Am I missing something? High GPU temps (60c load - EKWB 3080ti FE)

[toontoonizer]

Hello there, long time forum lurker - i finally went and bought the components for a custom loop. I am familiar with watercooling having used AIOs for a while and moving to the expandable Alphacool Eisbaer Extreme system to add the Alphacool Nexxos GPU block and have a semi custom loop running for a while.

Putting together the loop was ok - Removed all the old paste from CPU/GPU, flushed the new rads and washed out new components with distilled water. Ran leak test for 3 hours with no problems and powered up everything which ran ok. Loop goes as follows:

Pump >> GPU >> CPU >> 280x45 Rad >> 280x30 Rad >> Pump

I have a Founders Edition 3080ti and got the Quantum Vector FE block on. From initial testing of the first boot after setting this all up i was getting very high temps - 74 under load (390ish watts). I must have done something wrong so I removed the block and i didn't use enough paste + not enough mounting pressure. Cleaned all the gunk off with qtips and isopropyl alcohol and reapplied the paste and this time ensured the block was mounted properly.

Running some testing again and this time was much better - with the GPU maxing out at 60c. From the reviews I saw of this GPU block, this seemed too high - reviewers would max out at 50-55c.

Thinking i may have applied to much paste again, i removed the blocked, cleaned and reapplied the paste - similar story, though i feel it takes slightly longer to read 60c.

Am i missing something here? Why am i not able to achieve better temps - a max of 60c with fans at around 1000rpm is not bad at all, i dont mind it - but just wondering if i am doing something wrong here? This is the temp with no panels on - the case is completely open. Once all the panels are on this temp will rise by ~5c i imagine (this case has terrible airflow).

Any insights would be much appreciated. If i've missed a bit of information here please let me know.

These are the components in the loop - i used the EKWB Cryochill coolant. The reservoir is 90% full.

• EK Water Blocks EK-Quantum Vector FE RTX 3080 D-RGB Water Block - Black Special Edition
• Alphacool NexXxoS ST30 Full Copper 280mm Dual Fan Water Cooling Radiator V2 - Black
• EK Water Blocks EK-STC Classic 10/13 Soft Tubing Compression Fitting - Nickel
• Alphacool NexXxoS XT45 Full Copper 280mm Dual Fan Water Cooling Radiator V2 - Black
• EK Water Blocks EK-Quantum Kinetic TBE 200 D5 PWM D-RGB Pump / Reservoir Combo - Plexi
• EK Water Blocks EK-Velocity Digital Addressable RGB CPU Water Block - Nickel + Plexi
• EK Water Blocks EK-Duraclear Tubing 13/10 - Clear 3M
• Phanteks Pump Mounting Kit
• Alphacool Eiszapfen 13/10mm Threaded Rotatable 90 Degree G1 / 4 Fitting - Black
• Thermal Grizzly Conductonaut Liquid Metal Thermal Paste - 1g
• Thermal Grizzly Kryonaut High Performance Thermal Paste - 1 Gramm

PC Components:
- Intel 7820x (Delidded - conductonaut used between IHS and Copper EHS)
- 32gb RAM
- Nvidia 3080Ti FE
- Case: Phanteks Evolv X

 

[toontoonizer]

Depending on your fan RPM that mght just be what you're going to get with the amount of rad in your loop, especialy if your CPU is OC'd (??).

When you say 60, is that core or hotspot? As mentioned, 60 is pretty reasonable all in all so I wouldn't be too disappointed with it. If you're after better temps and you haven't already got them maxed out, you could test it by just ramping up the fans and see if it makes a difference.

that is for the core - the hotspot tops out at 70 when the core is 61. The CPU is overclocked - 4.5 on all cores at 1.15v - the Skylake X do run quite hot i know that.

Have you ensured all the air is out the system? I found the EKWB had a large air bubble on my GPU. Set pumps to 100% and rotate the case.
Also, are you undervolting or at stock? I have a 3090 and my temps on the core are about 55°C, but I am undervolting to 0.925V @1930 Mhz.

Yeah all the air bubble are out. I am undervolting but running 0.918v @ 2000Mhz - i could run it a little slower for sure.

Thanks everyone - just seems that this is my lot. Wanted to make sure i wasn't missing something obvious

 

[toontoonizer]

So a review was getting lower with that block. With the exact same amount of radiator and air flow? 60C with fans at 100rpm is impressive I would say. Why are the fans at 100rpm or do you mean 1000?

hah - yes i meant 1000rpm! This is a good point - I found another review which is more in line with what I am getting (slightly higher as they only used one rad to cool a CPU + GPU) so suffice to say 60c isn't too bad it seems

 

[toontoonizer]

Mine has peaked tonight at 58°C playing RDR2 at 1440p, everything maxxed out. Fans as yours, 1000 RPM

Hmm, that's interesting. Makes mine seem reasonable. I undervolted to 880mV @ 1930mhz and it tops out at around 54/55c

Problem is your case is an oven. Amazing build quality but awful airflow.

EDIT: reread the OP, something else is amiss here...

What fans are you using?

