JMC Wrote:I'm not shocked that it can fluctuate between 10 - 15 fps during a runthrough
Not only in this benchmark, but overall the performance varies a lot. And it's not "randomly inconsistent" but somehow "deterministic".
It's something like this:
1st run = 100% speed
2nd run = 115%
3rd run = 100%
4th run =100%
5th run =100%
6th run =100%
7th run = 115%
8th run = 100%
The other "it's just the nature of computers" variance is also present, but is pretty small, amost insignificant - about 1% (2% max.) on an optimized OS with very few background services running.
Make sure you're doing a boot from an in-game save, and not some ready made savestate, since things can get left in memory from previous runs that mess up testing.
And why are you complaint about speed increases? That's normally a good thing
No savestates (I never use them).
Performance should always be at 115% (100% in that example is just a randomly chosen value so you can easily see the relative difference in performance between runs)
JMC Wrote:I'm not shocked that it can fluctuate between 10 - 15 fps during a runthrough
10-15 fps of what? 40->55fps is a huge difference, while 150->165fps is relatively small.
You're free to debug this issue yourself if you're interested in it.
The same thing happened with the windwaker benchmark in the past, people and objects move in the background and appear or disappear from the screen creating fluctuations, the world isn't static. In the observatory stars are moving around, i think it is normal.
Kirbypuff: I can't seem to private message you, but I was a bit irritable earlier when I first woke up. If you want to sort out the benchmark stuff, please PM me, and I can discuss what I think is going on and make some builds for you. Thanks!
Also, it usually is about 150 - 165 fps fluctuation for me. In games like NSMBWii, it fluctuates 2 - 3 fps. I don't know if it's a Dolphin bug or anything.
Anyway, the minor D3D improvements build is something I can check into for you on my AMD.
Also, would you mind if I kind of combined all your threads later? I think it'd be easier to reply to you if we just had some "Possible AMD CPU/Video Card performance regressions" thread. As such, the Faster Hash thing I want to discuss here a bit.
I think it's totally possible because the AMD processors take more cycles to do the hash operations that maybe using the Faster Hashing is a lot slower. Could you possibly get the Virtual Console game "Castlevania III" since that is the absolutely best example of a hash limited game we have?