I really want to have the GPU physics enabled in this game but I just can't seem to get it to the point where the game is actually playable. I did a bunch of testing to see if maybe I could come up with some information that might be useful for the devs. I have a GTX 690 and a GTX 590 at my disposal. All of the tests were done in Amerish WG with about 30 people scattered around. I have all graphics settings completely maxed out with FOV at 60. Here are the configurations I've tried: First of all, GPU physics off with the GTX 690 in SLI results in 60+ fps at all times. Everything is very smooth. 1. The ideal configuration, GTX 690, SLI enabled, PhysX processor chosen as automatic in NVIDIA Control Panel (GTX 690 (B) is automatically chosen). With this configuration my fps is always below 40 when there are GPU particles in view, usually hovering around 25 - 35. Along with this, all movement in the game is very jerky (characters, vehicles, particles) and it's pretty much unplayable because of this alone. Not looking at GPU particles, the game plays similar to having GPU physics off. 2. GTX 690 with SLI off, GTX 690 (B) dedicated to PhysX. My fps with this configuration is pretty much the same as with configuration 1, only it maxes out at about 45 fps even when there are no particles around. All movement is smooth with this configuration. This is playable in the WG, but GPU 1 is being used 99%. This configuration is unplayable in large battles because the single card can't handle it. 3. GTX 690 with SLI on, CPU chosen as PhysX processor. With this configuration, the particle effects are glitched out. The teleporter particles spin, but they don't move upwards. Movement is fine but I get microstutter that is not present with GPU physics off. My fps while looking at physics particles usually starts out at about 30 fps and then gradually works up to 60 fps and stays there. If I start moving around and looking elsewhere it will lower back down to around 30 again then gradually climb back up. 4. GTX 690 with SLI on, GTX 590 (A) dedicated to PhysX. This configuration gives exactly the same results as configuration 1 and is unplayable. I'm not really sure what else to try. Configuration 2 seems to be the best, although the fps is pretty unstable and the particles are messed up so there really isn't a point. I should also mention I'm using a 2700k at 4.5ghz IBT stable. I have CPU parking completely off.
Unfortunately GPU PhysX is just not useable for me. My FPS stays around 40-45 but it introduces some kind of microstutter that makes the combat all but unplayable.
My suggestion (fixed PhysX performance issues for me on a much weaker rig then the OP's) would be to use 1 GPU, remove all SLI and dedicated PhysX cards, delete your user options, and work from the ground up. Tweak 1 setting at a time until you see a performance loss, but do it with GPU PhysX turned on from the moment after you delete your current useropts.ini
Alright so as per your suggestion, I did some more testing with a single card and a new UserSettings.ini. This time there were about 5 - 10 people at Amerish VS WG. During the tests I made sure to only have a jump pad and a teleporter in view. With Highest Settings: SLI off, GPU particles off, 60 fps SLI off, GPU particles on, GTX 690 (A) used for PhysX and rendering. Less than 40 fps SLI off, GPU particles on, GTX 690 (B) dedicated to PhysX. ~47 fps SLI off, GPU particles on, CPU used for PhysX. ~52 fps (particles are glitched but visible) With Lowest Settings: SLI off, GPU particles off. 130 fps SLI off, GPU particles on, GTX 690 (A) used for PhysX and rendering. 70 - 75 fps SLI off, GPU particles on, GTX 690 (B) dedicated to PhysX. 100 fps SLI off, GPU particles on, CPU used for PhysX. 112 fps After seeing that CPU PhysX was giving me better performance, I figured there must be something wrong here. I swapped out the 690 for the 590 and now I get 85 fps with everything maxed out and GPU physics on looking at the same spot as the tests above. So now I'm looking at the possibility that I have a defective GTX 690. If theres anybody else with a 690 kicking around, do you see the same results as me? EDIT: I should also point out that recently I've been getting horrible marks in the new 3DMark lately. I've heard that there was a Microsoft update to DirectX that might have messed some stuff up. I've even tried older drivers as well as complete display driver uninstall and clean.
