Occasional Hard Crash in certain games - video hardware error - any advice? GTX 960

[jamesalex]

I have a complicated problem with hard crashes and if anyone smart could give me some advice it would be great!


I occasionally get hard crashes while playing certain games including - Dragon age inquisition, Skyrim and the new Paladins game on steam. However it has never happened in the 30+ hours for Mass effect 3.

The screen just goes black and sometimes I get a funny noise through the speakers. My only option is to turn the computer off.

I have stress tested the GPU (GTX 960) with Furmark for roughly 40 minutes I think. (I do not know if this is sufficient?)

I have uninstalled and reinstalled my nvidia drivers.

When playing skyrim with an ENB installed it would crash all the time. Not sure whether it could be a graphical conflict? :S

I had experience black screen crashes with my previous graphics card (7770) which leads me to believe it is nothing to do with GPU. Although this happened less often and only in guild wars 2. Additionally the problem was solved by lowering the graphical settings. The RAM has also been replaced in this time.

I feel like I have pretty good airflow and temperatures, and have checked these whilst playing < 70degrees GPU.



When checking the reliability history in windows 7 it lists the error as a video hardware error. BCcode 117.



Specs:
i3 2120
GTX 960 ASUS strix 2GB
H61 mobo
430w corsair psu
corsair vengence RAM 8GB
120gb SSD sandisk
2tb WD green HDD



Can anyone solve my problem? 😀


Many thanks

 

[amtseung]

Try using DDU to wipe all display drivers and then reinstall the latest Nvidia drivers. I'm horribly unfamiliar with Nvidia graphics cards, as I've never owned one, but this is always a good place to start.

Modded Skyrim will crash. That's the nature of the beast.

40 minutes of Furmark is more than plenty. I feel sorry for that GTX960. You should pet it and tell it you're sorry and that you won't punish him anymore. I'm kidding, don't actually do that, but 40 minutes of Furmark is plenty tortorous. Even 5 minutes of Furmark is liable to actually just bricking your GPU. Try using Unigine Heaven or Valley in the future.

Actually, I'm thinking you might have a stability problem. When a screen just goes black, makes a funny noise through the audio, and you have to reboot to get any further response, I assume "bad overclock, gotta tone it down", which makes me think it's a stability problem. If it were a heat problem, you'd notice thermal throttling all across the board long before it shuts itself down. Heat affects stability, but so does power delivery.

The only other thing I can think of is that your PSU is insufficient. I'm assuming your PSU is the Corsair CX430 because it's the only 430W that Corsair makes. It says it has a 12V rail able to push 384 watts, and that's on a good day. Peak momentary power draw of the GTX960 can spike up to 300W, and quite often, according to the Tom's Hardware power draw charts on the GTX960, and if this occurs while gaming, then your CPU will be under load, somewhere in the neighborhood of 65-70W. Asking for that much power from a source only able to supply 384 on a good day could be the source of your inexplicable shutdowns.

Start with the complete wipe and reinstallation of your GPU drivers.

 

[jamesalex]

Thanks for your advice,

I have reinstalled the drivers with DDU but have since had a crash on The elder scrolls online.

So i looked at the power spikes for the ASUS strix 960 and could see that sometimes they get pretty high. Could that cause the game to crash in such a manner? Is there anyway i could differentiate a bad overclock from an insufficient power supply?

I guess my next step is maybe to slightly underclock the card?

Failing that I will have to pet the graphics card 😉

James

Try using DDU to wipe all display drivers and then reinstall the latest Nvidia drivers. I'm horribly unfamiliar with Nvidia graphics cards, as I've never owned one, but this is always a good place to start.

Modded Skyrim will crash. That's the nature of the beast.

40 minutes of Furmark is more than plenty. I feel sorry for that GTX960. You should pet it and tell it you're sorry and that you won't punish him anymore. I'm kidding, don't actually do that, but 40 minutes of Furmark is plenty tortorous. Even 5 minutes of Furmark is liable to actually just bricking your GPU. Try using Unigine Heaven or Valley in the future.

Actually, I'm thinking you might have a stability problem. When a screen just goes black, makes a funny noise through the audio, and you have to reboot to get any further response, I assume "bad overclock, gotta tone it down", which makes me think it's a stability problem. If it were a heat problem, you'd notice thermal throttling all across the board long before it shuts itself down. Heat affects stability, but so does power delivery.

The only other thing I can think of is that your PSU is insufficient. I'm assuming your PSU is the Corsair CX430 because it's the only 430W that Corsair makes. It says it has a 12V rail able to push 384 watts, and that's on a good day. Peak momentary power draw of the GTX960 can spike up to 300W, and quite often, according to the Tom's Hardware power draw charts on the GTX960, and if this occurs while gaming, then your CPU will be under load, somewhere in the neighborhood of 65-70W. Asking for that much power from a source only able to supply 384 on a good day could be the source of your inexplicable shutdowns.

