Graphics Card turns itself off while gaming. Urgent.

[StixuYT]

Hello, I have bought an EVGA GTX 670 2GB in December 2020 for a VERY good price(around 30$), but when I got it, had some very serious issues regarding temperatures, when the card was idle, the temps would hit even 60°C, when gaming the temps would go up to almost 100°C,making the card thermal throttle in every game that I played, with that happening, the games were running below expectations, a lot of stutter, terrible experience overall, that made me make the decision of cleaning the card inside and replacing thermal paste, I ended up using some MX-2 from Arctic, after the cleaning, the card was performing extremely good, temps were as low as 22°C on idle, and around 60-70°C when gaming, I was very happy at the time, but, after getting at least 60+Fps in almost every game I played, I started experiencing some issues, random crashes, restarts, stutter, temps were good, I started getting quite sad when I saw that my PC was having problems again, but after a while, everything just stopped happening, all was so good like nothing ever happened, but....something about this situation was about to change when I saw the first GPU CRASH, I was playing some CS:GO until my PC just showed a black screen for a second, and after that some random colors on both my monitors(for eg. green on my second one, and orange on my main one), I thought nothing of this at the time as it was the first time it happened, also, I had some crappy Intel board, Intel DQ67SW to be exact, and that board had some big problems, It would only accept SOME cards(for example, my ASUS GT 440 wouldn't work, black screen, not even my GTS 450, and then I decided to buy this GTX 670 which I was very happy to see a boot when I plugged it in), another problem it was having was that whenever I used the "Shut Down" button in Windows, the PC would restart, making me turn off the PSU manually from the back, but, the crashes started becoming more and more frequent, requiring me to restart the whole PC, and another thing, even if if I had an Gigabyte SSD, the computer would sometimes boot in 10s, and other times in like 5 minutes(still, because of the motherboard) but then I started saving money and I was able to buy another board, ASUS P8B75-M/CSM, I cleaned my whole PC, put the new motherboard in, put everything needed, and plugged in the PC, everything was normal, the GPU crashes were non-existent, and I thought "Oh, it was because of the motherboard, finally I can enjoy some peaceful gaming...well, no I didn't, after a while..like months..I downloaded a new game, called "Phasmophobia", I loved the game, and my PC would run the game at max settings at 60+fps, but...not so fast.. everything was running extremely fluent, but then... the first crash happened...the exact same way it was happening on the Intel board, 2 different colors on both my monitors(see photos below), Not even using the Medium settings that the game has to offer fixed this issue, Lowering the settings only reduced the amount of time the GPU would crash in. So I was forced to play with the lowest settings, which sucked, because, my GPU doesn't like low settings, whenever I use low settings in any game I have stutters, lag spikes, huge frametimes, but using the highest settings possible in any game would result in extremely playable experience, but not on this game...After a while, it started crashing even when using the lowest settings..which made me think of buying a new GPU(GTX 1050 Ti), but prices are HUGE..hopefully they will drop soon..

IMPORTANT NOTE: I use a Molex to PCI-E 6 pin because my PSU only has one 6pin connector.

Also, weird thing, my card is an EVGA model, but the guy I bought it from had an stock Nvidia cooler on the card.

Some stuff I tested:
-Another PSU(I was thinking that the GPU, being a beefy card, the power supply was not able to transfer the necessary power to the GPU, but that was not the case)
-Only one stick of RAM(Checked the RAM Stick with MemTest86 and it passed with great success, but unfortunately to no avail, crashes were still present)
-Only one monitor plugged into the GPU(same result, just one color appearing due to only monitor being connected)
-Changing A LOT of Nvidia drivers(nothing, only more frustration and time wasted)

Some stuff I haven't tested YET:
-Use the GPU in another PC, I only have pretty bad specs PCs laying around, and most of them don't even have the space necessary for the GPU to enter the CASE, towers), I might take the motherboard of one and then plug it that way see if any crash happens(I'll update)

PC Specs:
CPU: i5-2500 3.3GHz

GPU: GTX 670 2GB GDDR5

RAM: 12GB DDR3 1333MHz

PSU: FLOSTON 650wExtra 80+

BOARD: ASUS P8B75-M/CSM

[Pixelfie]

Quote

IMPORTANT NOTE: I use a Molex to PCI-E 6 pin because my PSU only has one 6pin connector.

 

Another PSU

Did you use the adapter with the other PSU?

