Certain Fermi chips 'downclocking' after extended playing?

[HollowRopes]

This is something that I'd noticed on three different GTX cards.. a Galaxy GTX 460 768MB, a PNY GTX 460 OC, and a Zotac GTX 465. The cards run the games fine for an hour or so, then the frame rate drops tremendously. Using GPU-Z, it shows that the clocks speeds for the core/shader/memory have all dropped by half, if not more. This has happened on two different machines. Tried different drivers, still the same issue with all three cards. Thee's no way all three cards are malfunctioning. Anyone know what gives, or does anyone have this same problem?

 

[HollowRopes]

By the way, using Afterburner to change the clocks back doesn't work; it shows that the clocks get set back, but GPU-Z still shows the clocks as being extremely low (i.e. 350/700/1200). The only way to get the clocks to go back to normal is to reset, to my knowledge.

 

[SickBeast]

Apparently if the Fermi GPUs are lightly loaded, they will downclock. You can change the power management settings in the drivers if you want, and it should prevent this.

 

[HollowRopes]

Okay, I have been using "Prefer Maximum Performance". Should I use "Adaptive"?

 

[HollowRopes]

Okay apparently "Adaptive" didn't work; frames dropped not even 3 minutes in Stalker: SHoC, and that's an older game.

 

[GodisanAtheist]

Yeah actually "Prefer Maximum performance" is the setting I used to fix the problem. How hot are your cards getting, make sure you don't have PCI-E link state management nonsense set to low in your windows power settings, and are you running multi-monitor (moot point as you had PMP enabled, but a question anyway).

This seems to happen when something on your desktop takes precedence over the game you're playing and the card drops back down to desktop 2D clocks.

 

[Silverforce11]

If you dont have proper ventilation in your case, the GPUs will throttle when its overheated. Otherwise i don't know what could be going on.

 

[Idontcare]

This is something that I'd noticed on three different GTX cards.. a Galaxy GTX 460 768MB, a PNY GTX 460 OC, and a Zotac GTX 465. The cards run the games fine for an hour or so, then the frame rate drops tremendously. Using GPU-Z, it shows that the clocks speeds for the core/shader/memory have all dropped by half, if not more. This has happened on two different machines. Tried different drivers, still the same issue with all three cards. Thee's no way all three cards are malfunctioning. Anyone know what gives, or does anyone have this same problem?

It's a setting in the drivers config in windows, I had the same issue. I'll try to find the thread/post where this was documented. It is solvable.

 

[HollowRopes]

My case is completely open; skeleton like, with two small case fans blowing towards the PCI-E slots. Max temps are 76 C. GPU, 49 C. PCB. Only using the single-monitor compatibility mode. I'll check my power management options in Windows but I'm pretty sure I have everything set for max.

Everything is set to "Gamer" mode for power settings, and the PCI-E link is set to "off" rather than power saving mode.

 

[Idontcare]

Found the thread, I kinda feel embarrassed that I had forgotten it was my own thread (lol, weaksauce memory).

http://forums.anandtech.com/showthread.php?t=2140877

 

[HollowRopes]

Thanks a lot Idontcare, I will definitely look into this.

 

[HollowRopes]

Erm... Seems like there's a solution to 'a' problem, but not mine. See, I don't use vsync and I don't believe in using it unless you can always have frames above your refresh rate and a gigantic monitor. Therefore, my vsync is always turned off and I have no desire to turn it on. This work-around is by enabling vsync and manipulating the power management and global power management modes. I'm trying to get around this without enabling vsync, I can't take a performance hit that bad as I'm only running a Core 2 Duo at 3.6GHz with 4GB of PC1333 RAM.

 

[MangoX]

Are you overclocking? I used to run a 460, and with a heavy overclock that "seemed" to be stable (none of the usual issues occur like artifacts, crashes) the core would clock down to 500mhz (or around there, I forget) and stay there until I rebooted the PC.

 

[BD231]

A lot of people are saying it's game specific, what titles have you noticed this problem on?

NBA 2k11 for me will only utilize 40-50% of my GPU no matter what I do, which is upsetting, but nVidia is well aware it's happening and they don't do anything about it. In fact these issues are all over the nVidia forums and I've not found a single thread that had a fix of any kind, nor does nVidia respond to the complaints on the matter.

DX9 titles seem to fair the worst. Just hard to pinpoint what your issue is because you say the titles run at full GPU usage for an hour beforehand, which isn't typical of the GPU usage issues that are rampant on the nVidia forums.

 

[HollowRopes]

I'm lightly overclocking (775/1550/1800) with all three cards... Could that be the issue? Friends of mine have similar cards running way higher than this and they're not having this problem...

