GTX 460 random GPU utilization issues?

[Am*]

I'm running into an annoying problem. The GTX 460 I'm using, performs perfectly about 85% of the time, but sometimes, especially in Mafia II, GPU utilization just drops randomly, and it takes it anything from a second to a while, to come back to normal levels. The worst thing is, it doesn't relate at all to my CPU, which I thought could possibly be the culprit. For example, my CPU could be at 80% utilization and the low GPU usage bug happens, or the CPU could be at 95% and it works fine and vice versa. I'm running PhysX on high though, and in Nvidia control panel, set the GTX 460 to do all the PhysX work, so GPU utilization should never be dropping to such low levels (drops as low as 20% sometimes for no reason whatsoever).

The problem lies with the GPU though, which is what I found out when I ran other games, like Killing Floor, which sent my 460 mad, randomly changing between 1% GPU usage to 70% (just for the record, my single core Athlon + GTS 250 rig, ran Killing Floor better than this GTX 460, despite running at lower FPS, being a lot more stable and playable). Similar thing happens in Source games, when the screen would freeze for a second or two in-game, GPU utilization drops to 0% then returns to normal.

Any ideas on how to fix this, or if it's worth reporting this to Nvidia? Because it took them ages to fix the clockrate throttling down issue (from 260 to 275 drivers I believe, it was a major problem), now that it's finally been fixed, I'm hoping this problem is next on their list of upcoming fixes.

P.S. I've attached some screenshots, one showing PhysX working fine and GPU utilization at normal levels (how it should be), and two showing the utilization bug happening in the EXACT SAME PLACE in game, one taken about 20-30 seconds earlier showing the bug, than the other that shows it working normally. Settings are all maxed, 1440x900, vsync off.

 

[Am*]

Just a quick note, the GPU utilization seems to have the direct impact on in-game performance. If, for example, GPU utilization is at about 50%, I get roughly half the framerates that I would get when it works at 99% utilization (the two above screenshots show that). I wouldn't mind if the GPU was at 99% and I had a poor framerate with PhysX on High, because most reviews said this game is demanding on PhysX, but I don't see that at all. This card doesn't have a problem in heavy PhysX areas, in fact if anything, it performs much better in PhysX heavy areas with a lot of debris laying around.

 

[xvi]

I think the first thing most people are going to say is to check your drivers. If you don't have the latest, upgrade. If you do have the latest, downgrade.

 

[Am*]

I think the first thing most people are going to say is to check your drivers. If you don't have the latest, upgrade. If you do have the latest, downgrade.

I am running the latest drivers, and on a clean OS install with nothing but Steam and MSI Afterburner on it, so no viruses or bloatware. Downgrading would make more problems, since those drivers suffer from GPU downclocking issues.

 

[Am*]

Anyone...?!

 

[erocker]

Windows up to date?

DirectX updated?

All your Microsoft Visual C++'s up to date? (Done through Windows Update)

 

[Am*]

Windows up to date?

DirectX updated?

All your Microsoft Visual C++'s up to date? (Done through Windows Update)

Yes, don't know and yes.

Also tried a few drivers tricks -- forced some supersampling, which drove GPU usage up but hasn't fixed the random framedrops and GPU usage that's all over the place. Core/memory and shader clocks stay the same but GPU usage is screwed. Also lowered the pre-rendered frames to 0 and found out the GTX 460 is bottlenecking my Celeron with default settings (framerate in L4D2 and a few other games went up by 10-30FPS)...:shadedshu I wish I never bought this piece of crap. I tested this whole build at a friend's house with his 5870 and it ran everything like a boss on a clean partition. Did the same for my GTX 460 (clean install, updates, even tried over/underclocking) and it's still the same useless paperweight that it's been since the late 260 drivers. Gonna mess around with it a bit more, if nothing works, I'll just give it to a friend and go back to my trusty GTS 250 till I can get a 6950 or something. Was gonna upgrade to a 2500K, SSD and W7 64bit soon -- but with this same old broken drivers BS (thanks again, Nvidia), that's gonna have to wait until I get a useable graphics card first.

