Unexpected shudowns when running games (but only in Windows)

Insane NiTEmare

Reputable
Apr 25, 2014
147
0
4,690
I've experienced shut downs in Windows for around 6 months, albeit for the last month I've used GNU/Linux and haven't had anything similar to this problem. Any time my computer is running Windows (tested Vista, 7, 8.1) and I'm running a game (full screen or windowed, borderless, OpenGL or DirectX) my system will shutdown. My CPU fans are at max all the time (I can't hear the fans in my case but I run them max anyway, CPU fans are stock from the Corsair H80i pump, the pump works) and ditto my GPU. My CPU doesn't heat up paste 45 degrees or so (monitored by coretemp) and my GPU with fan speed max never peaks 45 degrees either. I haven't overclocked the GPU and I'm currently running at only 3.4GHz on my FX-9370 but it doesn't seem to affect the system, and running stock clocks or 4.6GHz (no turbo) in Linux doesn't show any abnormal temperatures either. I'm currently using a stock Vista install with mostly basic components installed but it still shuts down.

Omega 15.03 Catalyst drivers (patched for Vista, last native Vista driver would not install)
Vista Ultimate x64 SP2 (also tested in Windows 7 Ultimate x64 SP1, Windows 8.1 Enterprise x64, same results)

FX-9370 (3.4GHz)
Crosshair V Formula-Z
R9 270X (ASUS 2GB Direct CU II TOP)
8GB (2x4GB) Kingston Hyperex RAM, 4GB (2x2GB) AMD Radeon RAM
Coolmax 1000w power supply (less than 6 months old, aware they aren't OEM but it's worked just as well as my Antec did)
Kingston Hyperex SSD (SH103S3120G)
3 hard drives with partitions accessible to windows (NTFS), 1 WD Green 1TB SATA 2, 1 WD Black 2TB SATA 3, 1 SG Momentus 500GB SATA 3. There are various BTRFS partitions on those and another drive dedicated to BTRFS and a swap partition, which is a WD Caviar Blue 320GB SATA 2.
H80i with stock fans running max speeds (~2200 RMP), the 270x runs at ~3000 RPM maxed as well.
 
Solution
The problem is, you're using a Coolmax PSU-shaped object with a "1000W" label slapped on it; big difference. Why that sort of consumer fraud hasn't been stamped out is baffling; my guess is there's not enough money in it for the parasites to act.
Are these shutdowns like someone pulled the power cord, or crashes to Desktop from within games or blue screens and/or reboots? If the first, your PSU is probably not able to hold up its end of the log, overheats, and shuts off. Coolmax is not a good brand.
FWIW, you'd extend your fans' lifespans by not running them at max RPM; try 90%. It doesn't appear that you have thermal problems. Are your VRM temps ok too?
 
The system appears as if the plug was pulled, but all the USB ports are still powered. It won't turn on until the switch on the power supply is turned off until the power is drained then set back to on. This only happens in Windows. I've never experienced this in Linux and I do far more demanding things outside of my Windows installs.
 
Gaming is what will increase the demand by your graphics card.
The USB ports may remain powered because they'd be getting power from the +5VSB rail; that's actually provided by a separate PSU inside the same box. The "main" part of it may shut down, but leave +5VSB available.
If you have a Kill-a-Watt meter, see what is being pulled from the wall at the time it dies.
A good 550W PSU should be enough for that system. Seasonic, XFX, EVGA GS, B2, or G2, or Antec would be good choices.
 
I've used a 600w PSU and it doesn't cut it. I'm not sure how much power I'm drawing and I don't have a Kill-a-watt meter to measure my power draw. If it was the power supply that doesn't explain why only windows fails in this regard and Linux can run the same program/game in OpenGL (or in Source games DirectX with Valve's OpenGL wrapper) with no interruptions.
 
The problem is, you're using a Coolmax PSU-shaped object with a "1000W" label slapped on it; big difference. Why that sort of consumer fraud hasn't been stamped out is baffling; my guess is there's not enough money in it for the parasites to act.
 
Solution
If I have to replace my power supply to use windows I'll just skip out on Windows; it's atleast $150 for a (good) power supply over 600w here right now and I've already used a 600w and it can't power the system (and it was one of the recommended OEM good ones, and no the psu wasn't dead).

(edit) If it was my power supply then my system wouldn't survive a night (ie at least 8 hours) of AIDA64 running over my RAM, CPU and GPU but it does. I don't think it's my power supply because if maxed load for at least 8 hours doesn't kill my system how can a mere 40 minutes do it. I had already run AIDA tests earlier this June to test my system but it just doesn't point to the pressure on my PSU.
 
For $150 there are Corsair and ECGA 700w to 1000w, for less than $120 there are units with A. less reviews and B. low ratings, and they don't support as many outputs as I use on my current PSU, albeit said models there are only 700w. I never turn of this computer and I need it on almost 24/7 so I need a PSU that has a lot of wattage to sustain capacitor wear (I am aware it would require even wear on all rails to be exact but that's not my point)
 
[strike] I've narrowed it down and It's highly unlikely it was the PSU. It's either a memory socket or a memory stick in my system that's malfunctioning (which would explain why I only get errors in windows, as I'm pretty sure my Kernel in Linux supported memory mapping in case of a bad address. [/strike]

(sorry for the glitch triple post) Nvm bad ram may be a part but it's not the main problem. Still not sure why this works in linux but I'm going to be rechecking memtest, AIDA64 and I've got a donor rig to test the graphics card in games and a different power supply to go with it.

I know someone who has a dormant 750w (label anyway, it's used) power supply but I don't know the make. I can't afford to buy a new power supply ATM myself though, not enough income to be feasible.