GTX 260 freezing in Crysis (blank colored screen)

[njdevilsfan87]

This only happens when playing Crysis. I can play Crysis for hours, but whenever I get killed by a grenade, or if a grenade just in general lands to close to me, the entire computer locks up and I get a blank single color screen. The temperatures aren't too high (usually mid 70s), but the fact that it only happens when a grenade lands near me makes no sense at all.

I played Oblivion, and that locked up once, but that happened when my core temperature went well into the 80s (this seemed like a normal lockup caused by overheating as my screen was multicolored boxes as opposed to a single color).

Everything else has been fine. I'm running 702/1512/1215 clocks on a 65nm vanilla GTX 260. I'm also now undervolted to 1.05V which has helped temperatures somewhat (which did not cause this instability, as this was already happening at 1.125V). I used to load in the upper 70s low 80s where now it is around 70C.

 

[MBrown]

I think the OC might be too high. I had a similar problem with my card and I lowered the overclock had haven't had any problems.

 

[Quiksilver]

Simple really. Your OC is unstable, either clock it back down or up the voltage.

 

[error8]

So, wow, you undervolted and overclocked your card in the same time and now you've made a thread here asking about what's wrong with your card. That is simply amazing.

 

[josh6079]

As others have said, your overclock is unstable.

 

[chizow]

Originally posted by: njdevilsfan87
Everything else has been fine. I'm running 702/1512/1215 clocks on a 65nm vanilla GTX 260. I'm also now undervolted to 1.05V which has helped temperatures somewhat. I used to load in the upper 70s where now it is around 70C.

Its built-in throttling, most likely triggered by shader heat or general instability from overclocked frequencies. You can easily monitor this by using GPU-Z and going to the Sensors tab. When the screen goes blank, the card cuts VDDC current from 40+A to 10-20A.

Undervolting certainly won't help this at all as it will just destabilize your core clock. If you want to stabilize your OC, you'll most likely just need to lower your shader clocks to 1404 or 1458. Either way you'll still be able to keep your core at 702 if its actually stable.

 

[njdevilsfan87]

Originally posted by: error8
So, wow, you undervolted and overclocked your card in the same time and now you've made a thread here asking about what's wrong with your card. That is simply amazing.

Please, just gtfo and stay out. No shit I have considered the possibility of an unstable overclock on my mind. Glad to have amazed you though.

I have yet to have such an isolated event, EVER. When I have overclocked CPUs and GPUs, they were either generally unstable crashing all over the place, or very close to completely stable. Never have I had such an isolated and prominent cause of a crash due to an unstable overclock with a GPU, CPU, memory, etc.

I wouldn't be making this thread otherwise.

Originally posted by: chizow
Its built-in throttling, most likely triggered by shader heat or general instability from overclocked frequencies. You can easily monitor this by using GPU-Z and going to the Sensors tab. When the screen goes blank, the card cuts VDDC current from 40+A to 10-20A.

Wish it was something I could monitor real time. Once the blank colored screen lock occurs, that's it. Not even Vista can reset the driver to get going again. Though I will text enable logging and see if anything shows up in the logs after a lock and restart. I undervolted because ATITool showed stable for 20 minutes even at 1.025V at these frequencies. I raised the voltage just a tiny bit for the cushion. Maybe it is time to start using Furmark to test GPU stability, because not even the undetected occasional yellow pixels show up while running ATITool.

Interesting that it does this though... I can force blatant instability by pushing the shader up some more, so that I can see errors being detected every second in ATITool, yet I can still load Crysis and play through it with visual artifacts (I see white flickering of textures - not all, like I may see a tree or some brush flicker) but yet it still will not lock unless a grenade lands too close to me... Though I have not played over an extended enough amount of time to see if anything else would cause a lock, or if a lock would just randomly occur.

 

[chizow]

Originally posted by: njdevilsfan87
Wish it was something I could monitor real time. Once the blank colored screen lock occurs, that's it. Not even Vista can reset the driver to get going again. Though I will text enable logging and see if anything shows up in the logs after a lock and restart. I undervolted because ATITool showed stable for 20 minutes even at 1.025V at these frequencies. I raised the voltage just a tiny bit for the cushion. Maybe it is time to start using Furmark to test GPU stability, because not even the undetected occasional yellow pixels show up while running ATITool.

Must be pretty unstable to produce a hard-lock, I'm able to monitor it continuously via a second monitor, and the game actually recovers by itself after a few seconds (depending on the shader overclock). The "black-out" period shows the VDDC current drop precipitously to 10-20A, then it goes back up to 40+A when rendering resumes.

I'd just set your core voltage to default, find stable shader clocks, then go from there on the core. Shader clocks/temps are going to limit you the most on GT200, adjusting Vcore won't really help you with that limit.

Interesting that it does this though... I can force blatant instability by pushing the shader up some more, so that I can see errors being detected every second in ATITool, yet I can still load Crysis and play through it with visual artifacts (I see white flickering of textures - not all, like I may see a tree or some brush flicker) but yet it still will not lock unless a grenade lands too close to me... Though I have not played over an extended enough amount of time to see if anything else would cause a lock, or if a lock would just randomly occur.

Its not really surprising, that's the point of stress testing on specific benchmark/stability software even though its unlikely any game will stress your hardware that much. You're testing for worst-case scenario so that your actual gaming experience isn't interrupted later from an unstable OC.

If you're unstable in ATT or Furmark and only see instability in actual games in intense scenes, that should be expected. A grenade or huge explosions obviously stress the GPU more, as that's a ton of shader/particle effects in a shader intensive title like Crysis, so its not surprising that would trigger artifacting/instability.