I was going to say about the fans

Also how is the Fans Intake and outtake?

So mine is Pump > GPU > CPU > RADS (FANS OUT) > RADS (FANS IN) > PUMP. I also have 2 extra Fans blowing air out. This is on a 5700xt which are known to be hot.

Negative air pressure will cause higher temps, it's about trying to get your case and temps to reach equilibrium.

I am in process of changing my exhaust fans to Artic P12 PWM - Pressure optimised fans. Best bang for buck for thermal resistance. you think your temps are hot on the junction I had 106c so it's coming up for the 6 month maintenance including changing thermal paste and pads, coolant, and cleaning rads. 5 hours work.

Using the same fan set up:
Pump > GPU > CPU > RADS (2x140mm FANS OUT) > RADS (2x140mm FANS IN) > PUMP.

All 140mm fans - plus one 140mm fan as another exhaust on the back.

The only thing I can think of is the fact I have 2x 280mm rads. I could have fit 2x360mm rads, or 1x280 and 1x420mm.

Not something in gonna do anytime soon

I do. Personally I set my fans up such that in a warm room, my GPU tops out at 50C. At the moment, 3080 + 5900x in a Corsair 5000X, dual thin 360mms and the fans spin at about 800rpm and the 3080 bobs between 45C to 50C.

The 3080 has a much lower TDP than a 3080ti - which can sustain just over 350w (400w if overclocked) - it makes sense it would run much cooler

 

[toontoonizer]

320W for my model, 380W when overclocked. 40W does not make up for a 10C difference.

What are your fans?

They are all Noctua NF-A14s

 

[toontoonizer]

Just out of curiosity when was the last time you changed the thermal pads? so not the paste, i have 4 thermal pads 1 on the VRAM then 3 on the chips.

I just replaced mine to 1.5mm also done the thermal paste on both the CPU and GPU, also flushed out the old Coolant

Temps before 80- 106c under load between clocks of 1850 and 2000mhz bearing in mind it's a 5700 bios flashed to 5700xt for the extra performance.

NOW idle 36c (so this is the equilibium temp) on load 1750 - 2000mhz 50'c that's a massive change in temperature almost 50% change. the only difference is that i changed the top exhausts to Antic p12 which are pressure optimised. i can only think the fans that i had before where Airlow instead of pressure.

I can only guess that it's probably the case airflow if you have changed the thermal paste, and in the right setup you could get 50'c on load however 60c on a GPU is pretty good tbh.

I put this loop in 5 days ago - cleaned and reapplied all thermal paste. I recleaned and reapplied paste to the GPU block twice after that. New coolant, new components (flushed with distilled water). My idle temps are ok: 29c on GPU and 35c on CPU. The P14s create abit more more pressure on paper! thats interesting.

I have a 3080ti with the EK block. I've just rebuilt into a new case with a new loop, and thought while I was at it I'd repaste the gpu with liquid metal and 'upgrade' the thermal pads.

My last case was the highly unoriginal 011D XL with 3 360 Rads, and I had a fairly consistent max gpu temp of 50 degrees. That was a water to core delta of about 15 degrees. It all worked well enough but it was noisier than I was happy with and I like to tinker.

I've jumped on the Mora bandwagon and I've not used it much yet but so far it looks like my water temperatures have gone down a lot (along with a lot of the annoying noise), but my GPU temps have gone up to 60 degrees which is a 40 degree core to water delta.

After a bit of trial and error mounting and remounting the block I figured out it's because the 1mm thermal pads I've used are less squishy than the ones EK supply. They don't compress quite enough so the block isn't making good contact with core. It's close, but it's not right!

My advice would be to get a water temperature sensor if you don't already have one to see how your block is performing from a core to water temperature perspective, as that should help narrow down whether it's the block or the loop as a whole causing the high(ish) temperature. If the water temperaure is 40-45 degrees you're in the right ball park so more rad / fans should help. If the core to water delta is high, check the pads / core contact.

I should have planned for a water temp sensor, but now it would be a hassle to install it - unless it went in the radiator drain port which is fairly easily accessible but not sure if thats a good place to put it tbh.


I think this is just how its gonna be for me. I don't think i have set anything up wrong - 2x280mm rads provide more than enough cooling power. 60c is fine - just wanted to see if i had missed something obvious

 

[toontoonizer]

Just for info, I'm gonna try remount the block again. Everything looks fine from taking it apart - paste covered the whole die on both sides.

Update - made no difference

 

[toontoonizer]

The two pads overlapping maybe causing the GPU core not to be making contact with the waterblock

I reapplied the pads again too (with new pads) just in case one of them was causing obstruction. No difference, in fact, it might be a little worse now. Running a few degrees hotter..oh well

 

[toontoonizer]

Something else must be wrong then if what he describes others getting much lower. Correct flow direction?

I'm looking through his specs and his radiators are 0.7mm thicker on one and 0.1mm more on the other compared to mine. Granted that's nothing.