Also, included a screenshot of the GPU usage in MSI Afterburner while running Lead and Gold (an Nvidia/PhysX game..*facepalm*), with 2x supersampling and better texture filtering (forced in drivers), was stuttering between 60-150FPS in-game at the time the screenshot was taken...CPU was barely breaking a sweat (around 60% utilization), if anyone has any suggestions, I'm all ears.

 

[Am*]

Almost forgot about this thread, thought someone would come up with something...guess not.

More great news, now this piece of crap is artifacting on me after playing certain games -- fails to draw d3d device or some crap after running L4D2, grinds to a halt after artifacting in UT3, with performance being completely different each day.



I'll be getting rid of this piece of shit and switching back to AMD, many thanks Nshittia. Maybe in 2 or 3 more years time it'll be somewhat useful/stable and they'll get their act together, like they did with their 8000 series (almost a year after release, if not more).

 

[LifeOnMars]

Surely a celeron is going to struggle pushing a GTX 460?

 

[Am*]

Surely a celeron is going to struggle pushing a GTX 460?

Not really. If it didn't struggle to push a 5870 to 80%-99% usage in the same game, why in the hell would a crappier card struggle?

 

[LifeOnMars]

Fair enough. I used to have a GTX 460 with a 1090T and had a few issues. You may just have a bad card mate, hope you get it sorted.

 

[Am*]

Fair enough. I used to have a GTX 460 with a 1090T and had a few issues. You may just have a bad card mate, hope you get it sorted.

That's the worst part, I don't even think it's faulty. There's just no way. On the day I bought it, I Furmarked it for 2 days straight without any artifacts, or the GPU temperature going above 65C. While it's almost always performed like crap since the 260.XX drivers, it never had any stability issues until the current drivers, which I am suspecting to be the problem. I can run any OpenGL game on it or application and it performs just as it should (90-99% GPU usage, no stuttering at all and FPS quite higher than my GTS 250), yet anything DirectX and it stutters like hell, making almost every game unplayable. Don't know when or how it will ever be fixed, besides just getting a completely different card.

 

[James1o1o]

The problem is the CPU. Straight off, no matter what you say about your old card, a Celeron can not play Mafia 2.

 

[Am*]

The problem is the CPU. Straight off, no matter what you say about your old card, a Celeron can not play Mafia 2.

Then you're either blind or just plain ignorant. Instead of assuming by the brand sub-name, you might want to look at some reviews first. This Celeron has about 80% of the performance of an i3 2100 in games (and a hair slower than a G620, 2-5FPS difference at best between them in the real world), since the only thing setting them apart is 700MHz and a 1MB of cache (and different on-die graphics/Hyperthreading, that mean jacksquat in gaming). Mafia 2 maxed out with PhysX off on this card, runs between 65-117FPS, and when I tested it on with my friend's 5870, never got below 70FPS. I have to turn PhysX on with my 460 to get the framerate down so the screen tearing wouldn't get so bad. Is 60FPS+ what you consider unplayable?

 

[James1o1o]

Then you're either blind or just plain ignorant. Instead of assuming by the brand sub-name, you might want to look at some reviews first. This Celeron has about 80% of the performance of an i3 2100 in games (and a hair slower than a G620, 2-5FPS difference at best between them in the real world), since the only thing setting them apart is 700MHz and a 1MB of cache (and different on-die graphics/Hyperthreading, that mean jacksquat in gaming). Mafia 2 maxed out with PhysX off on this card, runs between 65-117FPS, and when I tested it on with my friend's 5870, never got below 70FPS. I have to turn PhysX on with my 460 to get the framerate down so the screen tearing wouldn't get so bad. Is 60FPS+ what you consider unplayable?

Your looking at a very "unheavy" point in terms of CPU ultilization, I myself have only recently upgraded my CPU, I went from a Q8200 to a Core i5 2500k. I would say a Q8200 is better than a celeron, my framerate in GTA IV went up well in excess of 200%. The point under the train bridge is definetly the most heavy point in the game, went from about 16-20fps to well over 60fps.

