chump7431

interesting stats.  but i have 2 480's plus a 285 and i dont get close to 81fps.  and that is with one more 480 then what they used.

Is this during the demo or the benchmark? You won't get close to 81 fps in the benchmark, but the demo should run faster than that with your rig.

I'm only getting 35 fps avg in the benchmark is the physx not working properly in this game ? I'm only using one GTX 480 atm btw.

boredgunner

Saiyanjin2

I'm only getting 35 fps avg in the benchmark is the physx not working properly in this game ? I'm only using one GTX 480 atm btw.

Yeah that's better than what I got at max settings but no AA at 1920 x 1080 with one GTX 480.  Gotta get my new CPU quickly.

Yeah i installed the new physx drivers and the game play is a lot smother now, before it was a bit choppy in the benchmark. But I still ended with 36 FPS with everything at max. Thought when i play the game my frame rate is a lot better in the 50-60 for the most part.

kdurrell

Physx is fail unless they work together to get it to become a standard. By that I mean working with AMD and Intel. To me it only takes looking back to good old 3dfx and the use of Glide to really see how this story will play out in the long run.
 
As for Mafia II I haven't played the demo but would wonder if the full game might be a bit less taxing. The original is one of the few games I have beat multiple times and still play now and then.

Ya in a perfect world there would be one standard physics library and API that every dev and game on the PC used, unfortunately that's just not the reality of it as there's a lot of underlying egos and motives at play.  Even when it comes to game engines, platforms or programming APIs we see not everyone is on the same page, with numerous competing offerings out there.  The key is to support as many options as possible as simply as possible to make your offering better than the others.
 
Nvidia owns PhysX so obviously their interest is to expand the capabilities possible beyond the CPU to enhance its capabilities on their GPU, and they've done that.  Its clearly obvious the GPU PhysX used in games far surpasses anything we've seen done on the CPU, but for those who can only run CPU PhysX, the results are competent and comparable to anything else out there.   Intel owns Havok, so obviously they have no interest whatsoever in leveraging the abilities of the GPU, either ATI or Nvidia, as that would simply highlight the disparity in compute power between their CPUs and GPUs.  Any hope of GPU-accelerated Havok, which has been demo'd via OpenCL on both Nvidia and ATI GPUs, died with Larrabee.  ATI/AMD is stuck in the middle but they've chosen to support nothing as they would rather not support Nvidia's GPU tech openly while maintaining parity with their other competitor, Intel, when it comes to CPU support by simply sitting on their hands.
 
Also, standard APIs have been in place for some time now, both OpenCL and DirectX 11 (DirectCompute) offer a standard API path for all compliant hardware, both ATI/Nvidia and AMD/Intel, the problem is there are no tools in place to take advantage of it.  If given the choice, these devs are going to go with the most robust solutions available that require the least amount of effort, so that's been an easy decision: either PhysX or Havok, as they not only support x86 PCs, but all of the major consoles as well. 
 
I guess in the end Nvidia could sit around and wait for GPU-accelerated physics to fall in their lap, especially since both OpenCL and DirectX 11 were only ratified and rolled-out in the past 12 months or so and didn't exist when Nvidia offered GPU accelerated PhysX, but instead they decided to go ahead and offer the feature to all of their DX10+ GPUs in supported configurations, free of charge, using their existing compute architecture.  In the meantime, we don't have to wait for that great day that a physics standard arrives, we get to enjoy much nicer PhysX effects and eye-candy in some great games at no additional charge today, so for that I'm thankful.

BranC85

Does everybody else have a window saying "windows installer" implying a first time startup, and a stupid UAC message every time you start this game? I had it happen with the demo but i never thought much about it, but now with the full version which unfortunately requires steam(ugh) it does the same thing

Nope, not happening here.

This game is very hard on my system also.I am only getting around 60 fps at 2560x1600 with all settings maxed.I am getting the dumb steam install message at launce too.Is anybody else getting random crashes to desktop? The crashing is driving me crazy.I have cleaned drivers and physics and still getting ocasional crashes.

I haven't touched the game in a while, but what I do remember in the demo is that the problem was major inconsistency with framerates. Walking in the house, it'd chug down big time, also while walking out with so many people... but then, when you get in a car and drive a bit I hop right up to 60+ fps and performance generally seems ok.     

I got mad at this game last night and decided to do a format and clean install of windows and mafia 2.After reinstalling my fps shot up quite a bit to 83 on benchmark but I still got crashes.I decided to run everest for a while and low and behold I was getting memory errors.I just bumped my dim voltage one notch and that fixed my problems.I hope my memory or memory controller isn,t starting to get weak.

Delano79

After downloading the latest PhysX driver from Steam and updating, my performance is now quite good
 
I'm getting an average of 58 FPS at 2560x1600, max settings..
 
One thing I've noticed though, is that the recommended PhysX card for Apex high (a GTX 285) is absolute overkill.  Running the benchmark, and playing the demo, my GTS 250 never went above 67% usage, so it wasn't even being taxed.

Sorry to sound so noobish, just got back into PC gaming after 15 years or so..anyways, where do you get the steam physx driver?  I cant locate it and is it any different that the one from Nvidia?
 
Thanks-
 

BranC85

rwoodburn1

I got mad at this game last night and decided to do a format and clean install of windows and mafia 2.After reinstalling my fps shot up quite a bit to 83 on benchmark but I still got crashes.I decided to run everest for a while and low and behold I was getting memory errors.I just bumped my dim voltage one notch and that fixed my problems.I hope my memory or memory controller isn,t starting to get weak.

Have you still not had any crashing since upping the DIMM voltage? Did your settings pass memtest before the increase or did you have errors with it too? The crashing is still persisting for me. I have not tweaked any voltages. I know I am pushing my ram a bit hard but my settings have passed memtest multiple times.

I wonder why the new install helped your fps so much?? I know a clean install has less load on the CPU but a difference that drastic is pretty absurd!

 
 
I am not a big fan of memtest so No I havent tried it.I use Intelburntest and Everest and yes my system is 100 percent stable now without a single crash.As far as increasing frames thats kind of a no brainer.A clean fresh,stable system will always perform better than a cluttered system that has crashed multiple times.

This has to be the most utilization I have ever seen done on a dedicated PhysX card. I have two GTX 465 SC's in SLi and running a GTX 470 SC as a dedicated PhysX card, which is gpu 3 in the picture.

Q56_Monster

Pretty good optimization of SLI I would say! 

I wonder if I should dedicate one of my 480's to physx....humm  Or keep them in SLI and use a 9800 for physx.  Would 9800 be a good dedicated physx card?  Only other card I have laying around is an 8800 ultra.

I saw no difference in running two GTX 470's in SLI with one dedicated to PhysX and running in TRI SLI with PhysX set to Auto. With TRI SLI I get consistent GPU usage of about 40-45% across all 3 GPUs with everything set on the highest settings. A significant portion of the PhysX runs on the CPU.

I get an average of 34 fps with 1680x1050, Medium Physx, all options on high.

Oddly for most of the benchmark my GPU has a utilization of about 40% and all cores of the CPU are at 50%. So, if neither the GPU or the CPU are fully utilized what would the bottleneck be here?

HeavyHemi

Actually I'm going to completely retract what I just posted. I had entirely different results this time. With TRI SLI 54 FPS all settings maxed at 1920x1080. With SLI and GPU3 dedicated to PhysX I had 71.4 FPS. Here's two bench runs one after the other with SLI GPU3 dedicated to PhysX...

Very interesting...I thought that may be the case.  Thanks Hemi.