3080 FE temps - is this normal?

Doppleganger · 3 Dec 2020 at 16:41

danswan said:
Is freezing in here also. If it wasn't a bit of a faff, I'd definitely take a look at the contact the coolers's getting. It's the speed it goes to the peak temp, and also how it just steadily creeps up to 65 when I'm doing nothing that doesn't look right to me.

Doesn't seem right. Mine is sitting at 27 right now and unless I launch a demanding game/app it doesn't move. The fan isn't even on either.

Hakanese · 3 Dec 2020 at 16:50

Could you maybe get a pic of the inside of your case, how much cooler does it run with the side off?
Maybe the airflow into your case isn't enough to shift the heat from the gpu, do you have it vertical or horizontally mounted?

If you have your rad fans at a low rpm then that might explain why your gpu is heating up so fast

Aretak · 3 Dec 2020 at 16:55

77 is the most I've seen on mine after leaving Time Spy looping for an hour. That's in an old Define R4 case, which doesn't have the best airflow. At idle it sits in the low 30s. The fan takes ages to turn off after gaming, but that seems to be normal behaviour as TPU noted it in their review.

TNA · 3 Dec 2020 at 23:16

Undervolt it. Temps will stay in the low 60’s, fans will not go loud and you will maintain stock performance.

danswan · 4 Dec 2020 at 08:25

Had a check with/without side on case and no difference. So I just went for it and re-applied paste with thermal grizzly. There's maybe a small difference but not much.

Will undervolt tonight, sure that will help, hopefully without noticeable performance degradation

danswan · 5 Dec 2020 at 18:01

Update - yep, massive difference undervolting this thing. Still gets into the low 70s when gaming for a while but a run of Timespy there and the max it hit was 64°, whilst recording a record score.

Much more comfortable now, cheers for the advice

danswan · 16 Dec 2020 at 11:15

OK this was all ticking along nicely, max temp was around 75º but yesterday I noticed temps were going through the roof again. The only change I can think was updating Nvidia Drivers to 460.89.

So thinking it might be some sort of MSI afterburner issue I tried resetting defaults and re-doing my undervolt - no change.
I then tried installing the latest afterburner beta - 4.6.3Beta2 and again resetting and re-doing the undervolt settings - no change
So I rolled back the Nvidia drivers (uninstalled the old using the latest DDU) to the last set which were working fine before (457.51)- no change.

Here's a couple of screens of my MSI afterburner & the voltage curve:

Currently I'm maxing out at 86-87º, which is actually worse than it was originally, so something's clearly affecting things. I do have Corsair iCue and NZXT CAM installed, so have stopped both of these, to no improvement. As I'm fairly new to this, I'm hoping there's something obvious I'm missing here - any ideas?

nikoliath · 16 Dec 2020 at 11:23

danswan said:
I know this card runs hot, but this seems to be hotter than I'd expect.

Just idling on the desktop with 1 browser window open I can watch the temperature gradually creep up from 30° to 65° where it eventually stabilises.

In games it very quickly reaches the thermal limit which appears to be 83°. I've had a couple of incidents of RDR2 freezing and having to crash out, not sure if that's related or not though, otherwise performance is good and all seems stable.

Case is fairly well ventilated, but could be better I guess. 280mm radiator on the front drawing air in, and 140mm at the top/120mm at the rear. Last card in this case was a 5700XT which ran much cooler.

What sort of temperatures are other 3080 FE users seeing? I'm tempted to re-apply thermal paste, but seems a bit drastic for a 5 day old card

Seems a bit high based on the temps I've seen from most benchmarks however, my Zotac AMP HOLO runs at 83'c when under load and idles at about 60'c AFTER being loaded and 40'c idling from cold boot.

danswan · 16 Dec 2020 at 11:47

~~Think I've sorted it~~ - forgot I had to check 'Unlock Voltage Control' in the afterburner settings.

Not sure why it stopped working, ~~but back to a decent temperature now~~

Nope still the same as before, am back on latest drivers (460.89) and latest beta afterburner I can find.

