Did Far Cry 3 kill my GTX 570?

[knutover]

Hi

First my specs:

GPU: PNY GTX 570 (http://www.amazon.co.uk/GMGTX57N2H12ZPB-GeForce-GTX-570-PCI-Express/dp/B004ODQSA0)

CPU: i5 2500K

RAM: 8 GB (4x Corsair XMS3 DDR3)

PSU: 500W MIST 500BHE12

Mobo: Asus P8P67 PRO (with working sata ports)

All stock clocks and stock coolers


The story:

I got Far Cry 3 through Steam a couple of days ago and have been having fun blasting pirates and tigers in the beautiful jungle for over 9 hours in game in the last week, all without a single glitch or bug of any kind. The GTX 570 sounded like it was being pushed, but it was after all not an entirely new card anymore, and the settings were pretty high, but not ultra.

Then today I had played for a few hours when I get to a cut scene. Everything is smooth until the screen turns black and I get "Far Cry 3 has stopped responding". Restart the game, load save, everything is fine. Then total system freeze after maybe 30 seconds. Reboot, restart game, system freeze in the main menu. Reboot, rage, reinstall uPlay, update drivers, run CClean on the registry, switch to recommended settings, vacuum the card, restart the game, load, and "Far Cry 3 has stopped responding" after 5 minutes.

Cry

Then I decide to check if my GPU can still run other programs. I start Skyrim, since I have played it for ~150 hours on this system with almost no problems. The game loads and things look great. I press the left mouse button to bring up the animated flames on my hands, and the screen blacks out. Then it comes back again for 2-3 seconds, and then the screen blacks out permanently. Reboot, and decide to give up gaming and watch Game of Thrones. 5 minutes in: "Media Player Classic has stopped responding". Try again. Same result. The episode plays for 5 - 10 minutes, then the image crashes while the sound plays on until the error message pops up.

So, now I put the question to you: Have my 9 hours of Far Cry 3 somehow broken the back of my one and a half years old 570? Is it even possible for it to be broken this way, where the fan spins and desktop and Opera runs fine, and even the games and media run fine for a few minutes before they crash? And most importantly, is there a way to fix it?

Any help is very much appreciated!

I am going to go read a book. Hope it doesn't f***ing crash.

 

[lavaheadache]

What drivers are you running? Supposedly the latest drivers are causing headaches for some people.

 

[bystander36]

What are the temps of the GPU when running?

It also may be the PSU that died, and not the GPU. Without proper power, the same error can occur. I recently had my PSU die a month or 2 back while playing Far Cry 3 and had the same error you did. I thought it was my GPU at first, until I tested things further.

 

[crashtech]

Well, it's possible for a part to fail, but it was probably not any specific game or thing you did that caused it. After all, the card should throttle to protect itself from damage.

After doing some stuff like reseating the card, if you have a spare hard drive, you might load up a fresh install on it and see how the card acts that way. There might be something borked in your OS.

 

[knutover]

I checked the GPU temp now, just sitting in the main menu of FC3, and it rose from about 40 C idle to 78 in 3 minutes with the fan only on 48 % before i shut it down to prevent another crash. I think this might be relevant.

I was running the 314.07 drivers, and things crashed. I am now running the 320.18 drivers, and things still crash.

 

[bystander36]

That is not poor temps, so that shouldn't have been the problem. I have never heard of your PSU. I am more concerned with your PSU than anything.

 

[Keysplayr]

You can also try clearing the gaming cache in steamapps or check for integrity of the games cache files.

 

[knutover]

I have verified the integrity of the game cache, and there were no problems.

How can I find out if the PSU is causing my problems?

Ok, I see now that these cards can run over 90 degrees without immediate damage, so I'll try watching the temps in the FC3 menu again until it crashes and see how high I get.

 

[knutover]

It kept rising, and when it got over 75 GPUz started saying 0 sometimes, but it crashed at 79, so well below dangerous temps. The "GPU load" was 99 % the entire time though.