I have a 690GTX as well, and I also have insane framerates (no stuttering, very smooth graphics but no idea of what the numbers are). However, I'm not really a technogeek and I haven't tweaked it at all, just running 314.22 drivers in stock setup. But my experience has been as soon as the Phys-X update and I turned on the particle effect, I have been crashing out of the game abruptly over and over and over again. So today, I finally got so p'ed off I unchecked it and restarted the game. First session went well with no crash. I will save this thread and see if some of the suggestions help. I greatly appreciate people sharing them. Edit: If you tell me how to check my exact framerates perhaps I can provide more data for you. I guess you change the GPU core configuration thru the nvidia control panel? BTW, my CPU is an Intel i7-3770K@3.50 GHz but again, I don't fiddle with it, no overclock.
I have a single (although oc'd) 670, which only runs at about 80% usage with ultra on every non cpu limiting setting, and my cpu is the bottle neck as long as the fps stays below 110-120 fps, so my conclusion from this, and from your account above, the physx in this game doesn't like 690's, or there is something wrong with it. As stated above though, the physx make the game crash every two hours or so...
sli configs don't work to owell in this game let me explain. at the WG youll see your GPU's 100% load on each one that's great. now go too a big fight and youll notice that GPU useage drops. now before people go its the cpu ect ect it isent. For some Strange reason in big fights sli useage will lower and you will still be GPU bound don't ask me why ive no idea I have GTX 580's in sli and this happens too me I get maximum higher fps overall but lower fps overall over a single card. secondly. with gpu physics turned on I have the horrible stutter and frame loss also but heres something intresting. on windows 7 I can run with one card only no dedication for phys x on my other card and I get 0 fps loss I get a steady 60 fps at WG and 40ish in fights and I see cpu bound very very often. now with windows 8 and this very same setup I see horrible stutter and terrible fps dedicated card or not for phys x so I cant pinpoint the problem myself hope this helps you out
Can someone explain SLI to me? I know it's two nvidia GPU's running in tandem, however, with the 690GTX, the two GPUS are on one card (which is why I spent the cash for it, so I didn't have to mess with anything and figure it all out). But apparently I can still take the 690 out of SLI mode? Guess I need to open the nvidia control panel and have a look - just the thing I dread doing. LOL
Sure. The long and short of it - like you said - is that the 690 is two GPU's shoved on to one card. Those two GPU's still behave like two different graphics cards, there's even an SLI link built into the card. In theory, it's a cheaper way to get SLI's performance in a smaller (and cheaper) package than two 670's. In practice, each GPU can only use half the 16xPCIE link connecting it to the CPU, and the GPU's are downclocked a bit to get them running without melting. Oddly, config 4 should be giving you the best performance out of all of those options. But that said, SLI has always been a bit of a gamble - good in theory, but a lot of games just don't benefit from it. For reference, I'm running a 560ti as primary, with a 430 for physX. I don't lose performance using physX, even though the 430 is much (like, 20x) weaker than your 590.
The sweet spot for me in this game is to go in your nvidia profile and tell your machine to use single card. There is absolutely no reason to use 2 cards for the FPS boost in this game, as you wont get one. I actually get better fps by using 1 GTX 670 for the game and the second defaults to doing the physx since SLI is disabled. I get great FPS, well great for this game. Even with all the teleporters on screen in a spawn room or warp gate, i still get 80 fps or more. The second 670 laughs at the physx duty, and the other card is far more than enough for this game. With the 2600K running at 4.6 ghz, i was able to sustain a playable 50 fps in the biggest battles. In fact thats what you need the most in this game, pure clock speed. Going to see if i can get this CPU up to as close to 5 ghz as i can, as this game gets huge fps boosts with every clock bump. TLDR: Just running 2 cards, one for the game and one for physx yielded me the best performance, and i can enjoy the physx with no stutters and crashes.
