Mafia 2 Demo Benchmark

[evolucion8]

If I had a source I would have given it. But I think the bottom line is that Cuda itself doesn't 'do' SLI. Running Cuda on multiple GPUs is normally done by programming each GPU individually, not via SLI.
And there's a logical problem here:
"tie up the shaders and kill the fillrate."
For example:
1) Shaders aren't necessarily tied up all the time, there's still quite a bit of idle time during graphics. This time can be used by other workloads.
2) There's more to fillrate than just shader power.

But I usually enabling PhysX results in a performance drop, which means that there's not too much units idling at all, specially with nVidia's scalar approach. But I've never heard that PhysX calculations drop pixel/texture fillrate, while there's some scenarios where the shader have texture reads, that might experience a drop in performance, not sure though....

 

[Scali]

But I usually enabling PhysX results in a performance drop, which means that there's not too much units idling at all, specially with nVidia's scalar approach.

Not really.
Generally enabling GPU physics also increases the physics detail a lot... the GPU basically gets a lot more workload than the CPU would ever see (or be capable of handling anyway).
Obviously the GPU won't have THAT much idle time... but the alternative is letting the CPU do it, and then the GPU would sit INCREDIBLY idle, because you'd be completely CPU limited.
Obviously less or no PhysX effects will always be faster, but that's not the point.

 

[evolucion8]

Not really.
Generally enabling GPU physics also increases the physics detail a lot... the GPU basically gets a lot more workload than the CPU would ever see (or be capable of handling anyway).
Obviously the GPU won't have THAT much idle time... but the alternative is letting the CPU do it, and then the GPU would sit INCREDIBLY idle, because you'd be completely CPU limited.
Obviously less or no PhysX effects will always be faster, but that's not the point.

At least in Batman AA were impressive the lots of details, but in other games like Dark Void, it was very unimpressive, the same goes for GRAW 1 and GRAW 2 and yet, there's a performance drop (A smaller one), and Batman AA effects were more impressive than Mafia 2 (At least on the demo) and it wasn't so hard in the performance department, probably by the time that the game is launched, it will have some optimizations for it.

 

[evolucion8]

I played with phsyx on high and off. The destructive enviros are there regardless. The chunks out of walls are there regardless. The diff is the number of particles that are falling and stay on the ground. On high it is totally overdone. On Medium it looks more realistic. Off IMHO I don't miss it much. Yeah it does add a neat effect, but doesn't change the gameplay at all. Frankly I found myself more into what was going on in both scenarios and unless I stopped to go... cool at the crap on the floor then I barely noticed it. Cloth physx were nicer, but also didn't miss it, but did like it.

I used that trick of the Cloth folder renaming/deleting and I found a nice boost in performance.

Before trying the mod, everything maxed plus APEX PhysX at med:

Min: 13.0
Max: 62.5
Ave: 25.3

After the mod, using the same APEX PhysX at med:

Min: 15.6
Max: 111.1
Ave: 46.5

It looks good but far from rival Batman AA, the particles are overdone and they bounce like if they were made of clay.

 

[CitanUzuki]

4890 Crossfire: 1920x1080 Everything but physX on. i7 920 at 4.3ghz

max: 250.0
avg: 57.1
min: 9.3

Same as above Physx on high

max: 76.9
avg: 17.5
min: 10.2

Single 480 GTX: Everything but physX

max: 250
avg: 60.6
min: 5.1

Same as above PhysX on high

max: 142.9
avg: 40.0
min: 4.2

 

[sandorski]

Min 7.5
Max 125
AVG 38

1600*1200 16AF 0AA

 

[toyota]

4890 Crossfire: 1920x1080 Everything but physX on. i7 920 at 4.3ghz

max: 250.0
avg: 57.1
min: 9.3

Same as above Physx on high

max: 76.9
avg: 17.5
min: 10.2

Single 480 GTX: Everything but physX

max: 250
avg: 60.6
min: 5.1

Same as above PhysX on high

max: 142.9
avg: 40.0
min: 4.2

what was the point in running it with physx on with the 4890 setup?

 

[CitanUzuki]

What do you mean? Why not?

 

[toyota]

What do you mean? Why not?

because all that did was run it on the cpu and we all know that's not playable on any game that uses advanced hardware physx.

 

[CitanUzuki]

Um, ok? What does that matter?