 

[Am*]

Your looking at a very "unheavy" point in terms of CPU ultilization, I myself have only recently upgraded my CPU, I went from a Q8200 to a Core i5 2500k. I would say a Q8200 is better than a celeron, my framerate in GTA IV went up well in excess of 200%. The point under the train bridge is definetly the most heavy point in the game, went from about 16-20fps to well over 60fps.

Yes, it was a "heavy" point, as it was one of the most open areas in the game. If my CPU was the problem, my FPS would plummet there and then, when you try playing it on a underspec'd CPU, you'll know.

GTAIV has nothing to do with Mafia 2. Mafia 2 is an optimized PC game, GTA IV is a crappy console port made for Xbox 360's tricore.

FYI, your Q8200 is slower than my Celeron in anything dual threaded, which happens to be Mafia 2.

At least LOOK at the screens I attached to my other posts, before you post...

 

[micropage7]

does that change much if you lower the resolution?
if it happen at the same time in the game, i guess its not the card

 

[Am*]

does that change much if you lower the resolution?
if it happen at the same time in the game, i guess its not the card

I can try lowering it, doubt it will do anything though. Why lower it? As I said before, I'm trying to INCREASE GPU usage, not do the opposite.

 

[James1o1o]

I can try lowering it, doubt it will do anything though. Why lower it? As I said before, I'm trying to INCREASE GPU usage, not do the opposite.

He is trying to see if your CPU is bottlenecking.

Look at the benchmarks.

http://www.cpubenchmark.net/cpu_lookup.php?cpu=Intel+Celeron+G530+@+2.40GHz

Your CPU is lower than alot of MOBILE CPU's.

FYI, your Q8200 is slower than my Celeron in anything dual threaded, which happens to be Mafia 2.

The Q8200 chip had similar performance to that of a Q6600, in the benchmarks posted above, the Q6600 isn't even posted, looking at the top of the list, its in the higher end listing. Celeron is NOT a gaming CPU. The entire point of the Celeron brand is that its designed for budget computers.

 

[Am*]

He is trying to see if your CPU is bottlenecking.

Look at the benchmarks.

http://www.cpubenchmark.net/cpu_lookup.php?cpu=Intel+Celeron+G530+@+2.40GHz

Your CPU is lower than alot of MOBILE CPU's.




The Q8200 chip had similar performance to that of a Q6600, in the benchmarks posted above, the Q6600 isn't even posted, looking at the top of the list, its in the higher end listing. Celeron is NOT a gaming CPU. The entire point of the Celeron brand is that its designed for budget computers.

You're not seriously suggesting for that crappy in-accurate benchmark to reflect real-world performance, are you? FX-8150 scores 4x higher, if going by that useless benchmark; care to show me a game (ANY game) that it can run that much faster with my same GPU?

The Q6600, is also slower than my CPU in dual threaded games, since mine's around a stock E8400s performance level.

Budget has nothing to do with anything, I have no idea what you're trying to get across. If you're another nutter suggesting an upgrade to a 2500K for dual threaded games, save it. I already ran it on a 2500K and the same crap happens -- the only difference is, it recovers from the GPU usage drop/freeze and stuttering a few microseconds faster.

 

[James1o1o]

Now the denial sets in. I just want to point out that none of your benchmarks that you have posted show your CPU at all. Brand does have something to do with it, whether you like it or not, the entire point of "brands" is the fact that some are better than others.

If you're another nutter suggesting an upgrade to a 2500K for dual threaded games..

Yes I am. Do not be ignorant and completely ignore the obvious. You do a single benchmark, dual threaded whatever, a stock 2500k will beat your CPU every single time.

EDIT: That "crappy in-accurate benchmark" is one of the fundamental benchmarks used on the internet. It shows raw performance of a CPU, which yes I do agree is not the same as in-game performance but its close enough. The way you are making it sound is even though your CPU is slower than most of the mobile processors you are trying to claim it would still match a Core i5 in performance? No dude, your just talking out your back end now.

