RTX 2080 Super, Thermally Unstable?

[LazyDev]

Hello,

 

I've been pushing my 2080s on the memory clock. I've noticed that I can get high clock speeds on the memory (Can push towards +1850), but after a while, the card becomes unstable. I do believe that this is a thermal issue, as the card does not instant crash. As seen in the log below,

 

 

 

Igor's labs did a review of the Zotac 2080 TI variant, and did find issues with the thermals, https://www.igorslab.de/en/pad-mod-2-0-zotac-rtx-2080-ti-amp-extreme-with-few-handles-from-the-possible-danger-zone-liberated-igorslab/

 

Watercooling

I have the Zotac Gaming RTX 2080 Super Triple Fan (ZT-T20820H-10P), and was considering watercooling the card, but I cannot find a block for it. According to watercool.de, my 2080s uses a reference 2080 for the PCB? (http://gpu.watercool.de/WATERCOOL_HEATKILLER_GPU_Compatibility.pdf)

 

I'm going to replace the thermal pads and paste sometime this week, but was looking for other ideas?

 

Thanks.

[Applefreak]

7 minutes ago, LazyDev said:

Hello,

 

I've been pushing my 2080s on the memory clock. I've noticed that I can get high clock speeds on the memory (Can push towards +1850), but after a while, the card becomes unstable. I do believe that this is a thermal issue, as the card does not instant crash. As seen in the log below,

 

 

 

Igor's labs did a review of the Zotac 2080 TI variant, and did find issues with the thermals, https://www.igorslab.de/en/pad-mod-2-0-zotac-rtx-2080-ti-amp-extreme-with-few-handles-from-the-possible-danger-zone-liberated-igorslab/

 

Watercooling

I have the Zotac Gaming RTX 2080 Super Triple Fan (ZT-T20820H-10P), and was considering watercooling the card, but I cannot find a block for it. According to watercool.de, my 2080s uses a reference 2080 for the PCB? (http://gpu.watercool.de/WATERCOOL_HEATKILLER_GPU_Compatibility.pdf)

 

I'm going to replace the thermal pads and paste sometime this week, but was looking for other ideas?

 

Thanks.

Run the fans at 100 % and or play around with the clock and voltage to make it stable.

[tikker]

4 hours ago, LazyDev said:

but was looking for other ideas?

Don't push it as hard? +1850 sounds rather high. Are you actually seeing an effective increase in hashrate? Aside from potentially burning out your card, if I remember right the memory on 1000 series and newer is error correcting, so pushing it too high won't necessarily crash, but cause it to need to recover from more and more errors making it effectively slower. The memory error also makes me thing of an unstable overclock.

[LazyDev]

21 minutes ago, tikker said:

Don't push it as hard? +1850 sounds rather high. Are you actually seeing an effective increase in hashrate? Aside from potentially burning out your card, if I remember right the memory on 1000 series and newer is error correcting, so pushing it too high won't necessarily crash, but cause it to need to recover from more and more errors making it effectively slower. The memory error also makes me thing of an unstable overclock.

Clocking the memory higher and higher is pushing the hashing rates. At +1900, the card was pushing 47 MH/s, without instantly crashing. I remember it going for a good 20 minutes before it crashed, then it had to be clocked to around 1700 for it to become stable again.

[tikker]

2 hours ago, LazyDev said:

Clocking the memory higher and higher is pushing the hashing rates. At +1900, the card was pushing 47 MH/s, without instantly crashing. I remember it going for a good 20 minutes before it crashed, then it had to be clocked to around 1700 for it to become stable again.

Extra or better pads for better thermals won't hurt of course, but you're probably just pushing too hard. 20 mins would be stable enough for a benchmark, but for mining you'd like that to be infinity of course.