 

[3DVagabond]

Googled around a bit about your power supply. Results don't inspire confidence, unfortunately. First off, it's an old design. I'm not sure how long you've had it but the review I found was from 2006. No decently designed modern PSU would have the load distribution between the rails done like this one.

Assuming these specs are accurate, this PSU is at best a 400W unit. I would definitely recommend replacing it. Especially if it's old.

First thing to try is can you split the videocard's load between the 2x 12 volt rails. OCP for the 12v rails is set at 240w. That's a bit low for a 570. If you can divide the load between the 2 rails it might help. I would definitely replace that PSU though, even if you get it working.

Edit: Sorry for not being clearer on hook up. Typically the 24pin/8pin to the mother board is run off of the #1 12v rail. Hook the 2x6pin on the card to the #2 12v rail. If it's not clearly marked on the PSU/cables, it might be in the manual.

 

[lavaheadache]

lol, that psu is likely an issue.

 

[rgallant]

just thinking
you might want a custom fan profile using msi-ab 50% fan @ 50c ,60% @ 60c, 70%@70c or what ever -test at high fan speed to keep gpu cool as possible to see if the system still crashes.
-also check your stock cpu fan and cpu temps. while gaming.[core temps]
-imo Asus P8P67 PRO if using asus software for temps or anything some would say to uninstall it and use free 3rd party appts
-logging gpu-z might show the gpu voltage drop before a crash pointing to a bad power connector or psu.

 

[crashtech]

Oh, lordy. That PSU might not be the problem today, but it's definitely a time bomb.

 

[flexy]

Yeah, "Mist" In German means about the same as Sh!t or Cr@p, seems fitting

FC3 can be quite demanding and gets temps quite high, temps are a little on the higher side (compared to my 660 TI) but still "tolerable" and normal range, I also get into the mid 70s easily with FC3 and a ton of fans.

Run the usual stability tests like OCCT http://www.ocbase.com/ and test CPU, run on Medium and check CPU temps, also run the gfx test if you can and POSSIBLY furmark although I would be careful with graphics burn-in tests when there is something wrong with the GPU, watch temps!!

Also, check ram with memcheck.

If there is issues, try the usual reseating the GPU in the slot, and if you can and know what you are doing you could try to remove HSF of the card and apply new layer of thermal paste, although temps seem normal to me. On the other hand, the fact that problems occur AFTER some gaming leads me to believe it's a heat issue. But, yes that PSU has to go.

 

[Crap Daddy]

Definitely first the PSU must be changed. I'm amazed it worked until now.

 

[knutover]

Ok, thanks for all the help guys!

I agree that the PSU is looking suspicious. I talked with a friend, and I will try my card in his setup tomorrow, but I think I will be replacing the PSU in the near future anyway.

I also ran the OCCT power supply test. Graphs here (http://imgur.com/a/L9Qdu). Except for an error right at the start, where windows said that the grapics driver had encountered an error but recovered, and the screen saver kicking in at a little over 40 minutes, everything went alarmingly great. GPUz said that the GPU temp was at 91 - 92 degrees C during the entire test before the screen saver. Am I right in thinking that this removes a lot of suspicion from the GPU cooling?

Tried FC3 again. "Stopped working" after 3 minutes.

Will try the GPU : 3D test now.

 

[knutover]

That went fast. Graphs here (http://imgur.com/a/Xkwx5). The screen froze after 20 seconds, then it blacked out for a while and came back for 1-2 seconds at the 40 second mark before freezing again. Rinse and repeat every 20 seconds, with a "Graphics driver has stopped responding but recovered" message each time. Closed the test the seventh time I got the chance.

GPUz has the temp at 57 C for the first crash, and then never above 56 afterwards.

What does it mean?!

 

[Anteaus]

Do you have a factory OC version of your 570? That is one of a few games I know about that has trouble with some OC GPUs. I had that exactly same problem until I installed MSI Afterburner and clocked my card back to reference. The problems completely disappeared. The same thing happens with Just Cause 2. I think the probably is purely software related...whether its a game code or driver issue I can't tell you. Other than those games my card works fine with everything else I own.

Just something to look into.

