Scoob
New member
Hi all,
Some of you may recall that a while back I got myself a second GTX 570 (and Inno3d) to pair with my original Palit GTX 570 for SLI. I documented my experiences here and was very very happy.
Some time later, as a bit of a project, I decided to water cool everything. This also went really well and I was happy again. However, this didn't last long as I had a compression fitting fail, killing my original Palit GTX 570 - I replaced this with an EVGA 570 and was happy once more.
Now, quite some time later, I suspect my Inno3D GTX 570 is less than happy. It was never a very good overclocker, but managed 800mhz happily for an everyday OC @ 1025mv and could do 850mhz @ 1063mv, but I did not like going over 1050mv personally. My EVGA cards stock is 797mhz @ 1025, so both at 800 was a nice speed. Ironically my now dead Palit did these speeds with ease at far less volts *sigh*
Anyway, my Inno3d GTX 570 was the weakest of all my cards when it came to overclocking, not that I needed to overclock to play anything well of course, I did it just for fun. I started to notice some slight instabilties at my previously stable clocks of 800mhz - testing the Inno3d in isolation suggested this was the "bad" card. Not a huge deal, so I ran at the stock of 732mhz for a while without issue.
Now this card seems to be deteriorating. Generally it's 100% fine at stock, it'll bench away in things like Heaven 3.0 and 4.0 (with stressful settings) Vantage & 3DMark '11 as well as play many games just fine. However, Crysis & especially Crysis 2 often suffer from hard crashes during play.
Now, I know Crysis 2 in particular is renowned for pushing GPU's quite hard - I've read many many posts about people needing to give there GPU's a slight voltage bump even at stock to ensure stability. I did this and it appeared to help for a while, but the card is getting worse I think.
The issue is that the card usually performs flawlessly if loaded hard, so at at 99% in a benchmarking tool for a while. Even gaming in Crysis or Crysis 2 I can sit in a scene where the card is pushing 95% + load and it's perfectly fine. The hard crashes seem to happen when the card is only paritially loaded. Bumping the vCore seemed to help a little, i.e.
975mv @ 732mhz - regular crash at a certain point
988mv @ 732mhz - less regular crashing
1000mv @ 732mhz - even more stable, get a CTD rather than hard crash + reboot.
1025mv @ 732mhz - hard crash + reboot again.
The thing that confuses me is that the crashing is not about GPU load at all, it generally happens at fairly moddest load, rather than when the GPU is pushed. It's also not temerature related as the card rarely goes above 40c - in fact the Inno3D is cooler than the EVGA when run in SLI.
Additionally, when in SLI things are a little less stable than just running the "bad" Inno3D GTX 570 in isolation - consider that on it's own a single card is regularly at 99% load during benching & some hard scenes in game. When paired with the EVGA GTX 570 it has a far easier life.
So, while I'm sure from my testing (far far more than I mentioned here) that my older Inno3D GTX 570 is somewhat unstable, I'm not quite sure of the cause.
I have observed that during a recent NV driver update, my GPU's switch power states much more readily. They'll also happily go down to 50mhz core when idle with TWO screens running, they never used to do that on earlier drivers. I understand that NV have made a few changes to the power management of 500 and possibly 600 series cards recently, so I wonder if my Inno3D GPU is somehow throwing a wobbly as the power state changes at lower GPU loads.
I mean, if I get a slight blip in FPS as the game loads an asset / I save a game and the power (vCore on GPU) drops off too quickly then that could explain the crashes. Such hard crashes DO often happen in such situations.
I know I've rambled, I just wanted to get some input from you guys as to what might be causing this.
TL DR - GPU runs fine when pushed HARD in benching etc. even with a moddest OC, however it's NOT stable at mid-level loads - even at stock settings or with a slight vCore bump. As such it regularly passess benchmarks just fine, at at 90%+ load, yet will crash (hard + reboot) in games (admitted tough ones) where the GPU load is less, say in the 50-70% range.
Power is fine, I just got a new 860i PSU thinking the issues were power related, but it made not difference. I DO OC my CPU, but I reverted everything to stock for testing. I also was using W8 for a while, but dropped back to W7 to test. I've tried different drivers and the older ones do seem to be slightly better, but the problem still occurs.
After having to already replace one GTX 570 that worked brilliantly, and get another block, I hesitate to buy yet another. I would in fact likely just pick up a pair of 670/680's and two new blocks as this issue is bugging me.
Oh, I went through a phase of thinking that my motherboard VRM's were getting a bit hot, and causing the instabilities. Reason being I've game for a couple of hours ok. Then I'd get crash, reboot, reload game & crash again very quickly. Leaving the PC for a while usually saw things more stable. However, after going back to stock CPU and the Motherboard VRM heat sinks feeling much cooler, I still get the problem. As with many such problems it's slightly intermittant nature makes it harder to track down, though I have observed a pattern of sorts.
Any advice / suggestions of things to try welcome.
