Does anyone know how the RTX 3080 clock speeds work?

[Coldkil]

I did! From 78C, but I have to admit, I would need to remove it again and run thorough tests without and with because these gains seem unrealistically high.

Also, I wonder if one vertically aligned fan at the back would be better than the two horizontal fans I have inside.

Sketchup is horrible, so I need to find some new software to design a v2.0 shroud.

If you go with a single fan, at this point make a 30degree slope or whatever so the fan is blasting the air more directly to the gpu. I think that with a shroud like this the second fan (the one further away from the back of the pc) isn't actually doing much. Due to the small volume of the chamber and relatively limited intake space the additional pressure generated isn't significant; also since it's nearer the shroud it has even lessp space to get air from (which is mostly sucked but the forst fan anyway).

I'm thinking that the second fan may be even slightly detrimental. I would say putting the fan near the intake would definitely get air in the shroud "better" but it may be just an illusion. However this lets you print a much smaller shroud which means air has a narrower/smaller path towards the gpu, which means more speed = more airflow and less dispersions, for whatever it counts. We're not talking about a perfectly sealed environment.

[nocturnus]

If you go with a single fan, at this point make a 30degree slope or whatever so the fan is blasting the air more directly to the gpu. I think that with a shroud like this the second fan (the one further away from the back of the pc) isn't actually doing much. Due to the small volume of the chamber and relatively limited intake space the additional pressure generated isn't significant; also since it's nearer the shroud it has even lessp space to get air from (which is mostly sucked but the forst fan anyway).

I'm thinking that the second fan may be even slightly detrimental. I would say putting the fan near the intake would definitely get air in the shroud "better" but it may be just an illusion. However this lets you print a much smaller shroud which means air has a narrower/smaller path towards the gpu, which means more speed = more airflow and less dispersions, for whatever it counts. We're not talking about a perfectly sealed environment.

Definitely! This was the lazy approach, since I absolutely hate Sketchup; what a shit program man. Whenever I have time, I'll try to design a proper shroud with sloped mount. Ugh, already dreading it ^^

- - - Updated - - -

Version 2.0:





Let's print this and see.

[Coldkil]

Why the L shape at the narrower end? Cable fittings or someting else?

[nocturnus]

Why the L shape at the narrower end? Cable fittings or someting else?

Ah yeah, I should have mentioned that. That’s for the GPU anti-sag pillar. I also forgot to make a hole for the fan cable, but I can drill that in easily enough. The prints should be done in a few hours.

- - - Updated - - -

Snippety snap[/IMG]

What are you using for benchmarking?

[Ange]

What are you using for benchmarking?

The OSD? Thats just a custom RTSS overlay with only HWiNFO sensor data to quick check temps/fancurve and to see if everything is working as it should be.

For games I use CP2077 with RTX-ULTRA + psycho (global illumination) to get a good CPU/GPU heat mix, QuakeRTX for VRAM temps since its as hot as mining. For sustained cooling I just use flight simulator in 4k VR for a few hours and check temps from time to time. MSFS uses the CPU pretty hard and the GPU is basicly at 100% TDP even at 1080p.

[gaaara]

lol, they're totally fine.

Isn't 110C like the threshold for the vram temps? Its insanely high, mine peaks around 80C on Vram, 65C on GPU. This is on a 3080 TUF OC. GPU Clock sits around 1950~

I do run a pretty low fan curve, however this is on a o11xl with all fans populated.

My guess is perhaps the thermal pads for the VRAM isn't sitting right, as the GPU is in good range.

Who knows..he might be one of those guys who has a fingernail thing in the gpu...(Yea...that happened))

- - - Updated - - -

While we're on the subject of temperatures, without wanting to deviate too much, what would be a decent temperature for a GPU at idle? My 3070 (EVGA XC3 Ultra model) sits at 46-47c when idle and I was wondering if that was high, mostly as a curiosity. When gaming at full throttle it rarely ever goes over 75c (except for Cyberpunk which makes it go to 80c at times) which from what I know is fine but 46c at idle I'm not sure. Do consider I have no special cooling options beyond 2 case fans, no AC in my desk room and that it's pretty hot outside for my neck of the woods at 33c, in case that changes anything.

You probably have a zero fan curve that spins above 50C? On Zero RPM with light usage (eg desktop), that looks normal. If your worried you can just adjust the curve so it zero rpm's at maybe 40C.

