Question Max safe long term voltage for a 2070 super ?


Jul 31, 2018
I have seen a lot of posts saying pascal was 1.094v but not much on Turing. I also saw somebody say on here nvida is on the record saying anything under 1.064v I haven’t been able to confirm that. I also saw Nvidia say on a game or nexus interview turning the voltage slider up to 100% could reduce lifespan to one year but I hope that’s just them covering their butt. Some of you may recognize me from my other posts but a quick recap: I have an aorus 2070 super. I didn’t have the luck I had hoped manually overclocking with offsets, I can get +25 core at 2010mhz and +800memory before I start to get errors on OCCT 3D test. I found better results using the voltage frequency curve. My card runs mostly 1.050 stock and there I can turn it to 2070mhz and +800memory and not really worry sense that is what it does normally. However I can get up to 2100mhz at about 1.068v , and Maybe more but I stop there until I found out what safe long-term voltage is. I understand this varies card to card and model to model but I was looking for a general safe consensus. I also don’t get errors in OCCT when I use the curve this way , not really sure why but I think it has to do with I hit power limits instantly in that test which drops the voltage no matter what down to around 900 making higher clocks unstable. I guess when I only changed one point on the voltage frequency curve if I can’t do that it just goes back to what it normally does, but I’m no expert maybe you’ll know ? It may be worth mentioning to that one of the selling points if this car was “ Built for Extreme Overclocking 12+2 Power Phases” As opposed to nvidia 8+2. What is this really has any effect on lifespan at high voltage is I don’t know. Thank you for any help in advance !


Sep 28, 2018
If you have to feed it a significant amount of voltage for a small leap I personally wouldn't go for it. Like for my cpu just to get 25 more MHz I'd have to increase it's TDP by roughly 30W which isn't worth it. I'd research for your specific model that you have since some models can take more power than others. As for your other questions I'm not entirely sure somebody else would have to answer them.
Reactions: Phaaze88