Hardware GTX 580 Black Screens - Instability?

[gifi4]

Hey all, with my current rig, I've been gaming on it but lots of games tend to crash, I first noticed it in Dirt 3, then Batman A.C, F.E.A.R. 3, Need For Speed: The Run, and more.
[I'm using Fear 3 as an example]
What will happen is that the game will just go to a black screen with the sound still running and eventually give me the whole "needs to be closed" thing from Windows 7, or I'll end the process myself using Task Manager.

Take a look at this thread as is seems to be a similar issue: http://www.tomshardware.com/forum/327231-33-game-crashing

Specs:
CPU: Intel - Core i7 2600k @ 3.4GHz
RAM: 8GB Dual Channel 1300MHz
GPU: Nvidia GeForce GTX 580 1536MB (Gigabyte)
MOBO: Intel DP67BA - LGA1155
HDD's: 2 x 1TB Western Digital Caviar Black
SSD: Corsair Force GT 120GB (SATA3)
Optical Drive: LG - GH22NS70
OS: Windows 7 Enterprise 64bit
Case: CoolerMaster HAF X RC-942
Monitor: Benq G2450H 24"
PSU: Cooler Master Silent Pro M700 (700 Watts)

Extra Details:
Everything inside the case is all connected correctly (power to GPU, GPU to PCI port etc.)

As recon-uk stated in the other thread, it could be the voltage of the GPU, so I already tried increasing it by .25mv but it did not make it any better.

I'm up to following his BIOS settings but, will it fix it? Is it safe to do considering that was for an 850W PSU and mine is just 700W?
All drivers are updated and BIOS is latest version (Same issue on old drivers and BIOS)

 

[AlanJohn]

Are you using retail games, or pirated ones?
Anyways have you tried running in Administrator mode? I really doubt it can be anything with the hardware.

 

[gifi4]

Are you using retail games, or pirated ones?
Anyways have you tried running in Administrator mode? I really doubt it can be anything with the hardware.

Did you check out the thread from Toms Hardware?
It basically covers my problem and with one of the same games... Instability is what I'm assuming.
Recon left pics of how he changed his BIOS and it made the crashes stop for the user who was suffering from these problems.
I'm wondering if it's safe to do those changes considering I'm on a 700 W not an 850 W...
Anyway, Admin mode has nothing to do with it, the games run, they just crash to a black screen. Retail or pirated doesn't matter either, especially if the game runs fine for someone else (Using a crack if pirated)...

 

[moose3]

What they are saying in that topic about the BIOS settings is for the system memory, not the power supply, and they were setting the memory values to some sort of preset 'extreme' profile settings.

So yes, you can try something like this and see if it works. Just don't set your memory settings to exactly what he has (unless it is some default profile) as you list your memory as being 1300Mhz so it has different CAS timing settings than his at 1866Mhz. You should be able to look up what your memory is rated for online.

If the memory tweaks don't help and you are still wanting to tinker with the video card, you can also 'under clock' the card by reducing graphics card gpu and memory settings to see if it runs stable (meaning you have a defective card as it should run at stock speeds).

 

[Originality]

All I can say is that the "This program needs to be closed" thing from Windows is not a symptom of hardware fault. Neither is the screen going black (with sound running) when launching games. That all sounds much more like software faults (e.g. drivers, system files, DirectX, .net, etc).

That said, I've seen one case where it was happening because the display adaptor (read, graphics card) had crashed whenever it approached peak load. I got him to RMA the card and the replacement fixed the problem.

 

[gifi4]

All I can say is that the "This program needs to be closed" thing from Windows is not a symptom of hardware fault. Neither is the screen going black (with sound running) when launching games. That all sounds much more like software faults (e.g. drivers, system files, DirectX, .net, etc).

That said, I've seen one case where it was happening because the display adaptor (read, graphics card) had crashed whenever it approached peak load. I got him to RMA the card and the replacement fixed the problem.

The games run fine, but eventually it will crash, I played Batman A.C for roughly 3 hours without a crash, then I played it later and it crashed within 10 mins.
As for software, the games tend to crash the same amount on both Directx9 and 11, system files are good considering it's been happening since I built the pc, all I was using was Dirt 3 then but that would still have the same problem.
So what?
I have to RMA the card? atm that's not exactly an option, MOBO does not have built in graphics so my computer will be useless...