[DANK_AS_gay]

6 minutes ago, StixuYT said:

I have bought an EVGA GTX 670 2GB in December 2020 for a VERY good price(around 30$), but when I got it, had some very serious issues regarding temperatures, when the card was idle, the temps would hit even 60°C, when gaming the temps would go up to almost 100°C,making the card thermal throttle in every game that I played, with that happening, the games were running below expectations, a lot of stutter, terrible experience overall, that made me make the decision of cleaning the card inside and replacing thermal paste, I ended up using some MX-2 from Arctic

Ok, a few things to check, can you check the hotspot temperature? It could be that one part of your card die is getting super hot while other parts are being properly cooled. Also, did you replace the thermal pads? 

The only time I have seen stuff like this is when the VRAM on one of my ancient cards would get too hot during a longer gaming session, you may want to try new thermal pads, but also the proper cooler for this unit. 

 

Also, your post is very helpful, often people come here asking for help while giving us no information, while you give in detail everything you have done to this thing.

[StixuYT]

Just now, Pixelfie said:

Did you use the adapter with the other PSU?

Yes I did, I have no other option but to use it, I currently have no other PSU that has 2 PCI-E 6pin connectors

 

[DANK_AS_gay]

Just now, Pixelfie said:

Did you use the adapter with the other PSU?

I assumed he didn't but you are right, this could also be the problem, as the MOLEX rail can't supply enough wattage for a GTX 670 lol

[DANK_AS_gay]

Just now, StixuYT said:

Yes I did, I have no other option but to use it, I currently have no other PSU that has 2 PCI-E 6pin connectors

 

Ok, you are going to want to try that, there are cheaper name brand PSUs for $40-$60 that could work.

[StixuYT]

Just now, DANK_AS_gay said:

can you check the hotspot temperature?

How can I do that?

1 minute ago, DANK_AS_gay said:

did you replace the thermal pads? 

 

I did not, how can I acquire them?

[StixuYT]

Also, another thing, If I try and downclock the card, it adds time until it crashes when using high settings.

[DANK_AS_gay]

1 minute ago, StixuYT said:

How can I do that?

I did not, how can I acquire them?

HWINFO64 will pull up all of the temperature sensors in the system, and you can navigate to the GPU temperatures, there should be a category called "hotspot" or tjmax or something like that.

 

Try to measure the distance between the VRAM chips themselves and the cooler, and you can find thermal pads on Amazon for pretty cheap, just make sure to get the correct thickness.

[StixuYT]

Just now, DANK_AS_gay said:

you can find thermal pads on Amazon for pretty cheap

Thing is, I'm not from America, I'm from Romania and stuff like this are kind of hard to find.

Also HWInfo64 has some weird glitches because of my motherboard sensors, It will say that my 5th and 6th core of CPU are like 124 degrees(I have a quad core)

[DANK_AS_gay]

Just now, StixuYT said:

Also, another thing, If I try and downclock the card, it adds time until it crashes when using high settings.

Yeah that sounds like a temperature issue, try out HWINFO64 real fast, pull up Unigine Heaven or some other stress test and watch the temperatures closely. Specifically the Hotspot, Memory/VRAM, and the VRM.

[DANK_AS_gay]

Just now, StixuYT said:

Thing is, I'm not from America, I'm from Romania and stuff like this are kind of hard to find.

Also HWInfo64 has some weird glitches because of my motherboard sensors, It will say that my 5th and 6th core of CPU are like 124 degrees(I have a quad core)

You are only looking at the GPU though, it should be fine.

[StixuYT]

Just now, DANK_AS_gay said:

pull up Unigine Heaven or some other stress test

Oh and another thing, I have stress tested the GPU multiple times using 3DMark and not a single crash, also it says the card is just like the expectations.

[Pixelfie]

10 minutes ago, StixuYT said:

Yes I did, I have no other option but to use it, I currently have no other PSU that has 2 PCI-E 6pin connectors

I think this is probably the problem. Do you know someone you could borrow their PSU from to see if it's the problem? If not try 1 6 pin connector from the PSU connected to your PC and the other connector from the other PSU. Turn the second PSU on with the paperclip trick. How to: Test a power supply unit – Corsair

[StixuYT]

Where should I look exactly?

[Spotty]

9 minutes ago, DANK_AS_gay said:

Ok, a few things to check, can you check the hotspot temperature?

