Gaming Discussions

[mengelag]

So the wife recently got her rig up and running. Doing some benchmarking I'm noticing the temps at the junction point (hot spot) is going right up to 110c within a minute of running Aida 64'. Overall GPU temps were under 80c. I also did a custom fan curve that put the fans at 95% when temps hit 90C. GPU hot spot was still hitting 110c and staying there. I couldn't tell if it was throttling or not and doesn't seem to be when in-game. Aida is pretty tough on the card. I just don't like it getting that hot and I can't figure out why it is. I'll include a pic. Airflow is a bottom-up design, Case is the tower 500. Plenty of air intake and exhaust. The ambient room temp is 20c. Fan set up is two ML fans pulling in air from the front, 2 ml fans under the GPU pushing the air up. two ml fans in the back compartment exhausting air out the back and the 280mm radiator at the top also exhaust an exhaust. There isn't really anywhere I can mount more fans. What do you guys think? Airflow issue or card issue? I re-pasted it yesterday with Kryonaught and medium load temps were a few degrees lower as well as the idle temp a few degrees lower. Thank you all and cheers!

[Justifier]

Seems to be unfortunately common when mounted like that.

I'm running a Segotep Phoenix (modded so it doesn't have crap intake) with a similar layout, I think these GPUs are designed to have way more airflow over the "top" of the GPU (the backplate) coming from the front to dissipate heat than setups like this can provide. Since that isn't happening it's causing higher hotspots than normal.

1. Should likely do this anyways: undervolt it.
2. Best: (voids warranty) replace the thermal pads for high end ones (after EXTENSIVE research into literally every aspect of how because you can brick that thing so fast if you're not careful) NM see you already did this
3. Rigged solution: find the hotspots and glue/mount somehow some M.2 heatsinks (with thermal pads ofc) to the backplate and get some air flowing over it

[mengelag]

Cheer's @Justifier . I was wondering that as well. Seems to be a trade-off. GPU sag vs lower cooling. I have it undervolted on the adrenalin app. Don't know how to undervolt it more though. If I could I would. 

Thanks again

[Justifier]

The reason to favor setups like this isn't about GPU sag.

It's PCBs on these gargantuan cards cracking and the warranty not being honored by many AIBs, and it's not limited to AMD by any means:

https://www.youtube.com/watch?v=qWwVpnYlySw

whole video

Gigabyte RTX 3080 graphics card Repair [FAILED DIY attemp] #nvidia #gpurepair - YouTube

(seen @~4:15)

Companies like Powercolor even ship anti-sag mounts with them, but many popular cases have holes right where you'd need to mount them at rendering them useless

[Justifier]

Can use MSI afterburner if you're uncomfortable/unfamiliar with undervolting normally

Can use it to power limit the whole card until the temps get to more acceptable levels. I don't recommend using Adrenaline for it. it's yet to properly do the job in my experience

[BigAl01]

I think undervolting is the cheapest solution, but you might be able to fabricate your own mounting and get a PCIe 4.0 extender cable so you can try a vertical orientation for the video card.  The card's exhaust has to be channeled out of the case though, so it would take some thought.  It appears you already removed the video card's heatsink and reapplied thermal paste. 

If the card isn't reaching more than 80 C under normal loads or gaming, should you be worried about it?  

[Justifier]

" If the card isn't reaching more than 80C under normal loads or gaming, should you be worried about it?  "

Yes, OP should be worried, as the hotspot is hitting +110C

If you browse about some well reputed hardware repair shops who post on YouTube, such as " KrisFix Germany " or " Northridge Fix " You'll easily find many many instances of modern GPUs warranties not being honored, even in instances where the users did not mess with the cards in any way, and the OP did technically void the warranty by repasting the GPU.

Now that's not to say that at this point the brand wouldn't cover the warranty anyways, because some companies are just awesome like that, but no guarantees.

Best case scenario letting it run with a hotspot that high could lead to early degradation, worse case complete failure of the chip

Best option is to keep searching for ways to mitigate it in advance as they're doing (and I'm doing too, same issue on my Red Devil 6900xt)

Currently, I've kept mine under 100c by power limiting and undervolting, but when you're running at 4k that quickly impacts the performance so it's not optimal

[mengelag]

@Justifier Can't do the thermal pads on the hot spots because they are on the chip itself, unfortunately. The other sensors are in an acceptable range.

[Justifier]

one last thing you could try is to "vertically" mount it, if there's enough room to do so

from what I can see you've only got ~20mm of clearance at the bottom there so that's likely not an option either unfortunately