[SOLVED] RTX Water cooling temps

[Qwirks]

I added my RTX 2080 into my water cooling loop. For some reason, the idle temps are in the mid 30's, but under load, it hits 85C without any overclock on the card at all. I have remounted the block, tried different thermal paste, and I even tried to use a graphite thermal pad on the core. But no luck! I haven't been able to get any good temps on the card at all. CPU temps are fine, reaching 50C under full load at 4ghz. Any ideas? I'll post a picture of the card to make sure I'm not putting the thermal pads in the wrong spot.

EKWB Vector GPU and CPU Block
EKWB d5 140 res pump combo
1 240 slim rad, one 360 slim rad from EKWB
Gigabyte RTX 2080
Ryzen 1800X
Crosshair VI

https://imgur.com/a/xO3VXdT

View: https://imgur.com/a/xO3VXdT

 

[rubix_1011]

I have a 2080 in my loop and see about the same results, although mine are mostly high 70's to low 80's C. You even have more radiator space than I do; I'm running dual EK PE240 rads.

I'm considering pulling my card and reseating my block to see if there is any difference. I've watercooled cards for many, many years and this is the first time I've seen where GPUs don't benefit greatly from custom watercooling.

 

[1LiquidPC]

Something odd is going on here. Are you both using same block? If so, possible there is a defect and block not making proper contact with GPU core. I have my 2080 Ti in my single loop with 9900k. Using HeatKiller IV block and backplate. Idle is in the mid to high 20C range and full gaming load(example BFV RTX Ultra setting) is just above 40C according to GPU-Z real time logging. However, I do have a good bit of radiator capacity. I had similar temps with previous 1080 Ti and 1080 before that. Room temp is about 21C.

 

[rubix_1011]

I'm using the EK Vector 2080 RGB block (although I paid about $50 less than this): https://www.ekwb.com/shop/ek-vector-rtx-2080-rgb-nickel-plexi

My card is EVGA XC Ultra, part 08G-P4-2182-KR: https://www.evga.com/products/product.aspx?pn=08G-P4-2182-KR

My CPU is 9700k and it runs very cool, only seeing temps on the GPU around 35-40C at idle and upper 70's-low 80's at full load.

I agree, something is amiss.

Edit: I put a post out on the EK Discord support thread, I'll see what turns up.

 

[1LiquidPC]

Thanks for the update. Find it highly unlikely you two should have temps like that on custom loops especially. Plus the remounts, ect. and your tons of watercooling experience make it being a user install issue very unlikely. Manufacture defect perhaps on thickness?

 

[rubix_1011]

EK said to pull the block, use Thermal Grizzly Kryonaut (which I happened to have).

I did, block looked to have been making good contact with Arctic MX-4, but cleaned and replaced anyway. Rebuilt the loop, filled with water, off to the races.

Idles at 23-25C, I haven't seen loads over 40C yet.

I haven't ever seen an issue where a block didn't seat well and perform splendidly on the first go-round.

First time for everything, I suppose.

Update: played World of Tanks last night for a few hours solid with all settings on the highest setting (Ultra, I think) with the ability to now use the frame sync and triple buffering - dipped down to 28 fps with this before, now runs easily at 60+ fps, so I actually think my card was throttling at 80C+ and not able to really stretch its' legs. I'm thinking my initial disappointment with this card was simply due to a bad block seat and not cooling effectively.

 

[1LiquidPC]

Wasn't expecting that from a remount, but I suppose it can happen to anyone. We all make our mistakes time to time. Using Kryonaut myself, and happy with results as well. Glad to hear the huge boost in performance. You likely are correct on the FPS tanking from heat. I believe this generation throttles at 80C. What are your boost speeds now out of curiosity? My 2080 Ti boost to 2010Mhz on its own when being fully utilized in games. This is with no adjustments for power/temp targets or OC settings in third-party software. Thanks again for the update. Hopefully, the OP tries the same with another remount attempt and gets a good result as well. Was there anything different about second remount in your case? The OP did say they tried a few times themselves.

 

[rubix_1011]

No, I've done this so many times I can almost do it in my sleep. It also means that I really only do about 5 mins of leak testing each time, but mostly because I'm comfortable with knowing what I'm doing.

One odd thing I noticed on the EK RTX card - the 4 main mounting threads that surround the GPU itself are kind of threaded inserts that actually thread into the GPU block itself. I found that when I loosened the screws, they actually unthreaded from the GPU block itself...I had to use pliers to secure the inserts, remove the threads and then re-thread them back into the block. Otherwise, this was just another ordinary full-cover GPU block install...nothing different.

 

[1LiquidPC]

I see. Wouldn't expect the insert/thread issue either for typical install. I saw an article about the GDDR6 being different thickness(Micron vs. Samsung) causing issues with block mounting for some. Symptom was poor GPU core contact and high temps.

 

[Karadjgne]

From the looks (wish op had lined up the block mount with the card, stupid thing is upside down) it looks like the double row of vertical pads has 1 row too many. The taller VRM's seem to be hitting the block first, not allowing good contact with the block on the lower mosfets. The block itself only has the one thin section at that end running the full width of the block that looks to make actual contact.

That and possibly op isn't torquing the card down right, warping the card.