GPU at factory settings causing screen blackouts during game.

[BlurKillerEX]

Had almost the exact same experience with afterburner being the culprit of crashes.

Wasted a lot of time merry go round trying other solutions but the crashes only stopped when i uninstalled afterburner

Yah I really think this could be the issue. Lucky I never jump the gun to go buy new PSU like so many is suggesting.

 

[Richliao]

Any similar afterburner program to recommend?

 

[watzup_ken]

Yah I really think this could be the issue. Lucky I never jump the gun to go buy new PSU like so many is suggesting.

Never jump to buy a new part without validating it is the problem. That's why I suggested that you test it first before buying or you may end up wasting your money and not fixing the problem. Glad you resolve the issue.

 

[watzup_ken]

Just set power limit to 75% in afterburner. Works for me. I'm using sf750 BTW and before I did this, I was crashing like no tomorrow on all games. Now I can play for 6 hours straight without a single crash. Using latest drivers and msi 3080 ventus just like you. Don't need to undervolt or anything, just lower the power limit. Try it and let us know.

Not sure how much performance lost when setting power limit to 75%? It certainly is better to have the card to work without crashing, but is not the ideal solution. A good 750W should be able to run this card without any power issue.

 

[Reavl]

Yah I really think this could be the issue. Lucky I never jump the gun to go buy new PSU like so many is suggesting.

since u narrow down to afterburner issue then maybe some software in your background having issue with afterburner if you wanna use your afterburner, if not you can try software of your gpu to tune your gpu

 

[Lastwishes]

Just set power limit to 75% in afterburner. Works for me. I'm using sf750 BTW and before I did this, I was crashing like no tomorrow on all games. Now I can play for 6 hours straight without a single crash. Using latest drivers and msi 3080 ventus just like you. Don't need to undervolt or anything, just lower the power limit. Try it and let us know.

Using the same card. No crashes on stock settings, but the card does run very hot even with my case fully kitted out with fans. Your card may just be crashing due to overheating.

Out of curiosity, I lowered the power limit to 75% for mine and did a quick benchmark test. It's actually a significant drop in performance, with the core clock jumping all over the place (between ~1400 to 1900, averaging around 1500+ most of the time, with temperature about 74 degrees on mine). I suspect that if you set up a conservative undervolt, you'll be able to get it stable without losing much performance from stock; my core clock is sustained at 1845 throughout, with temperature around 71 degrees.

The different in the 3dmark GPU score between the two is about 2400 pts.

 

[Dark_Skye]

At 75% I'm getting about 1725mhz gpu clock, temp is 61c playing division 2 and ac on. Without ac almost close to 67c.
GPU load is 250w max.
With stock fans, stock curve, stock power limit the card hits 83c in my case without ac.
I'll run 3d mark tonight and update my score.
Tbh, I never bother running benchmark because all I wanted was games to play fine and not crash. Lol.

 

[Dark_Skye]

Ocp tripping causes psu to shut down, not only gpu. If its only gpu its probably something else. 750w should definitely be enough, maybe try uninstall afterburner. Also, do not daisy chain your gpu power connection if u r doing so, even if it save space in a sfx build.

My crashes are a wide variety. Cna be black screen during game, can be instant crash after loading games, can be 5 min into a game, sometimes can feel when the game is about to crash because it's like the game is loading assets and there's a slight stutter and then boom! It crashed.

I'll uninstall afterburner and see if there's a difference. If it still crashes because the power limit is no longer there, can I assume it's my gpu that's faulty or psu needs to rma?

Edit: when you say psu trip, meaning the com just shuts down with zero power inside and does a restart? If it's gpu, that means just crash or hang or bsod but com still running? If so, mine seems to be gpu crash and not psu because I don't see the com restarting or no power at all. My crashes all have the com still running but nothing on screen or ctd.

 

[Lastwishes]

At 75% I'm getting about 1725mhz gpu clock, temp is 61c playing division 2 and ac on. Without ac almost close to 67c.
GPU load is 250w max.
With stock fans, stock curve, stock power limit the card hits 83c in my case without ac.
I'll run 3d mark tonight and update my score.
Tbh, I never bother running benchmark because all I wanted was games to play fine and not crash. Lol.