 

1 minute ago, DANK_AS_gay said:

Yeah that sounds like a temperature issue, try out HWINFO64 real fast, pull up Unigine Heaven or some other stress test and watch the temperatures closely. Specifically the Hotspot, Memory/VRAM, and the VRM.

Older Nvidia cards don't support hotspot temperature readings. I can't remember if it was Ampere or Turing that it was introduced but definitely not Kepler. HwInfo might display an estimated value but it won't be a true hotspot reading.

[StixuYT]

Just now, Pixelfie said:

If not try 1 6 pin connector from the PSU connected to your PC and the other connector from the other PSU

You actually made me remember I wanted to try this. I know a lot of stuff about PCs and electronics, but this is just such a hassle.

[DANK_AS_gay]

1 minute ago, StixuYT said:

Oh and another thing, I have stress tested the GPU multiple times using 3DMark and not a single crash, also it says the card is just like the expectations.

 

Try running them for a while, like FurMark or Unigine Heaven, and give them like 30 mins. How long were you playing the game before it crashed?

[DANK_AS_gay]

2 minutes ago, Spotty said:

Older Nvidia cards don't support hotspot temperature readings. I can't remember if it was Ampere or Turing that it was introduced but definitely not Kepler. HwInfo might display an estimated value but it won't be a true hotspot reading.

Ah, I didn't know that, I just figured that such a reading would be on all cards since like 2005, but I guess not.

[StixuYT]

Just now, DANK_AS_gay said:

Try running them for a while, like FurMark or Unigine Heaven, and give them like 30 mins. How long were you playing the game before it crashed?

If your talking about Phasmophobia(the game in the recording) about 5m, but sometimes it crashes even when I just join the game and enter the house(If it crashed recently), so something about temperature has to be, cause why would it crash much faster only when it has already crashed a few minutes ago?

[DANK_AS_gay]

3 minutes ago, Spotty said:

 

Older Nvidia cards don't support hotspot temperature readings. I can't remember if it was Ampere or Turing that it was introduced but definitely not Kepler. HwInfo might display an estimated value but it won't be a true hotspot reading.

Also, would you agree on the memory temperature issue? He said the thermal pads haven't been replaced since he got it, and we don't know if the stock ones are still on there, and if they are, this issue sounds a lot like the issue I had running Need For Speed on my Radeon Mobility 9600 when the Pentium M got things too hot.

[DANK_AS_gay]

6 minutes ago, StixuYT said:

ut 5m, but sometimes it crashes even when I just join the game and enter the house(If it crashed recently), so something about temperature has to be, cause why would it crash much faster only when it has already crashed a few minutes ago?

So you are saying that if you reboot immediately and play the game again, it crashes faster? 

I mean, the thermal pads and PSU are things you will want to get in general to prolong the life of your components, as it sounds like you do not have much money to spend on them. So, while I think it is the thermals, I don't know 100%, and you likely want to just play the game, so I would order both. If the thermal pads work by themselves, great! Return the PSU (I would keep it, don't trust a DELL or HP PSU from some office computer), if the thermal pads do not work, then you already have the PSU to replace the old one with. 

[StixuYT]

Just now, DANK_AS_gay said:

So you are saying that if you reboot immediately and play the game again, it crashes faster? 

I mean, the thermal pads and PSU are things you will want to get in general to prolong the life of your components, as it sounds like you do not have much money to spend on them. So, while I think it is the thermals, I don't know 100%, and you likely want to just play the game, so I would order both, and if the thermal pads work by themselves, great! Return the PSU (I would keep it, don't trust a DELL or HP PSU from some office computer), if the thermal pads do not work, then you already have the PSU to replace the old one with. 

I will currently try what Pixelfie suggested as it's in theory the same thing, so If the crashes don't happen anymore while using both the pci-e 6pin connectors, then it's the psu. If it still happens, I will try and replace the thermal pads.

[Hairless Monkey Boy]

8 minutes ago, DANK_AS_gay said:

Also, would you agree on the memory temperature issue? He said the thermal pads haven't been replaced since he got it, and we don't know if the stock ones are still on there, and if they are, this issue sounds a lot like the issue I had running Need For Speed on my Radeon Mobility 9600 when the Pentium M got things too hot.

Based on my skimming the OP and the rest of the thread, that'd be my guess. Emphasis on skimming.

[StixuYT]

I connected the other PSU's PCI-E 6pin connector to my GPU while my PSU's PCI-E connector is on at the same time and it currently works, I'll update after 10 mins of playtime, if it crashes.