Cheers,
Scoob.
Some of you may recall that a while back I got myself a second GTX 570 (and Inno3d) to pair with my original Palit GTX 570 for SLI. I documented my experiences here and was very very happy.
Some time later, as a bit of a project, I decided to water cool everything. This also went really well and I was happy again. However, this didn't last long as I had a compression fitting fail, killing my original Palit GTX 570 - I replaced this with an EVGA 570 and was happy once more.
Now, quite some time later, I suspect my Inno3D GTX 570 is less than happy. It was never a very good overclocker, but managed 800mhz happily for an everyday OC @ 1025mv and could do 850mhz @ 1063mv, but I did not like going over 1050mv personally. My EVGA cards stock is 797mhz @ 1025, so both at 800 was a nice speed. Ironically my now dead Palit did these speeds with ease at far less volts *sigh*
Anyway, my Inno3d GTX 570 was the weakest of all my cards when it came to overclocking, not that I needed to overclock to play anything well of course, I did it just for fun. I started to notice some slight instabilties at my previously stable clocks of 800mhz - testing the Inno3d in isolation suggested this was the "bad" card. Not a huge deal, so I ran at the stock of 732mhz for a while without issue.
Now this card seems to be deteriorating. Generally it's 100% fine at stock, it'll bench away in things like Heaven 3.0 and 4.0 (with stressful settings) Vantage & 3DMark '11 as well as play many games just fine. However, Crysis & especially Crysis 2 often suffer from hard crashes during play.
Now, I know Crysis 2 in particular is renowned for pushing GPU's quite hard - I've read many many posts about people needing to give there GPU's a slight voltage bump even at stock to ensure stability. I did this and it appeared to help for a while, but the card is getting worse I think.
The issue is that the card usually performs flawlessly if loaded hard, so at at 99% in a benchmarking tool for a while. Even gaming in Crysis or Crysis 2 I can sit in a scene where the card is pushing 95% + load and it's perfectly fine. The hard crashes seem to happen when the card is only paritially loaded. Bumping the vCore seemed to help a little, i.e.
975mv @ 732mhz - regular crash at a certain point
988mv @ 732mhz - less regular crashing
1000mv @ 732mhz - even more stable, get a CTD rather than hard crash + reboot.
1025mv @ 732mhz - hard crash + reboot again.
The thing that confuses me is that the crashing is not about GPU load at all, it generally happens at fairly moddest load, rather than when the GPU is pushed. It's also not temerature related as the card rarely goes above 40c - in fact the Inno3D is cooler than the EVGA when run in SLI.
Additionally, when in SLI things are a little less stable than just running the "bad" Inno3D GTX 570 in isolation - consider that on it's own a single card is regularly at 99% load during benching & some hard scenes in game. When paired with the EVGA GTX 570 it has a far easier life.
So, while I'm sure from my testing (far far more than I mentioned here) that my older Inno3D GTX 570 is somewhat unstable, I'm not quite sure of the cause.
I have observed that during a recent NV driver update, my GPU's switch power states much more readily. They'll also happily go down to 50mhz core when idle with TWO screens running, they never used to do that on earlier drivers. I understand that NV have made a few changes to the power management of 500 and possibly 600 series cards recently, so I wonder if my Inno3D GPU is somehow throwing a wobbly as the power state changes at lower GPU loads.
I mean, if I get a slight blip in FPS as the game loads an asset / I save a game and the power (vCore on GPU) drops off too quickly then that could explain the crashes. Such hard crashes DO often happen in such situations.
I know I've rambled, I just wanted to get some input from you guys as to what might be causing this.
TL DR - GPU runs fine when pushed HARD in benching etc. even with a moddest OC, however it's NOT stable at mid-level loads - even at stock settings or with a slight vCore bump. As such it regularly passess benchmarks just fine, at at 90%+ load, yet will crash (hard + reboot) in games (admitted tough ones) where the GPU load is less, say in the 50-70% range.
Power is fine, I just got a new 860i PSU thinking the issues were power related, but it made not difference. I DO OC my CPU, but I reverted everything to stock for testing. I also was using W8 for a while, but dropped back to W7 to test. I've tried different drivers and the older ones do seem to be slightly better, but the problem still occurs.
After having to already replace one GTX 570 that worked brilliantly, and get another block, I hesitate to buy yet another. I would in fact likely just pick up a pair of 670/680's and two new blocks as this issue is bugging me.
Oh, I went through a phase of thinking that my motherboard VRM's were getting a bit hot, and causing the instabilities. Reason being I've game for a couple of hours ok. Then I'd get crash, reboot, reload game & crash again very quickly. Leaving the PC for a while usually saw things more stable. However, after going back to stock CPU and the Motherboard VRM heat sinks feeling much cooler, I still get the problem. As with many such problems it's slightly intermittant nature makes it harder to track down, though I have observed a pattern of sorts.
Any advice / suggestions of things to try welcome.
Cheers,
Scoob.