It kinda sounds like your card is just overheating badly. Lowering the power limit is similar to doing an undervolt, except you lose more performance by leaving it to the card to decide what the core clock is.

I also not too bothered about the benchmark scores I use them to test my settings where I optimise for fan noise and temperature instead.

 

[Dark_Skye]

It kinda sounds like your card is just overheating badly. Lowering the power limit is similar to doing an undervolt, except you lose more performance by leaving it to the card to decide what the core clock is.

I also not too bothered about the benchmark scores I use them to test my settings where I optimise for fan noise and temperature instead.

I did undervolt but left power limit alone and the card still crashes. Only power limit at 75% is able to ensure it doesn't crash. You think this is a gpu fault? I can send it back to Korea for rma.

 

[Lastwishes]

I did undervolt but left power limit alone and the card still crashes. Only power limit at 75% is able to ensure it doesn't crash. You think this is a gpu fault? I can send it back to Korea for rma.

Do you remember your settings for the undervolt? It might be just not conservative enough.

 

[BlurKillerEX]

At 75% I'm getting about 1725mhz gpu clock, temp is 61c playing division 2 and ac on. Without ac almost close to 67c.
GPU load is 250w max.
With stock fans, stock curve, stock power limit the card hits 83c in my case without ac.
I'll run 3d mark tonight and update my score.
Tbh, I never bother running benchmark because all I wanted was games to play fine and not crash. Lol.

My card at stock settings hit 78 deg at load. I think around 80 deg is still considered an ok operating temperature for gpu though.

 

[watzup_ken]

Using the same card. No crashes on stock settings, but the card does run very hot even with my case fully kitted out with fans. Your card may just be crashing due to overheating.

Out of curiosity, I lowered the power limit to 75% for mine and did a quick benchmark test. It's actually a significant drop in performance, with the core clock jumping all over the place (between ~1400 to 1900, averaging around 1500+ most of the time, with temperature about 74 degrees on mine). I suspect that if you set up a conservative undervolt, you'll be able to get it stable without losing much performance from stock; my core clock is sustained at 1845 throughout, with temperature around 71 degrees.

The different in the 3dmark GPU score between the two is about 2400 pts.

I am not surprised that you will lose a lot of performance with a 25% power limit reduction. Even at 100% the card is power limited and running around 1.9 GHz. Depending on the quality of the chip, the reduction in power limit may have a significant drop in clock speed.

As for the heat of the Ventus, did you try increasing the fan speed? MSI is known for using very low fan speed, so may be worth exploring it tinkering with the fan curve helps.

 

[Lastwishes]

I am not surprised that you will lose a lot of performance with a 25% power limit reduction. Even at 100% the card is power limited and running around 1.9 GHz. Depending on the quality of the chip, the reduction in power limit may have a significant drop in clock speed.

As for the heat of the Ventus, did you try increasing the fan speed? MSI is known for using very low fan speed, so may be worth exploring it tinkering with the fan curve helps.

Already using a custom curve, which is set a lot more aggressive than MSI's stock curve. I actually disassembled my card and reapplied a better thermal paste as well; this gave me a noticeable drop in temperatures.

I did a fair amount of testing and found that my card runs rather hot if left at stock settings, so much so that it throttles itself way before the power limit is hit (at stock settings, I almost never saw 1.9 ghz, while max clocks are actually set at 2.1 ghz).

That's why I was wondering if Dark_Skye's undervolt was conservative enough, because reading online guides, 850mv/1875 seems to be the range that most people would go to for an initial undervolt, which to be fair, is a good point to start with, unless the airflow of the case used is not great. In practice, I found that even at 850mv/1875 mhz, there was no difference in temperature compared to stock settings (about 80 degrees); in a poorly ventilated case, the card is no where close to touching the 100% power limit. Only when I hit about 825mv/1845mhz, there was maybe a 3-4 degree improvement.