 

[Am*]

Now the denial sets in. I just want to point out that none of your benchmarks that you have posted show your CPU at all. Brand does have something to do with it, whether you like it or not, the entire point of "brands" is the fact that some are better than others.




Yes I am. Do not be ignorant and completely ignore the obvious. You do a single benchmark, dual threaded whatever, a stock 2500k will beat your CPU every single time.

EDIT: That "crappy in-accurate benchmark" is one of the fundamental benchmarks used on the internet. It shows raw performance of a CPU, which yes I do agree is not the same as in-game performance but its close enough. The way you are making it sound is even though your CPU is slower than most of the mobile processors you are trying to claim it would still match a Core i5 in performance? No dude, your just talking out your back end now.

Then don't post at all, and you can leave your elitist, ignorant trolling at the door. My CPU to match a core i5? You're the one talking out of your arse now, show me where I said that. What you still don't get through your thick skull, is that I've already tested my system with a different card and my card on a different system, and the problem isn't the framerate, it's the random GPU usage drops that cause insane screen tear/freezing and stuttering. You completely ignore everything I posted about this, instead picking words to start an argument. Take your e-peen crap to someone else's thread who asks for it.



FYI, your benchmark is still worthless. I came from a Phenom 9550, that's shown to be considerably faster on your benchmark than my Celeron, when my current CPU stomps my old Phenom in any game or real-world application, period. Here's how my CPU does in games now, compared to a i3, which is not in any way considered a "slow" CPU.

My Phenom struggled even in dual-threaded games at stock, and would barely scrape 40FPS at best in Crysis 2.

 

[James1o1o]

Then don't post at all, and you can leave your elitist, ignorant trolling at the door. My CPU to match a core i5? You're the one talking out of your arse now, show me where I said that.

Potential disambiguation with your post but nonetheless it sounds like you are trying to say a 2500k performs the same as your Celeron.

If you're another nutter suggesting an upgrade to a 2500K for dual threaded games, save it. I already ran it on a 2500K and the same crap happens -- the only difference is, it recovers from the GPU usage drop/freeze and stuttering a few microseconds faster.

FYI, your benchmark is still worthless. I came from a Phenom 9550, that's shown to be considerably faster on your benchmark than my Celeron, when my current CPU stomps my old Phenom in any game, period. Here's how my CPU does in games now, compared to a i3, which is not in any way considered a "slow" CPU.

http://www.xbitlabs.com/images/cpu/celeron-g540-g440/first/crysis.png

Care to troll some more?

SIGH. First off straight away, your using a game that is a completely GPU dominant game, not CPU. Secondly, you have just automatically shown me that your CPU is 1/3rd the performance of the lowest end sandy bridge processor. You have done nothing except validate my own posts. Your CPU is shit. And I'm sorry for cussing because I'm not 'that' guy but its the only way you are going to get it through YOUR thick skull that your processor is the problem because it is designed as a budget office processor and not to be paired with a high end graphics card.

 

[Am*]

Potential disambiguation with your post but nonetheless it sounds like you are trying to say a 2500k performs the same as your Celeron.






SIGH. First off straight away, your using a game that is a completely GPU dominant game, not CPU. Secondly, you have just automatically shown me that your CPU is 1/3rd the performance of the lowest end sandy bridge processor. You have done nothing except validate my own posts. Your CPU is shit. And I'm sorry for cussing because I'm not 'that' guy but its the only way you are going to get it through YOUR thick skull that your processor is the problem because it is designed as a budget office processor and not to be paired with a high end graphics card.

Are you blind? The G440 you're looking at, numbnuts, is a single core at 1.6GHz. My G530, is a dual core, identical to the G540, only 100MHz slower (I can raise it to that same frequency anyway). The quote you handpicked yet again to try and start shit, it seems you need translated to you: NOWHERE does it say it performs the same, it says I get the same problems (not framerates, Sherlock). Got that through YOUR thick skull this time?

 

[James1o1o]

Yup I'm done here, good luck with your problem.