My waterloop runs about 34C on gaming for several hours, temperature probe coming out from the pump, two 360mm radiators with 3 fans running at 1000rpm and the top 3 at 1300 - 1500rpm under load. The 3080Ti runs around 44C on the core, memory 50C and hot spot 51C with a pump speed of 2500rpm which is just slightly more than 50% of its actual speed. This all being in a room 21 - 23C with the loop idling at 25 - 27C.

He remounted the block and got slightly worse temperatures, so it cannot be the block that's going to knock off 10C. As you can see the core contact it had.

How the hell do you get that? Liquid metal?

I've checked the inlet and outlet ports, they are all correctly connected so the flow direction is correct. Plus when I filled it up I could see the flow was circulating correctly.

Not really sure what's left apart from bigger radiators, or my CPU is dumping a huge amount of heat into the loop. Or my radiators are terrible? Nothing came out when I flushed them before putting them in

EDIT:
I did notice that the temperature jumps dramatically under load and then gradually increases - pretty sure this is normal?

 

[toontoonizer]

It's just ordinary Kryonaut. Perhaps it is the CPU. Try running GPU-Z with the Render Test to see how quick it increases the temperature, then let it cool and do the same for the CPU with Cinebench to see how quickly it spikes the temperatures. Perhaps the radiators aren't big enough for that CPU and GPU?

With Cinebench all cores 5Ghz on the i9 10900K it reaches about 70 - 75C. Gaming dependant on what it is, reaches between 55 to 65C.

The render test does nothing to the GPU - no change in temps.

When running Cinebench, the CPU does not go over 68c, in fact it stays around 65/66 mostly - when gaming it is under or at 50c. this is with a all core 4.5Ghz OC @ 1.15v

I'm baffed - does this point to the GPU block if the CPU seems to coping very well

 

[toontoonizer]

How does the jetplate look on the GPU block? It's placed where it's supposed to be? I saw somebody had one with it moved when it came from EK. No idea how it could be a faulty block though.

Looks fine? Right way round...slightly to the right but otherwise fine

However, I have noticed that the coolant runs between the block and the plexi glass essentially running over the channels rather than through them.lovely quality control from EK

 

[toontoonizer]

Does that GPU core cover in the above pic come out? I have a CPU piece that looks like a razor blade, is it fitted the correct way?
There seems to be too much thermal paste, try less, the paste is there to fill in the micro-gaps but still maintain metal to metal contact. I doubt much metal is contacting with that amount.
How are you screwing the plate on? Start with putting in the screws everywhere by a few hand turns, this helps with alignment. Then go around the four that surround the GPU core, one by one tightening a little more each time as you go around. Then the same method with the other screws. Are the screws tight enough?
(Be careful not to over tighten of course!) What stops you over tightening them, springs?

I don't think there is such a thing as too much thermal paste. Multiple real life tests have demonstrated this, with most concurring that more is safer for a GPU which is direct die and so complete coverage is important.

The jet blade is fitted the correct way but it's slightly off centre.

Yes, screwing in one turn at a time in a diagonal pattern. No springs, just a washer between screw and PCB. The screws are as tight as can be before they start stripping.

 

[toontoonizer]

I reckon you are chasing a ghost, in this triple test, https://youtu.be/ZjakYMT76jg?t=1751 the ek is doing 45C core in an 18C environment with a 3080, this is in a loop by itself with a chunky 360 and an open test bench, a 3080Ti will put out more power add in your CPU in the loop and a closed case which is probably pumping in heat from one rad to the next as well as the heat that radiates from the rear of a card and that probably puts you where you are.

VF numbers seem exceptional perhaps his environment is a factor.

Yeah perhaps indeed. Now that Ive noticed my misaligned jet plate, I'm gonna try to align it up and then call it. No more time wasted on this. My previous watercooling loop (which was an eisbaer extreme + alphacool eisblock aurora + 2x280 rads) ran at 65 on the same GPU - I think your view has merit.

That or I'm fundamentally incompetent at mounting GPU blocks.

 

[toontoonizer]

UPDATE:

OK SO - I drained the system and took apart the GPU block. I found something interesting - the jet plane was bent (significantly - around 40 degrees - obviously it wasn't like this in the assembled block but when it came out it almost flew out as i took the casing off), and not flat. I had to flatten it out.

I put it all back together and ensured that the block screws were tightened well to stop water passing over the channels and that the jet plate was properly in position. Tested it for leaks and then put it all back together.

I also took the opportunity to put the LGA2066 specific jet plate for my CPU block

Also attached 2 water sensors - one in the reservoir and one in the rad after the CPU + GPU. However, i only have one sensor connector on my motherboard so using the reservoir sensor - refilled the loop - leak test and air bubble removal.

Fired it up and lo and behold - the GPU hovered around 43 during destiny 2 and spiked to 47/48 whenever you die - for some reason this particular bit is particularly stressful. Water temps rise to 30c during gaming - idle at 27c

Ran the 3D Mark timespy extreme stress test - temps maxed out at 50 after 10 minutes with a water temp of 32.

This is much more in line with what I expected. Problem seems to be the misaligned jet plate and water not running through the channels. I shouldn't have to do this on a £200+ GPU block tbh but here we are. At least it is resolved