Another round of benchmarks

[NeoHippo]

Out of curiosity I also ran the benchmark on my laptop.

[Graf Zahl]

You left VSync on which makes the test useless. Nearly everything says 60 fps.

[BouncyTEM]

My settings said it was off for me. Weird.

I'll do it again later, when it isn't 2 in the morning over here.

[Enjay]

Can anyone provide me with direct links to a mirror of Phobia and Sunder please? I'm trying to download the files from work but the adaptive filter blocks me from those two - probably because someone swore in the idgames review or something. I just want to make sure that I have the exact versions required for the benchmark.

If not, I can get them at home, it'll just take longer (though I don't recall if either of these are particularly large files).

[Graf Zahl]

Phobia has never changed. If you got an older download it's still the same one. Sunder is 3.7 MB.

My settings said it was off for me. Weird.

I'll do it again later, when it isn't 2 in the morning over here.

My guess would be it's forced to 'on' in the display driver's settings.

[Enjay]

OK, here we go. I too got the Burghead error so this is for everything minus that particular benchmark.

Thanks for the batch file BTW. It made the process nice and simple.

[Graf Zahl]

Apparently the Burghead version I used was not the most recent one. Well, never mind. It wasn't the most important test...


I have to say that it's interesting to see how much faster your results are this time. My last round of optimizations after the last benchmark session seems to have helped after all.

[Enjay]

What kind of improvement are we talking about? I don't have the old figures handy. Sounds good though.

To muddy the water a little, I know that nVidia were talking about one of their driver updates having good speed increases too (figures as high as 15% were being thrown around IIRC), but I don't remember if that was after the last set of benchmarks or before. I *think* the last set were done on Vista and I am now on Win 7 so that may also be a factor. However, I don't know how much, if any, effect those two things would have on GZdoom performance.

[Graf Zahl]

The new version is ca. 10-15% faster. That were some optimizations I did during the 1.3.x testing phase but after getting your benchmark results. The speed improvement approximately matches what I get on my card.

[Enjay]

Interesting. That is a good improvement. I did notice that Sonic Map10 Flying was the only one with very low values. Sunder Flying was just under the old Doom 35fps @33 and everything else was above that value, most of them significantly so.

Good job.

[Graf Zahl]

The Sonic map is a hopeless cause. That scene runs with 10fps for me. There's so many hacky skybox effects in that level that the GPU gets constantly stalled. I was just interested how more modern hardware would cope with that but apparently not really much better than mine.

[BouncyTEM]

You were right about display driver settings having that.

Here's the new results.

[TIHan]

Here are two benchmarks on different video cards on the same computer.
Nvidia Driver: 258.96
ATI Driver: 10.9a


AMD Phenom II X4 955 BE 3.2ghz
4 GB DDR3 1333mhz
ATI 4870x2
Nvidia 9600gt

[SlayeR]

Figured I'd give this a shot, just for interest's sake.

System:
CPU: Intel Q6600 2.40ghz
Video: Nvidia GeForce GTX260
Driver: 258.96
Resolution: 1920x1080

[Graf Zahl]

Yes, that was very interesting. I got a much weaker graphics card but the same CPU so it was a great means to compare how much CPU bound GZDoom really is. Very much it seems. On the CPU heavy maps with a lot in sight your system isn't any faster than mine so it looks like in many situations even my rusty old GF8600 cannot be pushed to its limit.

It also tells me that it's a waste of money to upgrade my graphics card without the rest of the computer.

I think I really have to look into multithreading now to get any more juice out of the engine.