 

[tweakboy]

I think your system blew itself up! Not a game issue now. Or are other games play good. gl

 

[flexy]

>>
GPUz said that the GPU temp was at 91 - 92 degrees C
>>

Sorry some contradicting stuff here because once you say GPU-Z reports fine temps and then you say it's in the 90s?

Can you run afterburner or evga precision and really watch the temps/charts while you game (put the temps on OSD display so you see them in real time) ...respective GPU-Z should actually be fine and just let it run in the backgrouns. Just dont trust those OCCT GPU readings at the moment.

Did you check your memory with memtest? Also please use OCCT CPU test (medium) and let run for, say 30mins to an hour...simply so we can exclude CPU/mem test and can focus that's something with the GPU and not something else.

Then of course, if you have a modded card BIOS (?) or overclocked, there is ALWAYS a chance the OC doesn't work..really no point starting to look for h/w issues this way, you must test with a stock/original bios...and do the entirely problem searching strategically, step by step testing and eliminate components like mem/CPU etc. first.

PSU test in OCCT I personally never used, by the way. Since I once blew a brand-new OCZ PSU with furmark I am careful with burn-in tests and I especially dont like the idea to put "abnormal" loads and load changes on my PSU as the PSU test likely does.

 

[knutover]

Anteaus: No, I have no overclocks on anything, factory or otherwise.

flexy: The 78 temp was after 3 minutes of FC3 main menu. The 91 - 92 was during the 40 minutes of OCCT testing. I believe that the card would have reached the nineties in FC3 as well, if I hadn't aborted it, and from the fan noise during the time it worked I think it was at about that temperature before. Is this a problematic temperature?
I used GPUz's "Log to file" function to get the temps. It ran minimized in the background while I did the tests/played the games, and then I looked at the txt file after the crash. OCCT doesn't give any GPU temp readings that I can see.

Also, I pulled out the card and tested it in my friends setup. It has an i7 970 I think, a 650 W PSU and 6 GB RAM. In this setup the card did 20 minutes with no problems of the OCCT GPU test that crashed after 20 seconds on my system. 1080p video played without a hitch. The game "Dead Island: Riptide" however, stopped working after 5 - 10 minutes. Then I tried "Warframe", and it froze the computer.

The temps during all of this were around 85 - 86 at full load according to the GPUz log. This might be because his CPU is water cooled, so no hot air got blown on the card.

I'm thinking this indicates that something is broken in the card that somehow only manifests completely when it runs games. Does this make sense?

Thanks again for all the help!

I'm going to run the CPU and memtests now, though I don't think they will reveal anything, since the graphics card crashed in my friends setup as well.

 

[flexy]

Yes looks like a card issue, but I would lie now making a statement about 570 temps since I know only about GK104 and my old GTX 275, while for the 660 TI mid-90s would be FAR too high, my GTX 275 did 97C easily without a problem, so I can't say what's ok now for the 570. If i were in your shoes I would take out the card and do a arctic silver job and re-seat the heat sink again...but it depends on your proficiency with those kinds of mods, otherwise I'd look into RMA. It could well be that in time the HSF got lose or the thermal paste dried out and similar resulting in problems. You never know.

 

[amenx]

OP, I know you noticed the problems on 314.07, but 320.18 has probably the most ongoing issues/complaints of any driver yet released. Whatever problem your card has, there is a chance that 320.18 could be exacerbating it. FWIW, try 314.22, as thats probably the last most stable driver out there atm.

And btw, any temps under 90 are OK, no need to worry until it pasts that.

 

[Piotrsama]

The temps during all of this were around 85 - 86 at full load according to the GPUz log. This might be because his CPU is water cooled, so no hot air got blown on the card.

I'm thinking this indicates that something is broken in the card that somehow only manifests completely when it runs games. Does this make sense?

Maybe it's just dirty so it heats up a lot, anyway to try to clean it?
EDIT: Sounds like you already tried.

 

[flexy]

Don't go just after GPU temps. All it would take is a HSF not properly having contact with a voltage regulator, mem chip or similar shens...all possible.