Help!

danswan · 16 Dec 2020 at 14:52

Some further investigation - GPU-Z shows me that the undervolt is being respected, but it's having seemingly no effect on temperatures. Now I'm really confused. The heating isn't on and it's at least as cold in here is it was when it wouldn't exceed 73º. Yet it's still peaking at around 87º.

stooeh · 16 Dec 2020 at 17:58

Definitely seems too hot for 825mV, is that even with the case side off as a friend had some temp problems initially with a case that had noise filters? You mentioned repasting maybe take it apart again and check the paste was spread appropriately and cooler is making proper contact.

Disco_P · 16 Dec 2020 at 20:43

What was the power draw with the undervolt and what were you playing?

Found that at 0.825V, Cyberpunk will draw 260W but HZD is about 220W.

Dontrocktheboat · 16 Dec 2020 at 21:55

that waaaay too hot. Something aint right with that FE. Is your rad still blowing hot air on the card?
3090FE undervolted similar curve to you doesnt even hit 50C at 45% fans.

g67575 · 16 Dec 2020 at 22:59

Maybe cap the FPS at 60 or enable Vsync (60 herts). If it's getting close to the thermal limit at 60 FPS, there could be a hardware /cooling problem with the graphics card.

Play around with the fans, could be that they aren't running at high enough speed.

Would be good to know the RPM of your fans under load. What is the maximum RPM for the RTX 3080 FE fans?

Personally, wouldn't muck around with thermal paste - Not sure if this would effect the warranty or not.

danswan · 17 Dec 2020 at 09:13

Thanks for the replies. I'm kind of sorry I'm seemingly not missing anything obvious, so no quick fix. In other circumstances I'd simply return it, but who knows how long it might be before a replacement would be forthcoming? I've no spare card and even in it's current state it's light years ahead of my old 5700XT

Re thermal paste, I'm happy that and/or contact with the die isn't the issue - don't forget that it ran happily for a week or so and never went past 75º. Also the screws on the thing are terrible quality - almost all of them were starting to chew up when I took it apart last time, so I'd be pushing my luck trying that again.

The fact that it's just all of a sudden started doing this does point to a software thing of some kind for me, but I'm damned if I can guess where I should be looking next...

labbby · 17 Dec 2020 at 09:27

rarely see mine go above 60c while gaming but i do have a huge case with overkill air flow, that's with about 60% on the fans

nikoliath · 17 Dec 2020 at 09:50

tbh I wouldn't lose any sleep over it

danswan · 17 Dec 2020 at 10:16

nikoliath said:
tbh I wouldn't lose any sleep over it

Not losing sleep, I'm losing performance though as it's hitting thermal limits

nikoliath · 17 Dec 2020 at 11:02

danswan said:
Not losing sleep, I'm losing performance though as it's hitting thermal limits

will reducing the thermal cap improve performance significantly or just allow you to reach the next cap?

vRel = Reliability. Indicating performance is limited by voltage reliability.
VOp = Operating. Indicating performance is limited by max operating voltage(Hardware Limit).
Pwr = Power. Indicating performance is limited by total power limit.
Thrm = Thermal. Indicating performance is limited by temperature limit.
Util = Utilization. Indicating performance is limited by GPU utilization.

danswan · 17 Dec 2020 at 21:56

nikoliath said:
will reducing the thermal cap improve performance significantly or just allow you to reach the next cap?

vRel = Reliability. Indicating performance is limited by voltage reliability.
VOp = Operating. Indicating performance is limited by max operating voltage(Hardware Limit).
Pwr = Power. Indicating performance is limited by total power limit.
Thrm = Thermal. Indicating performance is limited by temperature limit.
Util = Utilization. Indicating performance is limited by GPU utilization.

In this case, it will improve performance significantly... turns out I'd somehow missed the fact that my AIO tubes had dropped onto the card and the fan towards the front wasn't spinning at all! Unsurprisingly now I've corrected that, things are much better. Caught out a bit by this special FE design, traditionally there's no fan facing the top of the case so it's fine for the tubes to rest on the back of the card.

Cheers for all the advice, thankfully nothing wrong with the card, just it's operator