Windows TDR resetting video card when multiboxing
OS: Windows 7 (applies to vista sp1 and later)
Game: Rift
Setup: 5 chars, 2 monitors
CPU: i7
GPU: HD Radeon 7750
Power: Thermaltake 600w
Graphics setup: 1 character set to low graphics settings as the main. 4 Slaves in low performance mode with all settings minimised.
Problem
Tried a "new" card in my system to upgrade from my pathetic Radeon HD 4--- series card. Was surprised when the same configuration my old card handled with no trouble caused the card to freeze and windows to reset the card (sometimes continuously). Even with no overclocking and even with all card settings at a minimum.
This would happen randomly when I loaded more than one character up with innerspace in Rift. On occasion it would run without issue for hours and often happen even before I had all the characters loaded up. Very strange. If you close all the windows and restart them after hours of trouble free playing....bang. And once it has happened there is no recovering from it. It would happen around 50% of the time.
The main symptom is that the screen would freeze and come back again. A windows popup in the taskbar would read: "Display driver stopped responding and has recovered". (i.e. Windows TDR)
Sometimes it would keep doing this and never recover and require a reboot.
In certain situations the computer would blue screen with error code: STOP: ...0x116.
(NB: Turns out this blue screen occurs is when TDR resets happens more than 5 times or lasts more than 60 seconds by default)
Random flailings of a very mad scientist
This was a very weird problem and I got all scientifical experiment on its ass to try and troubleshoot it.
At no time did the card or CPU get above 45 degrees or was there any other strangeness. Even in the Rift scenarios the only strangeness was the resetting.
Only when I ran 3 or more Rifts/NWs in innerspace was there a problem.
After much research, many reboots and many many hours later these are my findings:
I tried a lot of things that did not work to fix this including but not limited to:
- Upping "available power" to card (cannot change voltage on this one) without overclocking - caused blue screens instead of resets!?
- Underclocking card - only made it worse.
- Used the Unigine Heaven benchmark and assaulted the card - ran fine.
- Tried a more insane benchmark and ran 5 instances of Heaven at resolutions similar to my chars - ran fine.
- Bumped up settings to "insane mad scientist mode" on benchmark and ran 5 instances again - ran fine.
- Ran a single copy of rift inside innerspace and ran 4 copies of mad scientist benchmark in background - ran fine.
- Ran two copies of rift in innerspace with all settings maxed and FPS limiters off - ran fine.
- As above with some benchmarks in the background to the point where all 2 gig of the poor little cards memory was used up and cpu at 99% continuous- ran fine??!
- Adjusted TDR or "Timeout Detection Recovery" since this was what triggered the card resets. No dice, card is actually locking up.
- Fired up and configured Neverwinter for 5 box mode. Everything went fine (by chance?) until setup regions the same as for Rift...BOOM.
- Removed overlaps from both Rift and NW and restarted clients multiple times each...failed.
- Used the default configuration for Neverboxer without any window modifications...failed
- Loaded the Heaven benchmark in 5 ISboxer windows 5 times....all succeeded.
Not sure where to go now. The last ran fine but then the way the heaven benchmark works is probably different to a game.
I have a second hand (but different) video card in the mail that will hopefully turn up tomorrow. I will run all the slaves off it and see if that solves the problem.
Until then I will try playing a 3d game solo to see if there are any lock ups. I could try returning the card under warranty as "faulty" but if they test it they will see what I did - the card performs fine in normal scenarios.