Start with the complete wipe and reinstallation of your GPU drivers.

 

[amtseung]

An unstable overclock can accentuate the shortcomings of an insufficient power supply.
An insufficient power supply can accentuate the shortcomings of an unstable overclock.

Neither are a direct cause of the problem, but it is not sheer coincidence. When your PC requires more power than the 12V rail can supply, there's a certain margin for overcurrent, and then overcurrent/shortcircuit protection will kick in, and your PC will just black screen and/or shut off. When your overclock is unstable, your GPU ends up with so many random errors because it is choked and the silicon just can't switch on and off fast enough that it metaphorically throws its hands in the air and gives up, typically resulting in a black screen.

If you have overclocked, I would set the thing back to stock clocks and stock offsets, bring the power limit back to 100% (leave the temp target, that's fine), and hope you don't overwork the power supply. An i3 doesn't draw all that much power, so I'd imagine you should be okay if you don't overclock. You can try get a 550W or 650W PSU or something excessive and see if the problem goes away. If not, then it's probably not a hardware issue, and it's probably software related like drivers or something.

I hope you didn't brick your graphics card by running furmark without breaking it in. I've done that once, and thank god it was a $35 garbage card I was trying to kill to begin with.

 

[jamesalex]

Hmmmm.

I don't think I can get a hold of a higher wattage PSU unfortunately, without buying one.

I haven't OC'd the card any more than the stock OC and form what I understand, they do extensive testing to make sure they are stable. That said if I were to underclock the card (which I don't know how to do - I have MSI afterburner which i think will be needed) then I guess it would be using less power and therefore I could tell whether that is the issue.

Just a thought...I have had this PSU for 4 years now. Does the quality of the 12v rail degrade with time?

Thanks

An unstable overclock can accentuate the shortcomings of an insufficient power supply.
An insufficient power supply can accentuate the shortcomings of an unstable overclock.

Neither are a direct cause of the problem, but it is not sheer coincidence. When your PC requires more power than the 12V rail can supply, there's a certain margin for overcurrent, and then overcurrent/shortcircuit protection will kick in, and your PC will just black screen and/or shut off. When your overclock is unstable, your GPU ends up with so many random errors because it is choked and the silicon just can't switch on and off fast enough that it metaphorically throws its hands in the air and gives up, typically resulting in a black screen.

If you have overclocked, I would set the thing back to stock clocks and stock offsets, bring the power limit back to 100% (leave the temp target, that's fine), and hope you don't overwork the power supply. An i3 doesn't draw all that much power, so I'd imagine you should be okay if you don't overclock. You can try get a 550W or 650W PSU or something excessive and see if the problem goes away. If not, then it's probably not a hardware issue, and it's probably software related like drivers or something.

I hope you didn't brick your graphics card by running furmark without breaking it in. I've done that once, and thank god it was a $35 garbage card I was trying to kill to begin with.

 

[amtseung]

The more time you spend running a power supply near its rated 100% load, the shorter the lifespan of a PSU will be. Capacitors and inductors don't have an infinite lifespan, and stuff will degrade over time. There are people whose power supplies die in 4 years, and others whose power supplies die in 12 years. It depends on many things, from the component quality of the stuff inside to "capacitor lottery" (think silicon lottery, but for the parts inside the PSU), to how they're used. You could get a $50 PSU that works great for 10 years, or a $300 PSU that dies in 3 years. I personally run a 750W PSU in a system that should never pull more than 500W for this reason. Well, that and the Techpowerup review compared to what I paid for this unit ($55 on sale).

Power supplies aren't the most expensive things in a PC, and if you can afford to buy one from a local store with a good return policy (like Fry's or Microcenter) with the sole intent of using it to test and then return after (or just keep it if you like it), that's what I'd do. I don't have buddies to borrow components from, so I have to resort to these measures.

Stability at a certain speed relies on more than what the manufacturer can test. If your motherboard isn't up to par and can't even provide its 75W through the PCIe slot, then stability will go straight out the window. If your power supply can't provide a steady voltage and/or amperage, stability will go straight out the front door. If neither are up to par, you should notice your CPU performance decreasing and/or becoming unstable when both CPU and GPU come under load, like in gaming.

With how GPU Boost tends to work, there's no point really underclocking the card, as it's probably easier to undervolt the card and let GPU boost clock itself accordingly. If undervolting the card still doesn't fix your problem, we can point a pretty sure finger at power delivery.