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

[Nintendo Maniac 64]

I just kind of wish a new stable release would have come out since it also means .deb packages are also stuck at 5.0 still.

I had every intention of putting together a new benchmark once that happened but, alas, it seems Dolphin has moved to a rolling-release model and, by now, Dolphin is arguably not even demanding enough to be worth benchmarking for newer hardware anyway (I'd argue that both Alder Lake and Zen4 kind of fully make Dolphin a complete "non-issue" even for mobile).

...I however have been humoring the idea of running Dolphin in another emulator which itself would be running the Dolphin CPU benchmark as a way to be more demanding, but AFAIK all newer/more demanding emulators also use a sort of rolling-release model, so that's a bit annoying (though I believe some are less frequent than others).

Of course the irony is that Dolphin's actual "beta" releases seem to slowly be getting farther and farther apart between releases, almost like they're eventually going to turn into "stable" releases in all but name. In such a scenario I've humored the idea of simply using the likes of LLE and/or software rendering in order to make things slower which would additionally provide more precision on the resulting time.

I've also thought about a sort of GPU-compute test using exclusive ubershaders, but I haven't tested at all what'd be the best way to do that (I was thinking maybe the likes of the Quake II shareware running its timedemo or something, assuming the command works in the shareware version and it doesn't make it turn into a CPU benchmark instead).

[BlizzardToaBreeze]

Hello! I'm curious to know if the new release strategy means any changes coming from the benchmarking tool?

[Nintendo Maniac 64]

I've no new strategy.


As I stated in my previous post before it was made official, Dolphin had already then seemed to have moved to being rolling-release but just hadn't made it official. Sadly, rolling-release just isn't a great method for benchmarking since you need a single consistent version, especially when it's something being benchmarked by many other people rather than just a single person.

But because of the performance improvements of CPUs over time, Dolphin itself is actually starting to become a bit irrelevant in terms of something that is even considered "demanding". You only have to look at the graveyard that this "Hardware support" subforum has started to become in recent years—because if even low-end CPUs or dirt-cheap used can run Dolphin with even their integrated graphics at full speed (though maybe at 1x IR), then what help is there to be provided? It all "just works".

While this fact is moderately unimportant from a more scientific view, it's still worth noting that even slightly-newer emulators like RPCS3 have requirements that are much more in-line with modern hardware with support for quite a few more CPU cores, AVX512, and of course hardware demands that actually are worth "benchmarking" at all. I mean, if we think about it, Dolphin and the Wii itself is almost like the last generation of "classic" style consoles—there's only a single CPU core, no programmable shaders, there's no OS always running in the background just to name a few.


So I've basically decided to leave the Dolphin 5.0 benchmark "as is" to sort of fade away from importance. Also, I noticed that TechPowerUp actually does CPU benchmarks with RPSC3 and Ryujinx.