 

[Originality]

Seek out alternative solutions (such as reinstalling Windows or clean reinstalling your drivers), but remember that it is an option that may help if the graphics card really is crashing when reaching full load. You can also try a benchmark to see if it triggers it.

 

[gifi4]

Seek out alternative solutions (such as reinstalling Windows or clean reinstalling your drivers), but remember that it is an option that may help if the graphics card really is crashing when reaching full load. You can also try a benchmark to see if it triggers it.

So basically, if it is crashing under stress then a benchmark or stress test will be a good way to test it? If so, what program should I use, I've had a look around but the few that I looked at were Directx10 or something like only CPU's with one core (I don't know why but yeah) so what would you (or another member) recommend?

I might be able to get a replacement sooner (if need be) if the store I bought the part from allows it, hopefully I won't have to RMA it and just return it to the shop and get another one because the part is covered with a year warranty from the store and I think a 2 or 3 year manufacturers warranty...

 

[Originality]

I used 3DMark and 3D Studio Max on my brother's comp years ago. Can't remember the modern ones.

 

[gifi4]

I used 3DMark and 3D Studio Max on my brother's comp years ago. Can't remember the modern ones.

I'm just getting this error trying with 3DMark:

Workload Single init returned error message: DXGI call IDXGISwapChain::SetFullscreenState failed:

The requested functionality is not supported by the device or the driver.

DXGI_ERROR_NOT_CURRENTLY_AVAILABLE

edit: Got it working but had to close Firefox... Might get it cracked so I can try it out using 1920x1080.

 

[gifi4]

I used 3DMark and 3D Studio Max on my brother's comp years ago. Can't remember the modern ones.

So I just ran it, and it crashed with sound still running, so is it time to RMA the card?
edit: Sorry for double post, didn't realize

Edit2: Just did a second clean installation of all Nvidia Drivers, I'll test 3DMark again.
edit3: Just crashed again...

 

[Originality]

There are a few other things it could be, but they're all beyond my knowledge. I would RMA at this point.

 

[gifi4]

There are a few other things it could be, but they're all beyond my knowledge. I would RMA at this point.

I don't at all recall mentioning this but, the card is in a 16x lane? rather than a 32x (which would be pci-e 2.0 lane, if I'm not mistaken)
Would that have an effect such as this?
The store that I got the card from is open tomorrow so I'll probably give 'em a call and find out if I can get a replacement there otherwise I'll send it to Gigabyte.

Edit: I just ran a GPU monitor and made it log every 2 seconds:


09/04/2012 01:19:22 GeForce GTX 580: GPU0 temp:55C, FAN:2550rpm, GPU0 load:66%, Vmem Usage:34%, GPU0:0 MHz, Shader:0 MHz, Memory:2005 MHz
09/04/2012 01:19:24 GeForce GTX 580: GPU0 temp:53C, FAN:2580rpm, GPU0 load:99%, Vmem Usage:0%, GPU0:0 MHz, Shader:0 MHz, Memory:2005 MHz
09/04/2012 01:19:26 GeForce GTX 580: GPU0 temp:53C, FAN:2580rpm, GPU0 load:99%, Vmem Usage:0%, GPU0:0 MHz, Shader:0 MHz, Memory:2005 MHz

As you can see, it the time period of 4 seconds, it went from GPU load 66% and Vmem usage of 34% to 99% load and it stopped using memory.
Could this just be due to having graphics settings to high?

 

[Originality]

16 lanes is normal, 8 lanes is half bandwidth (south bridge limitation for secondary graphics cards) but full length, 4x and 1x are short. I don't know if PCIe 3.0 changes things, but PCIe 2.1 and older all are indistinguishable (afaik). Never heard of 32 lanes except in motherboard south bridge spec sheets.

