Black screens but not complete lockup

DavidVioMC

Honorable
Apr 25, 2016
403
1
10,865
I have been having this issue since I upgraded my GPU and PSU, where while gaming, I'll get a random black screen or a full red screen or full purple screen. When I get a coloured screen, I have to restart my PC, if I get a black screen, I can't do anything except Ctrl + Alt + Delete and close the game thru task manager, because the game dosen't stop responding or freezes, but I can't tab back in, when I try to tab back in, I just get a small black bar in top left coner and I'll hear live sounds but nothing else, when I close the game thru task manager, I can re-open it and carry on playing, but all my monitoring software will say 0 to everything GPU related, temp, clocks ect, and when I go into Afterburner, it will say "GPU1 lost connection" or something like that, but when I re-open a game, I expierence no performance drop and I have to restart my PC to see the temps ect again. This happened after I upgraded my GPU and PSU, I upgraded my GPU from my old 750ti to GTX 670 so I can play in sorround mode and I had to upgrade my PSU from EVGA 430W to Corsair CX600 since I needed one more PCIe power connector which my old PSU didn't have. I seem to get a black screen, once an hour which is very annoying and a red/purple screen maybe once a day, sometimes once every two days.My temps are fine, my PC is water cooled and my CPU temps are around 47c and my GPU is around 42c. My specs are;
i5 2500
EVGA GTX 670
Gigabyte GA-H61-DS2 Rev 2.0 Mobo
Corsair CX600 PSU
Kingston HyperX Fury 2x4GB DDR3 RAM
Drevo X1 120GB SSD
Toshiba 1TB HDD
PC water cooled by 4W pump, 1 240mm rad and 1 120mm rad with 6 Corsair SP120 fans in pull/pill config.
 
Solution
I don't know the specs for your specific GPU. So I can't give you specific advice about the voltage levels, although you could find them online, if you want to do more precise research. What I can tell you though is to lower the vcore one step at a time after you manage to find a stable clock spot. So for example if you find stability at -100Hz, try to lower the voltage 1 step. If that works try another step down and so on. Don't go to low though. Searching online about the 670 voltage limits (up & down) will certainly help you in your quest.

Now if you want to gain back some performance you could try to increase the GPU clock. So after finding a stable voltage, try to increase the clock speed by +20Mhz and so on. If you reach the...
Are both parts new or used? Your case points to a GPU failure but on the other hand that specific PSU model isn't top notch. It could be very helpful if you do manage to test your system with another PSU. If the issues go away then it's the PSU, if not, I'm afraid that your GPU is failing. In that case you could play with the voltage levers of your GPU and its VRAM in order to increase its stability but don't have high hopes.
 

DavidVioMC

Honorable
Apr 25, 2016
403
1
10,865




I bought both PSU and GPU used off Ebay, the GPU guy claimed the GPU was perfectly fine and gamed on it at 1080p till the day he put it up for sale and the PSU was from an IT store which claimed it was pulled out from a working PC, thats it. I did use DDU to un-install the old drivers from my old GPU before installing the new ones. I'm unsure but I dropped down the power limit in after burner from 100% to the minimum which is 71% and I have been gaming for about two hours without any issues, it could be just a lucky day or it actually fixed/improved the problem. I plugged in a watt meter to the PC and it only draws 293w at full load, so the 600w PSU shouden't have an issue unless it's damaged internally as I'm not an expert when it comes to PSU but I don't think a 300w would be too much for a 600w PSU even if it's poor quality. Unfrotunely I don't have another PC/PSU that I can test with, I can try buy a EVGA 500W PSU since I had a 430W version before I got the CX600 and I had no issues even tho it's a lower grade and the EVGA 500W new cost around the same as used CX600 so I wouden't be losing money, just 100w of power which I don't really need.