In the end, I settled with an undervolt of 800mv/1815mhz, because I preferred a lower power draw and fan noise, and also because my own set up holds the card back: I run 2 m2 ssds and 8 hard drives in the same case (a Fractal Design R6). While it's a good case and I'm fully kitted out with decent fans (noctuas and gentle typhoons), there's a lot of heat generated from the drives, and my preference is to optimise fans for silence rather than performance. I imagine that in a case with poor air-flow, the gpu temperatures would be equally bad.

For reference, with all the hard drives unplugged and fans running at full blast, my temperatures are under 70 degrees for 850mv/1875 running stress tests (usually 80+ with some clock throttling), which isn't great compared to some of the temperatures reported for other models. Except this was after repasting with arctic mx4, so I'm kinda wondering if the cooling for the Ventus is just not that great compared to say, an Asus TUF 3080. Either way, I'm not too bothered; I don't chase benchmarks, and the difference in actual fps between 1.9ghz and 1.8ghz is so negligible that I prefer to just keep my card cooler and quiet.

 

[watzup_ken]

Already using a custom curve, which is set a lot more aggressive than MSI's stock curve. I actually disassembled my card and reapplied a better thermal paste as well; this gave me a noticeable drop in temperatures.

I did a fair amount of testing and found that my card runs rather hot if left at stock settings, so much so that it throttles itself way before the power limit is hit (at stock settings, I almost never saw 1.9 ghz, while max clocks are actually set at 2.1 ghz).

That's why I was wondering if Dark_Skye's undervolt was conservative enough, because reading online guides, 850mv/1875 seems to be the range that most people would go to for an initial undervolt, which to be fair, is a good point to start with, unless the airflow of the case used is not great. In practice, I found that even at 850mv/1875 mhz, there was no difference in temperature compared to stock settings (about 80 degrees); in a poorly ventilated case, the card is no where close to touching the 100% power limit. Only when I hit about 825mv/1845mhz, there was maybe a 3-4 degree improvement.

In the end, I settled with an undervolt of 800mv/1815mhz, because I preferred a lower power draw and fan noise, and also because my own set up holds the card back: I run 2 m2 ssds and 8 hard drives in the same case (a Fractal Design R6). While it's a good case and I'm fully kitted out with decent fans (noctuas and gentle typhoons), there's a lot of heat generated from the drives, and my preference is to optimise fans for silence rather than performance. I imagine that in a case with poor air-flow, the gpu temperatures would be equally bad.

For reference, with all the hard drives unplugged and fans running at full blast, my temperatures are under 70 degrees for 850mv/1875 running stress tests (usually 80+ with some clock throttling), which isn't great compared to some of the temperatures reported for other models. Except this was after repasting with arctic mx4, so I'm kinda wondering if the cooling for the Ventus is just not that great compared to say, an Asus TUF 3080. Either way, I'm not too bothered; I don't chase benchmarks, and the difference in actual fps between 1.9ghz and 1.8ghz is so negligible that I prefer to just keep my card cooler and quiet.

I recall seeing this mod on the Ventus 3x that helped the GPU temps in a SFF.
https://www.youtube.com/watch?v=Y5WcIAa7Jv8
The problem it seems is the slow fan that came along with the Ventus it seems. The heatsink in my opinion looks beefy enough to cool the card at stock speed for sure.

 

[Dark_Skye]

I recall seeing this mod on the Ventus 3x that helped the GPU temps in a SFF.
https://www.youtube.com/watch?v=Y5WcIAa7Jv8
The problem it seems is the slow fan that came along with the Ventus it seems. The heatsink in my opinion looks beefy enough to cool the card at stock speed for sure.

I'm doing this. With 2 noctua nf-a12x25 running as exhaust. That's why at power limit 75% but no undervolt gpu full load temp is 61c.

 

[Dark_Skye]

Already using a custom curve, which is set a lot more aggressive than MSI's stock curve. I actually disassembled my card and reapplied a better thermal paste as well; this gave me a noticeable drop in temperatures.

