and quantitative finance codes in general are better suited for upcoming Intel Xeon 6900P Granite Rapids processors.

x265 benchmark with settings of Video Input: Bosphorus 4K. Xeon Platinum 8592+ was the fastest.
x265 benchmark with settings of Video Input: Bosphorus 4K. Xeon Platinum 8592+ was the fastest.
x265 benchmark with settings of Video Input: Bosphorus 4K. Xeon Platinum 8592+ was the fastest.

For the x265 video encoder that is well optimized for modern ISAs, the best raw performance for a single 4K video stream is with Emerald Rapids / Sapphire Rapids still. Though for those doing concurrent video encodes of multiple streams may find the Sierra Forest CPUs being better utilized. In any case with the one-stream scenario, the performance-per-Watt was similar between the Xeon 6766E/6780E and the M128-30.

Liquid-DSP benchmark with settings of Threads: 256, Buffer Length: 256, Filter Length: 57. EPYC 9754 was the fastest.
Liquid-DSP benchmark with settings of Threads: 256, Buffer Length: 256, Filter Length: 57. EPYC 9754 was the fastest.
Liquid-DSP benchmark with settings of Threads: 256, Buffer Length: 256, Filter Length: 57. EPYC 9754 was the fastest.

With the Liquid-DSP digital signal processing software, the Ampere Altra ARMv8 performance was much slower than the other tested processors while the Intel Xeon 6700E series was delivering the best performance per Watt of the tested Intel CPUs.


Related Articles