 
If that specific PSU (all the green Corsair CX models, not the newer grey ones) are notorious online for stability issues. That is because they use very low quality capacitors that age much faster and can cause stability issues after 1 or 2 years. If indeed you have such a model, used and you don't know how old it is, then you don't have a stable PSU. A PSU's Wattage output has nothing to do with its ability to provide clean and stable power. For example your PSU may be struggling to output more than 250-300W.

That said the GPU could also be responsible. It's used, you don't know how old it is, how "hard" it has been used by its previous owner and if it has been heavily overclocked. The fact that you managed to increase its stability by lowering the power limit [you could have done the same by lowering the GPU clock, or increasing the GPU voltage (which is more dangerous)], point to an aged GPU core, but you can't also rule out a possible power issue (PSU). That's why you must test your system with another PSU. You can also test it with another GPU with a similar power consumption (not a lower power one), but I think that this is more difficult. That way you'll draw conclusions about two components. Both of them are used and one of them is certainly bad or failing. Good luck
 

DavidVioMC

Honorable
Apr 25, 2016
403
1
10,865

So after trying different games, lowering the power limit has only helped reducing the black screen in certain games such as CSGO but they still do occur with the power limit at 71 just not so often, So I decided to under-clock the card just like you said, so I started off with 200mhz both on Memory and Core and that seemed to completely eliminate the whole black screen. I haven't tried lowering the underclock so I'm unsure if I can go to something like 100Mhz without black screens. Since lowering power limit only helped a little in certain games and under-clocking the card by 200Mhz eliminates the black screens, does it mean the card is at fault here, or does under-clocking it by 200Mhz uses a lot less power which is enough for the PSU to handle? If the PSU is the issue, I can still return it and I can buy anotherone tomorrow, I'm unsure if I can buy another 670 for as cheap as I bought mine for. If it's still un-clear if it's the PSU, I can buy it anyway and see if that helps it. I would rather buy a new PSU than a new card, espically since I have my card watercooled, I'm not looking forward to draining my loop and reverting the card back to original blower cooler.

 
Generally GPUs age must faster then PSUs. That's why buying used GPUs is like gambling. The GTX 670 is pretty old and you don't know how much it's been abused by it's previous owner. He may have heavily overvolted and overclocked the GPU core for years. That GPU has stability issues from aged electronics (GPU core or VRAM) or from unstable power delivery. So it could also be the PSU. You can't be certain. By lowering the GPU clock you manage to bring down the GPU power consumption thus stressing the PSU less. Now if the PSU's capacitors are in bad condition, they could be responsible for the stability issues. However I have a feeling it's more like a 70% GPU issue and 30% PSU. The lower clocks probably dropped the power consumption buy 30-40W so it doesn't make a lot of difference for a 600W PSU (unless it's in horrible condition). In your case testing with another PSU seems to be much easier. Don't buy a new one, also if the PSU is fine and you return it, you may get another used one in worse condition. Just find one from a friend for testing purposes.

Now if the GPU is indeed faulty and you can't return it for whatever reason, you can still tweak it a bit in order to make it more stable, but keep in mind that it may not last for more than a year. You have to find out what component is failing (GPU core or VRAM) and lower that specific clock. You have to do a lot of testing. Don't lower the GPU clock and the VRAM clock at the same time. That way you'll find who is causing the stability issues and you'll able to more precisely fine tune the clock speed of that component in order not to lose a lot of performance. By lowering both clocks (especially that 200 Mhz on the GPU core) you are giving up a lot of performance. You also have the option of playing with voltages in order to increase stability at a given clock speed but this could also make the situation even worse. So be careful.
 