I typically use EVGA Precision X for seeing my FPS but the game has the ability to show you it by using Alt + F. It will show you your FPS at the bottom left corner of the screen. I'm especially interested to see if you get similar results with your 690. I guess I'll have to do some testing with my 590 to see how that differs while I wait for some responses from 690 users.
I just did a few tests with my 590. Contrary to what my 690 would lead me to believe, the SLI scaling in this game is actually quite good. Again these tests are done in Amerish VS WG with about 20 people there this time. I never had anyone on screen in these tests, only the same teleporter and elevator. Highest Settings: SLI on, GPU physics off, 100 fps SLI on, GPU physics on, 590 (B) used for PhysX, 85 fps SLI on, GPU physics on, CPU used for PhysX, 95 fps SLI off, GPU physics off, 43 fps SLI off, GPU physics on, 590 (B) dedicated to PhysX, 43 fps SLI off, GPU physics on, 590 (A) used for rendering and PhysX, 38 fps SLI off, GPU physics on, CPU rendering PhysX, 40 fps I also played in some large battles with lots of elevators with everything on highest, SLI on, GPU physics on, 590 (B) used for PhysX. Everything was actually very smooth. I don't think it ever dropped below about 45 fps. The results with SLI off with the 690 were better across the board, but still not nearly as good as it should be. The 690 with SLI off should be twice as good as a 590 with SLI off. EDIT: I realized I never did a control test with only the elevator and teleporter in view with the 690 in SLI and GPU physics off. I get about 110 fps with that configuration.
I'll throw some tips personally regarding this overall issue. I run a GTX690 with a GTX 460 being dedicated to PhysX. * Applying PhysX to the processor will only give you weaker frame-rates to the main resolution cards, so that is unacceptable. * Applying PhysX to GPU "A or B" is not suggested due to a single GPU will easily reach 90%+ usage while the other GPU will reaming at around 50% The whole point of this card to take advantage of the SLI capability in this single card. * Applying PhysX to a dedicated card will relieve pressure off the GTX 690 and the GPU usage percentages should always be equal. - Keep in mind that the GPU usage can achieve over 50%, and that combined with PhysX can basically bottleneck the GPU, thus cause massive frame-rate drops. If you want me to test in various areas, tell me where. * Here is a image at a warpgate. - Ultra settings
I dont know why you guys think SLI scaling is good in a game where your cpu determines the FPS you get. Sitting in the warp gate was never the problem in the first place. I have seen my FPS rise by disabling my 670 4gbs in sli and just using them as one card for the game and one for the physx. using SLI results in worse fps in every single battle, and also dont even bother turning on the physx. Turning off SLI meant superior FPS in battles, and then i could use physx too. I might try throwing in my spare GTX 480 and use it as a physx card to see if this changes anything.
Massive battles. * Stable 60 FPS. If you can achieve higher than 60 FPS, in massive battles on a single GPU. * Then show some proof. Here is a couple screenshots from a couple battles I was just in today. 63 FPS 62 FPS 65 FPS
I am sure i could take some screens of worse FPS than 60 if i tried. I could also show 120 fps with a single card too. Since this game's performance drastically changes based on what you are looking at, that does not mean much. I am going to see what happens when i bring my CPU past 5 ghz. I dont think a 2600k at 4.6 ghz is high enough for my SLI to be used. Btw you can easily reach 60 fps on a single card, so why do you use SLI? Your own screen shot proves me right, your GPUs are at 50 % useage. Thats the equivalent of your SLI doing exactly nothing. Now disable it and try one, youll have even HIGHER fps in those same battles.
i believe, the GPU is used for rendering landscape while the cpu renders units and dynamic objects... So everytime i am looking at players, fps-indicatior shows CPU - when facing only landscape it shows GPU
Update: Since turning off the particle effects Phys-X checkbox, my game has not crashed once, and I played for long periods (2 hrs+ for multiple sessions) during double XP weekend. I probably won't investigate this issue further by tweaking, since I don't miss the particle effects and don't need the "eye candy". I won't be re-checking the option, I will simply leave it off unless reliable reports of it being fixed come in.
