[SOLVED] 3080 TI FE crashing when it boosts past 2100mhz any way to limit?

[masteve]

Hey i have a custom loop cooled 3080 ti FE, was playing with some OC recetly all temps/power draws are fine, just any time any game boosts past 2100mhz the game will crash with a driver error. I know this has been a problem for the 30xx series is there any work around/limit i can use to stop this happening? I would love to be able to have a limit to stop the gpu bosoting over 2100mhz. Some games at a 170mhz dont boost past 2100mhz and sit aroiund 2070 super stable for hours, other games with a 130mhz OC boost over 2100? Confused.
Thanks for your help : )

PC SPECS:
5950X.
Gigabyte X570 MASTER
Ram : Team Group RIPPED Edition 16GB (2x8GB) DDR4 PC4-28800C14 3600MHz (TDPPD416G3600HC14CDC01) x2
RTX 3080 TI FE
PSU: 1000W CORSAIR HX1000

 

[Koekieezz]

what are the specs?

 

[masteve]

what are the specs?

So sorry edited them in, thanks.

 

[sizzling]

As you know this is common on 3000 series. My 3080 is the same, if OC’d and it goes over 2100mhz it crashes. However it really doesn’t matter, at best it’s going to make a tiny FPS difference. I gave up overclocking my Gigabyte 3080 Gaming OC beyond its factory overclock, the performance difference was very small l, made no improvement to game play yet required a bump in power and significant increase in heat and noise. I’ve more recently undervolted it with surprisingly good results.

 

[Phaaze88]

If those LN2 videos are anything to go by, it could also be that thermals aren't low enough to sustain those higher clocks.
I understand that those samples may be too short - they do a run of Port Royal with their OC and are done after a pass? So yeah, not reflective of in game performance, but something to think about.

 

[masteve]

Yeh im currently running everythinmg under 70c as im watercooled, I guess this is how it has to be then, Its so frustrating that some games can have as 170mhz oc and not boost past 2100, and some have a 100mhz OC and boost past 2100mhz so Ill have to use different profiles for diffent games or just run a 80mhz OC on everything and ignore it. Quite frustrating to loose some potential of the card! Oh well, at least I got it at RRP here in the UK(£1049) which makes it £250 cheaper than most 3080 none ti's atm : ) Which makes my card still rly good price performance in the current markett! lol

Thanks for the replys I just wish there was some limit on the card that stopped it going over 2100mhz

 

[Phaaze88]

I just wish there was some limit on the card that stopped it going over 2100mhz

Overclocking hasn't been worth it on these cards since the 10 series. Nvidia broke it for more casual users with the Gpu Boost algorithm.
It'll make its own adjustments regardless of yours, so that 170mhz OC was never really stable to begin with - and you've no real way to prove it, because it can't be sustained in everything.
The hotter the core runs, Gpu Boost will yield shorter and lower boost bins.
The more frequently it runs into the board power limit - also worse boost bins.
Overclocking increases thermals and the frequency of running into power limits. Gpu Boost doesn't like either too much.
Undervolting has been more effective on these cards.

Run Msi Afterburner. Unlink the Power and Temperature Limits - there's a paperclip like icon next to it depending on the skin being used.
Max out only the power limit, and click apply.
Play your games with Afterburner's own hardware monitor running. Make sure both Core Clock and Gpu Voltage are visible.
If Gpu Voltage is not visible, you can open it by going into Settings > General tab, checking 'Unlock voltage monitoring', and then go to the Monitoring tab, find and check Gpu Voltage. Then click Apply.
After a few minutes of playing, check Afterburner's hardware monitor for the MAX Core Clock and Gpu Voltage. Memorize, or write 'em down.
Please close the game, and open Afterburner's Curve Editor. Take the max gpu voltage you recorded, and subtract 0.05v from it. Find the voltage point in the Curve Editor that matches it, or is the closest match, and click on it.
Then use the up arrow key and raise the frequency back up to the max core clock the gpu touched. Lock it with the L key, and click Apply again.
Save the settings in one of the numbered profiles and lock them. Then click the reset key.
Done.

When you want to start up a game, open Afterburner, click the numbered profile it was saved at, and click apply. When you're done, click the reset button and close Afterburner.

