[Unofficial] New Dolphin 5.0 CPU benchmark - results automatically updated!

[Nintendo Maniac 64]

The plan is still to release a 6.0 at some point, but it's not happening in the near future (I wouldn't expect it to happen this year, at least).

Well outside of 5.0 being older, AFAIK it isn't actually causing any issues so I'd rather leave it be then. I mean, Cinebench R15 was the "most recent" version for 5 years up until Cinebench R20, so it's not like there's no precedent.

[Lucario]

Hmm, this seems wrong (AMD Ryzen 7 5800x), I'll test more later.



Edit: Looks like someone else got similar results with a 5600X so I guess that's just how it goes.

Running benchmark once after launching Dolphin emulator has been a fluke it seems, pay attention to the FPS and restart emulation (don't close Dolphin entirely) until there's highest FPS you can get then let the benchmark finish, leave the mouse/keyboard alone.

I cannot replicate my score (262 sec) any more with my i7-7700K @5.0GHz, best I can get is like 268 sec... I tried un/overclock CPU and RAM and everything, still can't, even have reinstalled Windows 10. Could be because I'm on Windows 10 dev beta, the card was upgraded from GTX 670 to RTX 2060, or a slight performance penalty coming from CPU vulnerability patches.

Strange anyone isn't doing i7-10700K or i9-10900K yet!

[ExtremeDude2]

Running benchmark once after launching Dolphin emulator has been a fluke it seems, pay attention to the FPS and restart emulation (don't close Dolphin entirely) until there's highest FPS you can get then let the benchmark finish, leave the mouse/keyboard alone.

I cannot replicate my score (262 sec) any more with my i7-7700K @5.0GHz, best I can get is like 268 sec... I tried un/overclock CPU and RAM and everything, still can't, even have reinstalled Windows 10. Could be because I'm on Windows 10 dev beta, the card was upgraded from GTX 670 to RTX 2060, or a slight performance penalty coming from CPU vulnerability patches.

Strange anyone isn't doing i7-10700K or i9-10900K yet!

I'm actually in the process of (slowly) tweaking my CPU with some of the new stuff that has come out for Zen3, so I'll probably try this again soon.

[Nintendo Maniac 64]

Strange anyone isn't doing i7-10700K or i9-10900K yet!

Perhaps because they're just yet-another refresh of Skylake? I mean the IPC isn't exactly different outside of some spectre mitigation stuff and the minor change in latency due to varying core and cache amounts.

[Lucario]

i9-11900K @ default (5.1GHz does similar score)
4x8GB DDR4-3200 CL14 RAM
GTX 670
Windows 10 Pro 20H2

overall time: 206 seconds

[Nintendo Maniac 64]

I've removed the time posted by "Hannes Berboth" due to running with a much newer Dolphin revision:

• https://imgur.com/a/2tOUZRs

It'd be nice if someone tried running a Ryzen 5000 on Linux, because as it stands the fastest Ryzen time posted is actually a 3800X on Ubuntu 20.04 with a time that's a single second faster than a 5900X on Windows 10.

[belegdol]

I've removed the time posted by "Hannes Berboth" due to running with a much newer Dolphin revision:

• https://imgur.com/a/2tOUZRs
  

It'd be nice if someone tried running a Ryzen 5000 on Linux, because as it stands the fastest Ryzen time posted is actually a 3800X on Ubuntu 20.04 with a time that's a single second faster than a 5900X on Windows 10.

I could try, but the problem is that running such an ancient build on a modern distro is problematic due to dependencies having moved on. Would a self compile of the commit be acceptable, provided it even builds?

[Nintendo Maniac 64]

Either AMD is referring to some other "Dolphin Bench" or senpai has finally noticed us and deemed us worthy of wielding Mjölnir their attention:

• https://youtu.be/WcH_7xsYtUk&t=12m29s

click image to view full size


So about that Dolphin 6.0...

(also, um, wow is that a big per-GHz performance uplift in Dolphin between Zen 3 and Zen 4; that's like Ivy Bridge to Haswell levels of emulation performance gain, and that's without considering that Zen 4 clocks considerably higher than Zen 3)


EDIT: The funny thing is, if AMD used the actual Dolphin 5.0 benchmark, then that predates AVX-512 altogether so then this is just straight-up traditional IPC (do newer versions of Dolphin even take advantage of AVX-512?).

Now the question is, just how much performance does the small 2 CU RDNA2 iGPU on the I/O die have, and can it manage exslusive ubershaders at least at 1x IR... (it should at least be faster than Intel's 6th-10tn gen iGPU since even the 3CU Vega 3 on the Zen-based Athlon APUs was a bit faster, and the 8CU Vega 8 on a Ryzen 4800U can get to around 3x IR with exclusive uber shaders)

[Yahfz]

Pretty crazy that they used dolphin, I wonder how you guys felt. Also, thanks to them it made me submit a new result as well, kindly update the result list when you guys have time, thank you!

[ExtremeDude2]

I thought it would be fun to do a simple comparison between how the benchmark runs on 5.0 vs the latest beta version (5.0-20347)

Spoiler:

5.0:

5.0-20347:

About a minute faster!