Suggest a methodology for benchmarking CPU performance in PCSX2

[jana519]

I want to benchmark CPU performance accurately in PCSX2 (a PS2 emulator) using multiple CPUs. So far, searching has found this as the best example of CPU benchmarks. It uses a cutscene from FFX-2 which is helpful, but doesn't offer an idea of real CPU usage and performance. I'm good at reading benchmarks but I have no idea how to set one up. Can AT offer some suggestions?

 

[TheELF]

but doesn't offer an idea of real CPU usage and performance

What do you mean by "real CPU usage and performance" a PSX2 emulator only emulates psx2 games so the only thing you should care about is how fast X cpu is at playing whatever game it is you like.

If you are searching for a heavy game to really stress the CPUs I recommend "The Warriors"

 

[jana519]

How do I test CPU usage within PCSX2? The benchmark I linked above is benching a cutscene, and my understanding is a cutscene DOESN'T place a great strain on the CPU, so shouldn't there be a better test of CPU usage than watching a cutscene?

Like I said, I have no clue how to set up benchmarks. The goal would be a good, consistent PCSX2 benchmark.

 

[SPBHM]

can you record demos on the emulator? that would be very helpful
if not, what about a game that allows you to save a replay? perhaps a gran turismo game?
in any case, PCSX2 can run the games 100% on the CPU with software emulation, but most of the time people will run with D3D/OGL which makes the games look very different from the original, mostly a lot cleaner, but also at times missing key effects, I would be interesting in seeing both modes

I'm not sure what would be the most intensive PS2 game to emulate?
I remember shadow of the colossus being really rough on the original hardware

 

[jana519]

Yes you can save replays with GT3. So that's a possible choice.

What about choosing an intensive game aka "Warriors" or Shadow of the Colossus. I can run the game from the beginning to a specific point, say 15 minutes. Now to test the CPU performance, should I turn graphic settings all the way up or down?

Regarding the software emulation, that shouldn't be necessary due to using a dGPU.

 

[TheRyuu]

Now to test the CPU performance, should I turn graphic settings all the way up or down?

Doesn't matter since I doubt even if you use 2x/3x/4x Native or your monitors native resolution that it's really even stressing your card that much depending on what you have. You'll probably wind up using the OpenGL hardware renderer since that's the most complete and generally provides the most accurate rendering. Just make sure you're not using software rendering since that's not what you're trying to benchmark here.

The only other thing I can really think to say about this is don't pay attention to the CPU usage since we're not really concerned about efficiency here, just if it can actually run the game at 100% speed. Even if this means pegging all your cores at 100% so long as it runs at full/native/100% speed nothing else matters. You could possibly devise a benchmark which measures how many times it dips below full speed. I don't know if turning off the frame limiter would be useful for this.

 

[jana519]

Just make sure you're not using software rendering since that's not what you're trying to benchmark here.

This makes sense.

The only other thing I can really think to say about this is don't pay attention to the CPU usage since we're not really concerned about efficiency here, just if it can actually run the game at 100% speed.

OK, I see your point, but wouldn't discovering some sense of scalability require a lower power CPU to dip below 100%? I guess what I mean is, if all the CPUs can run it at 100%, how would scalability be discovered? If that makes any sense.