CPU Benchmark Performance: Legacy and Web

In order to gather data to compare with older benchmarks, we are still keeping a number of tests under our ‘legacy’ section. This includes all the former major versions of CineBench (R15, R11.5, R10) as well as x264 HD 3.0 and the first very naïve version of 3DPM v2.1. We won’t be transferring the data over from the old testing into Bench, otherwise, it would be populated with 200 CPUs with only one data point, so it will fill up as we test more CPUs like the others.

The other section here is our web tests.

We are using DDR4 memory at the following settings:

  • DDR4-3200

Legacy

(6-1a) CineBench R10 ST

(6-1b) CineBench R10 MT

(6-2a) CineBench R11.5 ST

(6-2b) CineBench R11.5 MT

(6-3a) CineBench R15 ST

(6-3b) CineBench R15 MT

(6-4a) 3DPM v1 ST

(6-4b) 3DPM v1 MT

(6-5a) x264 HD 3.0 Pass 1

(6-5b) x264 HD 3.0 Pass 2

Looking at Ryzen 7 5800X3D's performance in our legacy section of our CPU test suite, it does seemingly lack behind the Intel 12th Gen Core series in terms of overall grunt. The most interesting element is that the discrepancy is clear to see between the clock speeds of the 5800X3D and the vanilla 5800X; the 5800X is slightly faster and is marginally better than the 5800X in compute-heavy tasks.

Web

(7-1) Kraken 1.1 Web Test

(7-2) Google Octane 2.0 Web Test

(7-3) Speedometer 2.0 Web Test

Even in our web-based tests, the regular Ryzen 7 5800X outperforms the 5800X3D with 3D V-Cache quite consistently. This shows that the additional L3 cache has no bearing on performance in this area.

CPU Benchmark Performance: Encoding and Compression Final Words
Comments Locked

125 Comments

View All Comments

  • nandnandnand - Thursday, June 30, 2022 - link

    Great results for the 5800X3D in Dwarf Fortress and Factorio. Clearly it does not have enough cache for the World Gen 257x257 test.
  • dorion - Thursday, June 30, 2022 - link

    It shows the exceptional predictors in Intel's architecture, they don't have 96MB of L3 cache either(duh) and yet they whip that world generator. Wonder what odd coding of Tarn's the CPUs are butting against. And exceptional amount of civilizations and monsters to simulate for 550 years in the 257x257 world?
  • AndreaSussman - Sunday, July 31, 2022 - link

    Hello
  • Samus - Thursday, June 30, 2022 - link

    Also impressive is how much the cache improves WinRAR performance. Going from last to 2nd place - with a lower clock speed
  • nandnandnand - Thursday, June 30, 2022 - link

    Compared to some of the earliest reviews like Tom's, this one found more productivity/code scenarios where the extra cache helps it edge out over the 5800X, despite the lower clock speed. Obviously there are niches where the 5800X3D will do really well, like the workloads that Milan-X can boost by >50%. You won't usually see them all in one review.
  • DanNeely - Sunday, July 3, 2022 - link

    WinRAR has always been extremely dependent on memory performance. That a huge cache benefits it isn't a big surprise.
  • emn13 - Monday, July 11, 2022 - link

    Sure, but it's more than just memory performance that's an issue here: in very abstract principle compressors need to find correlation across broad swaths of memory. It's actually not at all obvious whether that's cache-friendly; and indeed in 7-zip it appears not to be.

    After all, if your compression context significantly exceeds the L3 cache, then that cache will largely be useless. Conversely, if your window (almost) fits within the smaller L3-cache, then increasing its size is likely largely useless.

    The fact that this helps WinRAR but not 7-zip is not obvious. Given the compression ratio differences, I'm going to assume that 7-zip is using more context, and thus can't benefit from "just" 96MB of cache. And perhaps that WinRAR at higher settings (if it has any?) wouldn't either.

    That does make me curious how the 3d-vcache impacts the more high-throughput compressors such as zstd or even lz4 perhaps.
  • brucethemoose - Friday, July 1, 2022 - link

    Just wanna say I am ecstatic over the DF/Factorio tests. Stuff like that is where I'm most critically CPU bottlenecked these days, as opposed to CPU Blender or Battlefield at 720p.

    I'd like to suggest Minecraft chunk generation as another test (though its a bit tricky to set up). Expansive strategy/sim games like Starsector, Rimworld, Stellaris and such would also be great to test, but I don't think they're fully automatable.
  • 29a - Tuesday, July 5, 2022 - link

    I’d also be interested in a Stellaris benchmark.
  • ballsystemlord - Thursday, June 30, 2022 - link

    @Gavin , is it just me, or do you have two sets of identical WoT Benchmarks at 1080p? BTW: I'm looking at the print view.

Log in

Don't have an account? Sign up now