Not sure who defined "overheating" for you but FYI 65-75C are excellent temps, nowhere near overheating.
That would be above 85C when the card starts throttling down the core clock.
Anyways, GLWS
Graphic Cards RTX 3080 temps
- Thread starter enthusiast29
- Start date
That would be above 85C when the card starts throttling down the core clock.
Anyways, GLWS
I noticed if I don't undervolt, the GPU goes above 270w , and starts thermal throttling.
I started cyberpunk 2077 without setting the undervolt and after 10 minutes I noticed frame drops . by the I hit pause and opened msi afterburner it has already went under 70C.
I think it was running hotter earlier at 270w, with my Undervolt I usually get 230w occasionally reaching 70C.
Then it's the overheating of vram do check those temps as gddr6x is prone to that very easily and causes throttling. The moment it will cross 105 it will cause fps drops and throttling even though vram temps threshold is more.
Last edited:
I have an RTX 3080 Founders edition. The memory chips on those cards can well exceed 95+C under gaming load and are safely rated up to 110C before throttling kicks in. Check your memory junction temps. GPU temps are useless.
You will face power limit-induced throttling before temperature (if your card is in good condition).
I have my power limit set to +115, core overlock of +140, and memory +750. In most (intense) games, the power draw is around 350W+. Temps normally look like this:
GPU - 60-62C
Hotspot - 72-74C
Memory Junction - 92-93c
The above is with the fan on auto, and inside a LianLi Lancool 216
Once the GPU goes over 365W, my clocks start throttling a bit. According to GPUZ, it's due to power limitations, and not temperature.
You will face power limit-induced throttling before temperature (if your card is in good condition).
I have my power limit set to +115, core overlock of +140, and memory +750. In most (intense) games, the power draw is around 350W+. Temps normally look like this:
GPU - 60-62C
Hotspot - 72-74C
Memory Junction - 92-93c
The above is with the fan on auto, and inside a LianLi Lancool 216
Once the GPU goes over 365W, my clocks start throttling a bit. According to GPUZ, it's due to power limitations, and not temperature.
The stock thermal pads are garbage. You should see better memory temps if you replace them.
+750 on the memory? GDDR6X memory is error-correcting so you won't see a crash if you clock the memory too high, but performance should still degrade. That and generally low benefit from memory overclocking has led me to not touch it. Have you ran any benchmarks to confirm that +750 on the memory alone actually improves your perf versus lower/stock?
I got the card last week, put an undervolt/overclock on it of 1920 Mhz @ 0.875V and it works great while only consuming ~250-270W in cyberpunk and a maximum of ~300W in Furmark.
I know it's auto-correcting, and I won't see a crash. Beyond a certain point, performance will just start degrading.
Yes, I have settled on +750 only after running benchmarks and confirming gains (although not much). My methodology involved setting a fixed (stable) core clock first and then clocking the memory higher to see for additional gains. Beyond 750, results start diminishing.
I don't care too much about power consumption, so I've just maxed out the card.
Just out of curiosity, are the temps you reported based off a run/session in winter? Provided of course you live in a region with a defined winter.
Yes, these are winter temps. I live in Kolkata, so pleasant winter weather.
Nice. What were your benchmark results like with that overclock? I get about 11,870 on Port Royal.
I haven't run Port Royal yet. My GPU score in Timespy is 19105, and 45271 in Firestrike
That's zotac rtx 3090 trinity stock score XD. Great for rtx 3080. 92-93 is bit high for prolonged use but for gaming that's totally under control. I have added 5-10mm heatsink on the back with a small fan that keeps the temps around 82-86c during summer (78-82 in these winters) for rendering on rtx 3090.