I did a fair amount of testing and found that my card runs rather hot if left at stock settings, so much so that it throttles itself way before the power limit is hit (at stock settings, I almost never saw 1.9 ghz, while max clocks are actually set at 2.1 ghz).

That's why I was wondering if Dark_Skye's undervolt was conservative enough, because reading online guides, 850mv/1875 seems to be the range that most people would go to for an initial undervolt, which to be fair, is a good point to start with, unless the airflow of the case used is not great. In practice, I found that even at 850mv/1875 mhz, there was no difference in temperature compared to stock settings (about 80 degrees); in a poorly ventilated case, the card is no where close to touching the 100% power limit. Only when I hit about 825mv/1845mhz, there was maybe a 3-4 degree improvement.

In the end, I settled with an undervolt of 800mv/1815mhz, because I preferred a lower power draw and fan noise, and also because my own set up holds the card back: I run 2 m2 ssds and 8 hard drives in the same case (a Fractal Design R6). While it's a good case and I'm fully kitted out with decent fans (noctuas and gentle typhoons), there's a lot of heat generated from the drives, and my preference is to optimise fans for silence rather than performance. I imagine that in a case with poor air-flow, the gpu temperatures would be equally bad.

For reference, with all the hard drives unplugged and fans running at full blast, my temperatures are under 70 degrees for 850mv/1875 running stress tests (usually 80+ with some clock throttling), which isn't great compared to some of the temperatures reported for other models. Except this was after repasting with arctic mx4, so I'm kinda wondering if the cooling for the Ventus is just not that great compared to say, an Asus TUF 3080. Either way, I'm not too bothered; I don't chase benchmarks, and the difference in actual fps between 1.9ghz and 1.8ghz is so negligible that I prefer to just keep my card cooler and quiet.

Yes, that's the voltage I used when trying the undervolt. However, not much difference for me as the card keeps crashing. Power limit at 100% with undervolting doesn't work for my card.

Edit: about 860mV with 1850mhz agar agar can't recall exactly how much the undervolt was

 

[Lastwishes]

I recall seeing this mod on the Ventus 3x that helped the GPU temps in a SFF.
https://www.youtube.com/watch?v=Y5WcIAa7Jv8
The problem it seems is the slow fan that came along with the Ventus it seems. The heatsink in my opinion looks beefy enough to cool the card at stock speed for sure.

Yep. Have seen this before. They're right when saying MSI's fan curve is too conservative. It's a very nice set up, but only works if I'm changing my case. Maybe when I retire the R6 for a DIY home server, which will take away all the heat from the hard drives as well

Yes, that's the voltage I used when trying the undervolt. However, not much difference for me as the card keeps crashing. Power limit at 100% with undervolting doesn't work for my card.

Edit: about 860mV with 1850mhz agar agar can't recall exactly how much the undervolt was

Ah I see. From what I observed, undervolting my Ventus at 850 mv was no different from running at stock (same temperatures, except that the core clock doesn't flit up and down between 1815 to 1875 depending on temperature). I'm not too sure about why it crashes repeatedly though (when nearing the temperature limit, the card actually decreases the core clock on its own). Perhaps can consider trying a more conservative undervolt, maybe start at 800mV/1780mhz, while keeping the power limit at 100%, and see if it still crashes? If it's stable, then at least you know it's not the power limit. You can consider going up to 800mV/1800 mhz (which is still conservative, but much less of a performance hit compared to a -25% to power limit) and keep it there.

 

[Dark_Skye]

OK. I'll try that 800mv tonight and see how it goes. I have a 6900xt merc that's coming tomorrow so I'm going to try that as well see if I get any weird crashes.

 

[Dark_Skye]

Thanks to all bros and gurus here. Using afterburner, power limit was set to 100% and then a undervolt with 800mV and 1850mhz was set. Card crashed within first 3 minutes of Divison 2. Decided to try and uninstall afterburner, guess what? No more crashes. Not one at all. Games load at first try, played Divison 2 for 3 hours straight with no issues. Cyberpunk seems fine as well. What gpu tuner program do you guys recommend if i still want to undervolt the card?