 

[error8]

Originally posted by: njdevilsfan87

Originally posted by: error8
So, wow, you undervolted and overclocked your card in the same time and now you've made a thread here asking about what's wrong with your card. That is simply amazing.

Please, just gtfo and stay out. No shit I have considered the possibility of an unstable overclock on my mind. Glad to have amazed you though.

It's amazing because you're such a smart boy that undervolted his card to reduce its temperature, so that it can be clocked higher. If you do this, you shouldn't ever mess with your hardware, because it's clear that you don't know anything about it. It's like you've hit your videocard with a hammer and now you're asking everyone, why isn't the card working properly.

Anyway, the voltage should be raised for a stable overclock, not dropped. And scanning with ATI tool for just 20 minutes, is nowhere near stable. Put the voltage where it was, mess with the fan speed, so that you won't go over 70-75C and start ocing the long and hard way ( first the core, then the shaders and last the memory). When you think you've reached a stable overclock, make an 1 hour ATI tool artifact test. If it passes that then you can further use Furmark and see if the doughnut crashes your PC. After one hour of that it should be quite stable. Play some games and see what happens.

 

[Hauk]

error is right OP. May have rubbed you wrong, but your OP was a head scratcher.

1.05v is for low power/low clock, like 2D only. That OC will probably be fine if you reset voltages. Uninstall the voltage software and keep Precision installed. I think a public release of voltage tuning software was not a good idea. It should be left for those brave/stupid/experienced enough to mod the bios..

 

[badnewcastle]

In my experience with OC'ing that card, I found that the shader clock is the most delicate, followed by memory and then core. Have you tried OC'ing each clock individually to find the limits independent of the others?

 

[njdevilsfan87]

I would have been fine with the response if he didn't come off a dick about it. (I also had my busiest school day of the week so I was just tired, and cranky - I am sorry about that)

But since I have no idea how to overclock, here is what I did, so everyone knows not what to do:

*Overclocking*
0) Initial overclocks set.
1) Card ran very hot at default 1.125V, passing 80C on a regular basis during games. Crysis was experiencing this freeze, at default voltage.
2) I considered the following: unstable overclock due to high clocks, overheating, or a combination of both. Others: game re-install, driver re-install, PSU
3) Game re-install was ruled out, and as for drivers, I have the latest and do not wish to go to an earlier version, even if they are the cause.
4) PSU was ruled out. I used a multi-meter while loading the GPU/CPU and the 12V rail was stable, at 12.01V under load.
5) To reduce heat, I slowly underclocked until signs of instability became prominent in ATT.
6) This actually did not happen until 1.000V. At 1.025V ATT would detect an error every few minutes (it would also miss some). At 1.05V no yellow pixels were seen in the first and last 2-3 minutes in a 22-23 minutes run. Card temperature never broke 70C, a very nice improvement there for me.

*Crysis*
1) Only happens when a grenade lands to close.
2) Huge explosions say 20 feet away of more, did not cause the screen lock.
3) Game has never randomly locked up, even after hours, only when a grenade lands very nearby.
4) NEW: Only happens in DX10 mode. In DX9 mode, this does not occur.

*Furmark*
1) Finally had something show more obvious signs of instability.
2) When the card passed 70C, which took it only 2.5 minutes to do, I began to see flickering artifacts here and there on the screen.
3) I let it keep running because I actually wanted it to lock up just like Crysis did to confirm, but after 15 minutes I just stopped it. Card reached 77C.

Now here is the main reason I am asking: I did a search on Google and I have found posts on different forums where people had a similar issue (blankcolored screen - not black screen, or loss of signal screen). But their events did not seem isolated, like mine is. I have NEVER, EVER, experienced instability so isolated like this before, even overclocking graphics cards way past beyond the point of stability. GPU instability I experienced in the past was completely random, NOT PREDICTABLE. Normally I would have gone with instability, but NOTHING else is showing any signs of it.

I also just realized I made a huge mistake on my first post...

This only happens when overclocking the card. I can play Crysis for hours, but whenever I get killed by a grenade, or if a grenade just in general lands to close to me, the entire computer locks up and I get a blank single color screen. The temperatures aren't too high (usually mid 70s), but the fact that it only happens when a grenade lands near me makes no sense at all.

I meant to put, this only happens when playing Crysis. If I read someone else type that, I would also kind of think to myself... talk about not being able to type quickly enough to keep up with your mind. That was a total mess up on my part which I believe lead this topic slightly down the wrong path.

I might as well just play at stock clocks for a few days and see what happens. If no grenade lockups, then I will be convinced it is the overclock causing the instability. If MBrown had a similar problem, and chizow is right on, then that would explain it... but it still just feels like an insufficient explanation... not trying to offend, its just how I am when it comes to wanting to know exactly what is going on because for now I see - unstable overclock -> nearby grenade explosions -> tons of particle and shaders effects in a short period of time - > VDDC current cut -> blank colored screen and system lock... while everything else in Crysis, and other games run fine and dandy... heh

 

[nOOky]

Honestly, trying for a max overclock on a video card usually ends up with no noticeable difference in games. It's like everything else computer-related, you can get to 80% of the performance for 20% of an effort, but to achieve that last 20% performance requires an 80% effort. Some cards overclock great, some don't. Most users on here come in wondering why their overclocks aren't like everyone else's. Probably because most everybody else exaggerates their claims.
In my experience, overclocking a gpu can maybe get you that next higher resolution or a bump in some texture/quality setting, but rarely does it make a game playeable when it wasn't before. I guess my point is that trying to push these things that last 20% usually isn't worth it, so just back the overclock off to a stable setting and be happy with your uberfast card the way it is