Dolphin, the GameCube and Wii emulator - Forums

(03-03-2017, 05:06 PM)Nintendo Maniac 64 Wrote: [ -> ]Well it might not be necessary because that result is in-line with the "slower than Haswell" result of luabench, and puts it just a bit faster per-clock than Ivy Bridge.

So now the question is, when will https://dolphin-emu.org/docs/faq/#which-cpu-should-i-use be updated?  The FX8000 series is no longer the best that AMD has.

While I am OK with my purchase, anyone who is looking for an AMD chip to use specifically for Dolphin should probably wait for the 4/4, 4/8 and 6/12 chips to come out. I still have a good feeling about that 1600X lol.

Hey, random thought - what happens if you run the two benchmarks with SMT disabled?

Oddly enough, some (non-emulation) benchmarks are actually showing performance improvement by doing that, so I wonder how Dolphin would react.


Also, I'd appreciate it if you'd submit your results for the Dolphin 5.0 benchmark; surely you're not too lazy to do that.



EDIT:

(03-03-2017, 10:30 AM)admin89 Wrote: [ -> ]AMD Turbo Core still sucks then . Last time i tried , it didn't work, i had to disable it , then set the CPU ratio and Vcore manually

(03-03-2017, 01:22 PM)Gregtastic Wrote: [ -> ]I would agree with this sentiment. Turbo is apparently 3.7 GHz yet when running the benchmark it never went over 3.12.

If I'm understanding correctly, it seems like setting the Windows power scheme to "high performance" works around this issue. Supposedly Windows keeps trying to control any p-state changes when in actuality Ryzen itself is supposed to handle that.

BTW, what motherboard are you using?

Ryzen 1800X, thanks to andymanic1 posting results on reddit
TL;DR 472s

I'm hearing that there are a lot of things that are a little dodgy with many BIOS versions randomly disabling things, and thread scheduling seeing SMT and enabling Intel hyperthreading-specific optimisations which aren't helpful. Supposedly AMD are having a big event on tweaking software to go faster at GDC, and they're claiming that when they've done their own tweaks on stuff that goes slower than expected, they can make it go a lot faster without hurting performance on other CPUs.

(03-03-2017, 11:56 PM)utack Wrote: [ -> ]Ryzen 1800X, thanks to andymanic1 on reddit.
TL;DR 472s

Could you disable SMT and  test again? AMD's  SMT seems to be poorly scheduled at this point  and it  decreases single  threaded performance in games. Perhaps Windows has to be updated.

http://www.gamersnexus.net/hwreviews/2822-amd-ryzen-r7-1800x-review-premiere-blender-fps-benchmarks/page-7

I have seen STP hit as high as 2129 here

http://cpubenchmark.net/cpu.php?cpu=AMD+Ryzen+7+1800X&id=2966

Its the metric that PCSX2 uses not sure how well it works for Dolphin. At its current rating it would put it on par with my i5 3570k at stock speeds.

Also Ramoth those are NOT utacks numbers since he got them from a user on reddit. so yeah nothing he can do about disabling SMT as its not his system that was tested

Seems that turning off SMT actually hurts Dolphin performance:

https://www.reddit.com/r/pcmasterrace/comments/5x3cvv/ryzen_review_mega_thread/degunni/

More benchmarks...



1700 @ 4Ghz = 7:46

1800X Stock = 7:34

http://www.legitreviews.com/amd-ryzen-7-1700-overclocking-best-ryzen-processor_192191/4

(03-04-2017, 07:27 AM)Nintendo Maniac 64 Wrote: [ -> ]Seems that turning off SMT actually hurts Dolphin performance:

https://www.reddit.com/r/pcmasterrace/comments/5x3cvv/ryzen_review_mega_thread/degunni/

Any  idea why Ryzen  has  such  discrepancy without SMT?

(03-03-2017, 05:34 PM)Nintendo Maniac 64 Wrote: [ -> ]Hey, random thought - what happens if you run the two benchmarks with SMT disabled?

Oddly enough, some (non-emulation) benchmarks are actually showing performance improvement by doing that, so I wonder how Dolphin would react.


Also, I'd appreciate it if you'd submit your results for the Dolphin 5.0 benchmark; surely you're not too lazy to do that.



EDIT:

If I'm understanding correctly, it seems like setting the Windows power scheme to "high performance" works around this issue.  Supposedly Windows keeps trying to control any p-state changes when in actuality Ryzen itself is supposed to handle that.

BTW, what motherboard are you using?

Going to submit them now. I was very busy yesterday and didn't get a chance to be on the computer.

I have the ASUS PRIME X370-PRO.