DavidVioMC

Honorable
Apr 25, 2016
403
1
10,865

Thank you for such a detailed post. I guess I learned it the hard way when it comes to used hardware, but I don't really have another choice since I can't afford new cards and barely can afford older used cards due to the whole cryptocurrency apocalypse, I know couple of years ago I was thinking about getting two GTX 980 and SLI them since they were affordable, now, I can't even afford one due to ridiculous prices. The GTX 670 is a good enough card for my use anyway, I only play games like CSGO which gets me around 100fps even with the card being underclocked by 200Mhz, I only noticed a slight drop in games like GTA V but I don't play it often so it's not a big of a deal. I guess I'll spend this whole week testing the card, I'll bring it to it's stock clocks and to 100% power limit and I'll start from just underclocking the VRAM by 50Mhz every time I get a black screen, if I get to -200Mhz and I'm still getting them, I'll bring VRAM clocks back to stock and start underclocking the GPU core by 50Mhz, if again I get to -200Mhz and still getting black screen, then I'll try lowering both VRAM and GPU by 50Mhz, I'll reply with the result. Also a small side question, if it was indeed the PSU having stability issues and causing the black screens, wouden't it also effect the CPU and possibly giving a BSOD from time to time along with the black screens? Or is it because the i5 2500 has a TDP of 95w while the GTX 670 is 170w? Thanks

 
As I said before you can play with the voltage levers (GPU core, VRAM) of your card. I wouldn't advise you to increase them as it may me too dangerous for your old card but on the other hand you can decrease them. Let me give you an example. Say you reach stability levels with your GPU core down 200Mhz. You can then decrease the GPU core voltage a little bit, since it doesn't need the previous voltage level at the new clock speed. What would you gain by that? You may be able to gain back some GPU clock speed by a small amount (50-80Mhz), thus reducing the overall performance hit a little bit. The biggest gain though is that you''ll definitely extend the GPU lifetime by a year or two. Lower voltage levels increase the longevity of electronic components. However be very careful, because if you decrease the voltage a lot you'll lose again stability.

As for CPU vs GPU power consumption question, there are two reasons why a graphics card can stress a PSU more than a GPU. The obvious one is the bigger power consumption of a GPU vs a CPU. The 2nd reason is that all modern GPUs in order to be more power efficient and to decrease their overall power consumption (careful, not the maximum one), monitor the GPU load, clocks and temperature all the time and adjust power consumption accordingly. The goal is to have the lower power consumption for any given load. The result is a constant fluctuations in the power they draw from the PSU in a 1/10 sec. So for example a GPU could go from 30W to 180W instantly and then back to 30W and it could it 10 times within a second. As a result many PSUs (bad, low quality), can't handle this. CPUs on the other hand don't cause so much sudden power fluctuations and are more PSU "friendly".
 

DavidVioMC

Honorable
Apr 25, 2016
403
1
10,865

Thank you for clearing that up, I never used core voltage even when I OC'ed my previous GPU's since I heard if too much is applied, it can cause damage just like you said so I left them at stock voltage. I use MSI Afterburner for changing clocks and I assume -1 on core voltage is 1 millivolt? Unless otherwise, also what would you recommend changing it to? It seems that VRAM is not at fault here, I was adding -50Mhz with every black screen until I got to -200Mhz and I was still getting them so I reverted back to stock speed and started with GPU clock, I'm currently at -50Mhz without black screens but I'm still testing it and can't say for sure.
 
I don't know the specs for your specific GPU. So I can't give you specific advice about the voltage levels, although you could find them online, if you want to do more precise research. What I can tell you though is to lower the vcore one step at a time after you manage to find a stable clock spot. So for example if you find stability at -100Hz, try to lower the voltage 1 step. If that works try another step down and so on. Don't go to low though. Searching online about the 670 voltage limits (up & down) will certainly help you in your quest.

Now if you want to gain back some performance you could try to increase the GPU clock. So after finding a stable voltage, try to increase the clock speed by +20Mhz and so on. If you reach the point of instability you have the option of increasing the voltage one step up (still down from the default voltage), or decrease the clock by -10Mhz depending on what you want. After a lot of testing you should find your desired stability level.

That last step is optional and from what I can tell at -50Mhz you may not be able to find stability after lowering the voltage. If you want to increase the GPU's longevity you may have to go a little lower say -80Mhz in order to be able to bring down the GPU core voltage a little bit (one step). Good luck.
 
Solution