[Ange]

Isn't 110C like the threshold for the vram temps?

Tj (max) is at 110°C, but throttling and extrem fan curve starts at 106°C and peaks at 110°C thats where heavy throttling starts.
T (tot) is at 120°C (dead silicon) but degradation rises till that point of course.

We want <95°C at worst case (mining, quake2rtx) with hot ambient temps. Thats pretty easy to reach with just thermal pad replacements. If done right you end up <90°C with 27-30°C+ ambient.

Stock what ever 3080 or 3090 you look at is somewhere in the 100-110°C range with the same load + ~20-22°C ambient. There are no magic thermal pad fixed variants out there, just some people not able to test propperly or using lower ambients for measuring.

[gaaara]

Tj (max) is at 110°C, but throttling and extrem fan curve starts at 106°C and peaks at 110°C thats where heavy throttling starts.
T (tot) is at 120°C (dead silicon) but degradation rises till that point of course.

We want <95°C at worst case (mining, quake2rtx) with hot ambient temps. Thats pretty easy to reach with just thermal pad replacements. If done right you end up <90°C with 27-30°C+ ambient.

Stock what ever 3080 or 3090 you look at is somewhere in the 100-110°C range with the same load + ~20-22°C ambient. There are no magic thermal pad fixed variants out there, just some people not able to test propperly or using lower ambients for measuring.

err i don't mine so no idea what all that was about, but mine in wow memory temps is close to 60-70C

[nocturnus]

I’d have tried using the shroud for anti-sag… a tab or two inside which the GPU contacts should have done the trick.

That would've been the best solution; this was quick and dirty but I'll definitely do that in the definitive version!

[nocturnus]

Looking forward to the results of your experiments.

I should have addressed all flaws in this one, but you never know. I'll start the print tomorrow and we'll see

[nocturnus]

I honestly doubt you need the pillar if you have adequate thickness. It’s like a bridge with just support being transferred to the cliff walls.

You really need the pillar, or it'll fall over. Besides that's also the GPU support. My current shroud is being kept upright by the GPU power cables running under it, but you can still see a slight sag at the front. I could of course just make a bracket that screws in the back so I won't need frontal support. But that would require more measuring.

I might do that tomorrow before I start the print; it is a more elegant solution.

[nocturnus]

Why would it fall over? If it’s properly designed all the weight is transferred to the back of your case via a rather large surface area. If it falls over the entire thing broke due to a thin wall. Hence my earlier comment. Hell, the majority of the force would be transferred to the bottom of the shroud.

I was still thinking about my current setup; two fans floating on top, hence no weight on the back. With this design the 120mm fan is in the back which will be responsible for counterweight and thus would make the pillar superfluous. I had forgotten about that change. /facepalm



- - - Updated - - -

Okay, so the shroud doesn't do much compared to just having two 120mm fans spinning below the GPU. In both scenarios I measured max temperatures of 75C with a custom fan-curve of maximum 40%.

Without either, the temps surpass 80C with the stock fan-curve, which boosts fan speed up to 70% from 60C and 100% after. No bueno.

All in all, pretty good news because I didn't really like the look of a plastic shroud in my case.

The next step is replacing thermal pads. Is there consensus about which pads are best? I know reference cards need 1mm. Which thermal paste would you advise for the CPU and what about cleaning the old thermal pads and cool paste?



Around the 10 min mark the guy starts complaining about the stickers and the plate not making contact but I wonder if it's supposed to make contact there. I can imagine that's intended for the GPU to dissipate heat. If it contacts the VRAM, the GPU temps could start rising.

[nocturnus]

Okay, so I changed the thermal pads and thermal paste and went from

100% load

~80C GPU -----> 73C GPU
~105C VRAM -----> 84C VRAM.

All the while, fans were spinning at 35% to 40%

The old VRM pads were 1mm thick and were clearly not making any contact with the headsink above. You could see this because the top section of the pads was virtually untouched. I replaced these with 1.5mm pads. I added 1mm pads to the VRAM and backplate and reapplied Kryonaut thermal paste.

My current setup is two virtually silent 120mm fans under my GPU, spinning at about 600-700 RPM and my GPU's fans at 20% to 40% under load.

It was an interesting project; even designing the useless shroud was kinda fun