 

[toyota]

Um, ok? What does that matter?

well what or why does anything matter? lol. you posted the results and I commented about them. it seems there was no point in running physx on your ATI setup since it was just going to run on the cpu anyway.

 

[CitanUzuki]

well what or why does anything matter? lol. you posted the results and I commented about them. it seems there was no point in running physx on your ATI setup since it was just going to run on the cpu anyway.

I was interested to see what the crossfire(cpu if you like) set up could do with the setting on. Perhaps someone else will be interested as well.

 

[toyota]

I was interested to see what the crossfire(cpu if you like) set up could do with the setting on. Perhaps someone else will be interested as well.

any ATI user that enables physx without a dedicated physx card will be running the effects on their cpu and nobody knowledgeable would do that in the first place though.

 

[sandorski]

Decided to try it with PhysX On just to see if my Thuban offered anything extra. Apparently it does not. Same settings as before + PhysX

Min:4.7
Max:58.8
AVG:13.5

 

[CitanUzuki]

any ATI user that enables physx without a dedicated physx card will be running the effects on their cpu and nobody knowledgeable would do that in the first place though.

Quit being retarded, I said crossfire set-up. I realize that the ati gpu's will not be doing any of the physx calculations. Like I said, I am interested to see how well physx runs on the cpu, and maybe someone else will as well. Move along.

 

[toyota]

Quit being retarded, I said crossfire set-up. I realize that the ati gpu's will not be doing any of the physx calculations. Like I said, I am interested to see how well physx runs on the cpu, and maybe someone else will as well. Move along.

I can see running it on the cpu just out of curiosity but we are all capable of doing that ourselves.

 

[Barfo]

Screenshots/video anyone?

 

[toyota]

Mafia 2 Apex physics comparison video

http://www.youtube.com/watch?v=_gN3fgKlymI

sorry but I think its a complete waste of framerates just for that.

 

[topmounter]

Any idea why when I change any setting in the "video" menu they don't take? I get the same benchmark measurements no matter what changes I make and when I return to the "video" menu all the settings are back to the default.

Radeon HD 5770 / Gigabyte 8900GPA-UD3H / Phenom II x4 955 BE

 

[sandorski]

Mafia 2 Apex physics comparison video

http://www.youtube.com/watch?v=_gN3fgKlymI

sorry but I think its a complete waste of framerates just for that.

Better comparison IMO

Adds some nice Detail, but I can live without it.

 

[cusideabelincoln]

I can see running it on the cpu just out of curiosity but we are all capable of doing that ourselves.

Not everybody has the same processor............................

And... http://physxinfo.com/news/3628/mafia-ii-demo-tweaking-physx-performance/

It actually works so long as you're willing to give up cloth physics.

 

[cusideabelincoln]

Any idea why when I change any setting in the "video" menu they don't take? I get the same benchmark measurements no matter what changes I make and when I return to the "video" menu all the settings are back to the default.

Radeon HD 5770 / Gigabyte 8900GPA-UD3H / Phenom II x4 955 BE

I noticed that too. What you have to do is start a demo session and change the settings while you're officially playing the game. Then you can exit your session and run the benchmark. Repeat this process if you want to run the benchmark at different settings. PhysX of course requires you reload the game.

 

[toyota]

Not everybody has the same processor............................

And... http://physxinfo.com/news/3628/mafia-ii-demo-tweaking-physx-performance/

It actually works so long as you're willing to give up cloth physics.

there wont be more than a couple of fps difference between any modern cpus though. I ran it on my cpu and got 16.5 fps myself. the point was its not going to be playable in its standard form on the cpu.

 

[topmounter]

I noticed that too. What you have to do is start a demo session and change the settings while you're officially playing the game. Then you can exit your session and run the benchmark. Repeat this process if you want to run the benchmark at different settings. PhysX of course requires you reload the game.

Ah super, got it. I didn't realize I played through the demo on the highest settings (Physx off)... which is good since it felt smooth and perfectly playable... but my benchmark with those settings was "D" at an average of 34.9 FPS ("C" at 52.3 FPS on the default settings).

 

[tviceman]

Mafia 2 Apex physics comparison video

http://www.youtube.com/watch?v=_gN3fgKlymI

sorry but I think its a complete waste of framerates just for that.

Then turn it off and you don't have to worry about it. I appreciate the differentiation from the console versions.