At stock speeds, no matter how high the demands on the graphics card, the only consequence should be frames. It should never shut off unless something was going wrong. This usually means heat, but if it can't even last long under full load, it means somewhere in the digital backbone of the daughter board, there is a hardware fault and that's not something I can diagnose without having it in my hands (and even then it's down to uncanny intuition rather than electrical knowledge). Sure, it might be heat, it might be that the heatsink or thermal paste isn't working properly, but rather than try and fix that yourself (voiding the warranty in the process) its best to contact technical support with a view to RMA.

 

[yellowsnowcone]

Those temps look fine, and the driver should back down and warn you in the GPU overheats.

Have you run memtest86+, or HCI memtest? I would try to verify the system RAM, first. If that passes, then assume there is a video card/driver issue. If you mobo has adjustable memory timings, try raising the three main ones by 1, as well. 9-9-9 to 10-10-10, FI. Chances are good it's either system RAM, video card RAM, or video card drivers.

If system RAM can be fully ruled out, then RMA the card. But, why waste the RMA time and effort if it's not the GPU?

 

[gifi4]

Those temps look fine, and the driver should back down and warn you in the GPU overheats.

Have you run memtest86+, or HCI memtest? I would try to verify the system RAM, first. If that passes, then assume there is a video card/driver issue. If you mobo has adjustable memory timings, try raising the three main ones by 1, as well. 9-9-9 to 10-10-10, FI. Chances are good it's either system RAM, video card RAM, or video card drivers.

If system RAM can be fully ruled out, then RMA the card. But, why waste the RMA time and effort if it's not the GPU?

I can't make a bootable CD at this current time as DVD's aren't allowing it to become bootable (memtest) and I don't have any CD's. Gotta check if I have a spare USB anywhere...

 

[gifi4]

Those temps look fine, and the driver should back down and warn you in the GPU overheats.

Have you run memtest86+, or HCI memtest? I would try to verify the system RAM, first. If that passes, then assume there is a video card/driver issue. If you mobo has adjustable memory timings, try raising the three main ones by 1, as well. 9-9-9 to 10-10-10, FI. Chances are good it's either system RAM, video card RAM, or video card drivers.

If system RAM can be fully ruled out, then RMA the card. But, why waste the RMA time and effort if it's not the GPU?

I can't make a bootable CD at this current time as DVD's aren't allowing it to become bootable (memtest) and I don't have any CD's. Gotta check if I have a spare USB anywhere...

Quoting myself...

Anyway, I found a memory card, put memtest on and ran it for 5 passes and there was not a single error. So, safe to say it's not the RAM?

 

[yellowsnowcone]

Anyway, I found a memory card, put memtest on and ran it for 5 passes and there was not a single error. So, safe to say it's not the RAM?

Absolutely, no, but the chances are awfully slim, if it makes it through several passes, compared to the video card itself, or a driver issue, which is basically unverifiable without a replacement card.

 

[gifi4]

Sure, it might be heat, it might be that the heatsink or thermal paste isn't working properly, but rather than try and fix that yourself (voiding the warranty in the process) its best to contact technical support with a view to RMA.

So, earlier today I unplugged the card completely and took it down to store where I got it, asked me what was wrong, told them what was happening etc. and they said they'd have to take it and it would be about a week till I got the replacement. He offered me a cheap, old 16MB gpu because he knows my mobo doesn't have onboard graphics (free, just return when I get the new card), took it and was happy. Got home, installed it, all was good until, my monitor decides the HDMI port no longer wants to work so now I have to take that back tomorrow, and the gpu they let me use only works on HDMI, the D-sub port does not work. Luckily I have my old GPU that is 128MB GeForce 6600 GT and am temporarily using the D-SUB on that.

Damn, what are the chances... =D

So, to-do list:
Take monitor back to a different shop to get a replacement (Got a 3 year warranty on that and it includes wear-and-tear of the ports)
Wait for phone call telling me to pick up a new GTX580
Return their GPU.

When I return their GPU, do you think I should tell them the D-Sub port doesn't work? It was like that when I hooked it up...

Lastly: I'm sure the 580 was faulty because they told me (and my father) to wait a minute while they test it out to see if anything was wrong, then they came back and told me it will take a week before I get a new one...

 

[Originality]

Wow, you have worse luck than my brother. And yeah, based on what you've said, it is broke.