How to improve thermals in my 570x Crystal?

Hello All,

I built a new PC a few months ago including a custom water loop. I had issues in getting the pump to work and I eventually used an EK Quantum FLT pump/res combo. My GPU is also quite long which made placement and/or size of my pump and front rad difficult.

My GPU thermals at 100% load in game have reduced from a steady 84 degrees to around 70 degrees but occasionally increasing to 75. My CPU was only at 20% because it was more graphic-intensive and sat at about 60 degrees. Idle the GPU/CPU are around 34/45 degrees. I don't think at least the loaded values are right for my custom loop and I am going to try the following, so please let me have your thoughts on these proposals -

- replacing the front 360 rad with an Alphacool 360mm 45mm depth

- replacing paste and pads with thermal grizzly/minus pads 8

- some how introduce an exhaust fan despite the limited space (?!)

- removing the 2x Corsair ML120s from the pull config on the front rad

- install thermometer in one of the G1/4 ports and link to CoPro, and set SP fans speeds to coolant temp with custom curve (which I have to calculate/set up too)

My current set up is as follows -

- Pump: EK Quantum FLT 120 D5, running ar around 2,000 rpm

- Front rad: 360 Bykski 28mm depth

- Top rad: 240 Bykski 28mm depth

- CPU block: Corsair XC7

- GPU block: Bykski

- Thermal paste on cores: Arctic MX4

- Thermal pads on GPU VRMS etc: Arctic 1.5mm

- Fans: 5x Arctic P12 PMW (push config on rads) and 2x Corsair ML120 (pull config on rad)

- Cooland: EK Cryofuel

The loop runs pump outlet --> GPU --> front rad --> top rad --> CPU --> pump inlet.

My computer spec is on my profile (I don't want to bulk this post out). I have a commander pro, node pro and DeepCool RGB converter/controller.

Thanks :)

What's the airflow direction? (both radiators are exhaust, both intake, or front in, top out, etc.).

I don't like the end temps on the GPU. That seems to be far too high for full water and that is more likely to be contact/block issue than coolant temperature. However, try playing/testing at a higher pump speed like around 3500-4000 rpm. Normally I would expect little difference, but if the loop is very restrictive or you have an unknown blockage this may help indicate that. It is possible to have unusually high component temps if the liquid slows down too much in the block.

Definitely get a coolant temp probe somewhere in the loop. You have everything else you need to make it work as a control or even just for information. Easy £5 upgrade.

yea 2000 RPM starts to be a bit slow. I tried a little on my end and i lose like 2° compared to just bumping to 3000 rpm.

That's 2 degrees accounted for. now there's 40 more to explain :p

Even if the radiators you have are not top of the line, you shouldn't see such temperatures, even in a case as restrictive with airflow as the 570x.

When you mounted teh GPU block, did you just put a small pea of paste in the middle of teh die or spread it manually?

With direct die cooling it's usually better to spread to make sure all of the silicon has paste. every little bit of bare silicon will heat up otherwise.

What's the airflow direction? (both radiators are exhaust, both intake, or front in, top out, etc.).

 

I don't like the end temps on the GPU. That seems to be far too high for full water and that is more likely to be contact/block issue than coolant temperature. However, try playing/testing at a higher pump speed like around 3500-4000 rpm. Normally I would expect little difference, but if the loop is very restrictive or you have an unknown blockage this may help indicate that. It is possible to have unusually high component temps if the liquid slows down too much in the block.

 

Definitely get a coolant temp probe somewhere in the loop. You have everything else you need to make it work as a control or even just for information. Easy £5 upgrade.

yea 2000 RPM starts to be a bit slow. I tried a little on my end and i lose like 2° compared to just bumping to 3000 rpm.

 

That's 2 degrees accounted for. now there's 40 more to explain :p

Even if the radiators you have are not top of the line, you shouldn't see such temperatures, even in a case as restrictive with airflow as the 570x.

 

When you mounted teh GPU block, did you just put a small pea of paste in the middle of teh die or spread it manually?