 

[RodroX]

Yeh im currently running everythinmg under 70c as im watercooled, I guess this is how it has to be then, Its so frustrating that some games can have as 170mhz oc and not boost past 2100, and some have a 100mhz OC and boost past 2100mhz so Ill have to use different profiles for diffent games or just run a 80mhz OC on everything and ignore it. Quite frustrating to loose some potential of the card! Oh well, at least I got it at RRP here in the UK(£1049) which makes it £250 cheaper than most 3080 none ti's atm : ) Which makes my card still rly good price performance in the current markett! lol

Thanks for the replys I just wish there was some limit on the card that stopped it going over 2100mhz

You have such a powerfull system:

5950X.
Gigabyte X570 MASTER
Ram : Team Group RIPPED Edition 16GB (2x8GB) DDR4 PC4-28800C14 3600MHz (TDPPD416G3600HC14CDC01) x2
RTX 3080 TI FE
PSU: 1000W CORSAIR HX1000

If I were in your place, I would remove any OCing, and just enjoy what I got without trying to break it for +5 FPS. Really is not worth it.

... Unless of course you have soo much money that you don't care.

Where I live people would kill to have such a high end system... and I really mean kill.

Cheers!

 

[masteve]

Thanks for the replys guys, I do apreciate what I have, ive allways tweaked my stuff since my first PC in the late 90's I montior my stuff pretty heavily and run stress tests/temp test to make sure im super ok. I have a custom loop, with a corsair waterblock on the GPU (the EK ones where just tooo overprices for a FE version), anyways this is my current OC profile that defualts with windows boot.

It never goes over 375W draw and hardly ever over 1v temps are around 55-65c on the core and hotspot/ram goes stays around 60c but can spike to 77c.

This is my 3DMARK firestike result if anyone is intrested. With a 180mhz OC and no crash. (Now u see my frustration )

And a picture of the rig with my old 2080TI in it.

Overclocking hasn't been worth it on these cards since the 10 series. Nvidia broke it for more casual users with the Gpu Boost algorithm.
It'll make its own adjustments regardless of yours, so that 170mhz OC was never really stable to begin with - and you've no real way to prove it, because it can't be sustained in everything.
The hotter the core runs, Gpu Boost will yield shorter and lower boost bins.
The more frequently it runs into the board power limit - also worse boost bins.
Overclocking increases thermals and the frequency of running into power limits. Gpu Boost doesn't like either too much.
Undervolting has been more effective on these cards.

Run Msi Afterburner. Unlink the Power and Temperature Limits - there's a paperclip like icon next to it depending on the skin being used.
Max out only the power limit, and click apply.
Play your games with Afterburner's own hardware monitor running. Make sure both Core Clock and Gpu Voltage are visible.
If Gpu Voltage is not visible, you can open it by going into Settings > General tab, checking 'Unlock voltage monitoring', and then go to the Monitoring tab, find and check Gpu Voltage. Then click Apply.
After a few minutes of playing, check Afterburner's hardware monitor for the MAX Core Clock and Gpu Voltage. Memorize, or write 'em down.
Please close the game, and open Afterburner's Curve Editor. Take the max gpu voltage you recorded, and subtract 0.05v from it. Find the voltage point in the Curve Editor that matches it, or is the closest match, and click on it.
Then use the up arrow key and raise the frequency back up to the max core clock the gpu touched. Lock it with the L key, and click Apply again.
Save the settings in one of the numbered profiles and lock them. Then click the reset key.
Done.

When you want to start up a game, open Afterburner, click the numbered profile it was saved at, and click apply. When you're done, click the reset button and close Afterburner.

I would play around with manual votlage but the OC scanner seemed to have done a good job Thanks! I will look more into this : D

 

[Phaaze88]

I would play around with manual votlage but the OC scanner seemed to have done a good job Thanks! I will look more into this : D

Scanner appears to be broken for me, and the few times it did work, it always gave 90% confidence... now that I undervolt, I don't use it anymore.

Here's a quick rundown if you want to give it a try:
Run Msi Afterburner. Unlink the Power and Temperature Limits - there's a paperclip like icon next to it depending on the skin being used.
Max out only the power limit, and click apply.
Play your games with Afterburner's own hardware monitor running. Make sure both Core Clock and Gpu Voltage are visible.
[If Gpu Voltage is not visible, go to Settings > General tab, check Unlock voltage monitoring, then go to the Monitoring tab, find Gpu Voltage, and check it.]
After a few minutes, check Afterburner's hardware monitor for the MAX Core Clock and Gpu Voltage. Memorize, or write 'em down.
Please close the game, and open Afterburner's Curve Editor. Take the max gpu voltage you recorded, and subtract 0.05v from it. Find the voltage point in the Curve Editor that matches it, or is the closest match, and click on it.
Then use the up arrow key and raise the frequency back up to the max core clock the gpu touched. Lock it with the L key, and click Apply again.
Save the settings in one of the numbered profiles and lock them. Then click the reset key.
Done.

When you want to start up a game, open Afterburner, click the numbered profile it was saved at, and click apply. When you're done, click the Reset button and close Afterburner.


Memory clock, I don't touch anymore. It's not as forgiving as fiddling with gpu core clock.
Once a running Vram OC starts to show signs of instability, things start to go downhill from there.