The games that I'm playing are the Call of Duty series, Stalker series, Metro 2033, Just Cause 2, Dirt 2, and.. that's about it. Occasionally I'll throw in another game and try it out but it always ends up as the same scenario, frames dropping to the point of it being unplayable. Also, I'm limited to 1280x1024 as my max resolution as I only have a 17 inch monitor.

 

[3DVagabond]

I'm lightly overclocking (775/1550/1800) with all three cards... Could that be the issue? Friends of mine have similar cards running way higher than this and they're not having this problem...

The games that I'm playing are the Call of Duty series, Stalker series, Metro 2033, Just Cause 2, Dirt 2, and.. that's about it. Occasionally I'll throw in another game and try it out but it always ends up as the same scenario, frames dropping to the point of it being unplayable. Also, I'm limited to 1280x1024 as my max resolution as I only have a 17 inch monitor.

When trying to troubleshoot the first thing you should do is return everything to stock settings. While the odds of you having 3 cards that wouldn't do 775MHz is really remote, (although the rhetoric that they'll all do 850 EASY is pure fantasy) you still need to do the basics. Make sure that Windows, direct x, drivers, are all up to date. Apply the latest game patches. Operate everything at factory settings.

 

[Will Robinson]

That can't be right because HP and some others here were so adamant that they would all do 800+ and were fantastic overclockers which is why they just pwned the ATI/AMD equivalent.

 

[Arkadrel]

This is something that I'd noticed on three different GTX cards.. a Galaxy GTX 460 768MB, a PNY GTX 460 OC, and a Zotac GTX 465. The cards run the games fine for an hour or so, then the frame rate drops tremendously. Using GPU-Z, it shows that the clocks speeds for the core/shader/memory have all dropped by half, if not more. This has happened on two different machines. Tried different drivers, still the same issue with all three cards. Thee's no way all three cards are malfunctioning. Anyone know what gives, or does anyone have this same problem?

Someone has bad air flow in his pc case(s)?

Sounds like a heat issue anyways.. but I doubt it, the 460s arnt really that hot cards. You must have really horrible case airflow in both of them, only thing that makes sense.

Unless its like a driver issue or something with a few titles as some others have said.

 

[sandorski]

Another Vote for Case Cooling being the likely culprit. It's not just sufficient to have air going to the Vidcard, but you also need sufficient venting of Hot Air out the Case. Just 3 weeks ago I had to Upgrade my Case as my old TT XaserIII was unable to vent Hot air fast enough with a HD 5870 running in it. After awhile my CPU and Mobo Chipset began overheating. Switched to new Case, Zalman Z9 Plus, and all's been fine since.

 

[Idontcare]

Erm... Seems like there's a solution to 'a' problem, but not mine. See, I don't use vsync and I don't believe in using it unless you can always have frames above your refresh rate and a gigantic monitor. Therefore, my vsync is always turned off and I have no desire to turn it on. This work-around is by enabling vsync and manipulating the power management and global power management modes. I'm trying to get around this without enabling vsync, I can't take a performance hit that bad as I'm only running a Core 2 Duo at 3.6GHz with 4GB of PC1333 RAM.

Unless I am horribly misunderstanding your post then I think you are misunderstanding the thread I linked.

My issue was I that I could not run w/vsync unless I changed my settings...but the problem was the same.

Did you try the graphics driver changes recommended by Notty22 in the thread? I had to make the changes to get my v-card from getting stuck at low clockspeeds, regardless the Vsync choice.

(Vsync is only mentioned in the thread because I was mistakenly thinking vsync was the cause)

 

[bryanW1995]

I've never noticed this before, but last night I played DAO for about 3 hours then decided to try out DA2 some more. I went to the exact spot where I benchmarked the game to compare different drivers the other day, and my frames went all the way down to 10 fps (from 30.22 avg the other day). I didn't check the afterburner clock settings, gpu-z, etc, but a reboot did solve my issue. I haven't used gpu-z in a while, guess I'll reinstall it and check it out if I see that again.

 

[HollowRopes]

Idontcare;

Sorry for the misunderstanding; I thought you were specifically referring to this happening while vsync was on. I am now going to try the driver fix. I'l report back.

 

[96Firebird]

That can't be right because HP and some others here were so adamant that they would all do 800+ and were fantastic overclockers which is why they just pwned the ATI/AMD equivalent.

You've really helped the OP with his problem! :awe:

I'm guessing this is something like what GodisanAtheist said, where something in Windows is taking precedence over the game and clocking the GPU down. Look for any new processes that pop up while gaming.