With direct die cooling it's usually better to spread to make sure all of the silicon has paste. every little bit of bare silicon will heat up otherwise.

Thanks both.

The 3 fans front and 2 top are all intake, in a push config. I would have had an exhaust on the rear 120 slot but had to put my pump/reservoir there and there isn't enough room for a fan as well dye to clearance with the cpu block.

I bumped it up to 4000 rpm and got around 65 degrees on the GPU, so I don't think it's a restriction in the loop (?) That said, I wish the first connection between the pump outlet and GPU inlet was cleaner, it was too tight to go straight into the topside GPU from either the front or bottom pump outlets.

When I get the new front radiator I will also re-seat the water blocks, both cores and also the pads on the GPU. I will take a photo to show the previous spread of paste (and will manually spread it when I re-seat, as you suggest).

Will let you know how it goes.

Incidentally, I have just installed a DRGB 3-pin-to-corsair connector, I think it was £11 for 2 of them on Amazon. I used one of them to get my EK Quantum synced with the rest of my corsair RGB via the compro/node pro. It works well! :) Just need to do something similar with my 4-pin GPU block RGB or take out the existing strip and swap it

yea the 4 pin RGB is 12V analog, so it won't work on corsair.

but a bit of ARGB strip with some encouragement from a soldering iron will do the trick in 5 minutes flat.

You could set the top fans to be exhaust, there would be a lot more air flow instead of doing a heavily positive pressure setup.

Even if the top rad will have air slightly warmed by the front rad, it will help.

Still may not be enough to account for such high temps, but should be better.

Also, don't be afraid to run the fans on pull. It is veeeeeeeeery slightly more efficent (barely really) but it also makes cleaning the rads way easier since you dont have to dismantle to remove the dust (maybe on the top rad, to set it to exhaust for testing).

Regarding the pump speed, it could be that one of the waterblocks is very restrictive.

The XC7 uses microfins like AIO coolers, but the Byksky uses custom loop style thick fins for high flow pumps.

It could have an incidence, but i am really not convinced it would account for even 65°C on a 2080 with two rads.

Maybe make sure you used the correct thickness pads on the mosfets and chokes when installing the GPU block. a mistake here could prevent good pressure on the die.

I checked the CPU and GPU paste and pads to Thermal Grizzly's paste. My previous paste was Arctic MX4 and pads were Arctic 1.5mm thick.

Looking at the old paste, it looks like it congealed and stuck to the cooler plate in disassembly.

I also installed a 45mm thick Alphaool Nexxus at the front which arrived yesterday, and took the opportunity which it was drained to put that sensor into the EK reservoir as well.

You can see that in pic 4 (on Witcher 3, High settings) my temp is much reduced. I had the case glass off whilst I was testing the water loop but it is back on now and I will test it with the glass back on tonight. Will also test other games to see what the temp is like.

LeDoyen - I can't do the pull config at the front because unfortunately my drain port is at the back of the front rad, so if I take the fans off of that side it will make the rad flush against the chassis and the port inaccessible. And you might have been right about the pads being too thick and preventing good contact with the die! Will consider the GPU plate and whether it restricts water flow, but the Bykski is the only one I have found for the Zotac 2020 Super Amp : /

Thanks for your help :)

all that mass of paste on the die and the block look like they were almost touching but not quite.

Did the block come with pads? or with a manual telling what thickness is required?

half a mm can ruin your day, as you saw.

all that mass of paste on the die and the block look like they were almost touching but not quite.

Did the block come with pads? or with a manual telling what thickness is required?

 

half a mm can ruin your day, as you saw.

Yeah it looks weird how it is all congealed too. Nah no information in the manual about the height of the pads. Pads were supplied but they were in an unbranded bag, stacked on top of one another. Thought it best to go with Arctic or another known brand. I think you're right about 1mm vs 1.5mm

I bought a 5v DRGB strip for the GPU plate and I look forward to finishing my final mod (well, until the next mod I think of! ;))