2080 ti - Added Water, Lost Performance. Thoughts please?

GilesGuthrie · 29 Dec 2019 at 10:18

I have an MSI Ventus 2080 ti, which pre-overclocking on the stock air cooler was doing 120 FPS at 5120x1440 across a few games, including Horizon 4. Temp was hovering around 70 degrees.

I took off the stock cooler and replaced it with an EK-Vector RTX RE Ti RGB as part of a CPU/GPU loop.

I'm now seeing temp under load locked at 70 degrees, but it's struggling to hit 80 FPS in the same titles. I've used MSI Afterburner to set the temp limit to 88 degrees, but it's not going over 70 deg at any time. This makes me think that there is some sort of throttling at work.

Water is moving through the loop, the CPU load temp has dropped 15 degrees from the stock air cooler, so I think the loop is working OK. Loop fluid temp is steady at 10 degrees over ambient when the system is under load.

My initial feeling is that I should recheck the seating of the waterblock on the GPU. However, I'm interested in the fact that the GPU is not approaching the thermal limit set for it. Is there a setting I need to change as part of the change to water cooling? I'm wondering whether it's self-imposing some sort of limit because its cooling fans are gone?

Do you have any thoughts? Should I just get on with draining the loop and checking my block installation?

Mcnumpty2323 · 29 Dec 2019 at 11:29

Temp shouldn't be still at 70c if it was 70c before watercooling it
My gtx1070 only runs 45c to 50c under heavy gaming which is well below what it was with the huge factory installed heatsink on it

And no mine didn't impose any sort of limit from no longer having the fans on it

wookiee87 · 29 Dec 2019 at 13:14

uninstall any gpu overclocking tools and it wont hurt to re install up to date drivers first, do this and see if your fps improve, i had a 2080 ti under water using a ek vector waterblock and at stock speed the card would boost near to 2ghz and my frame rate at 1440p was over 160fps, but i locked it down to 144fps so it sync'd to my moniter. temp wise was no highter that 49 degrees under load at stock.

i installed evga precision x gpu overclocking utility and proceded to up the power target (112% on the card) and managed to hit 2090mhz on the core and pushed a extra 1400mhz on the ram, saw some huge improvments in 3dmark scores, temps did increase to 56 degrees under load.

going back to your card if the software side of it has been uninstalled and re installed and getting same results, it may be worth taking the block off and checking contact on the core, its very important to get this right as it can casue weird issues, if the tim contact is good re build and start a game and check clocks especially the core if its stuck at 1350mhz there may be a problem with the card and it will not boost properly, rare but it can happen.

LuckyBenski · 29 Dec 2019 at 13:24

Wonder if there's another bit of the card getting too hot due to poor mount such as VRMs. Does it have any other temp sensors?

b00merang69 · 29 Dec 2019 at 13:39

Load up GPU-Z or Precision or Afterburner Riva Tuner Statics Server and see what is throttling the card, must be heat somewhere not sure if VRM's give a temp readout on your card or not.

wookiee87 · 29 Dec 2019 at 13:43

the attached image is the msi ventus 2080 ti card with all locations which need to be covered, pads for everything bar the core use paste, double check everything, and fix any missing areas if any

Jay85 · 29 Dec 2019 at 14:05

Sounds like bad contact to me, take the block off and see if any pads are mis-aligned or missing and also how the contact is between the block and the core by looking at the paste.

GilesGuthrie · 29 Dec 2019 at 14:08

Thanks for your thoughts everyone. My initial instinct was to disassemble and check contacts but I wanted to check if I’d missed any configuration steps that would save me the effort and/or prevent the reassembly not fixing the issue.

Cheers

Mcnumpty2323 · 29 Dec 2019 at 14:14

GilesGuthrie said:
Thanks for your thoughts everyone. My initial instinct was to disassemble and check contacts but I wanted to check if I’d missed any configuration steps that would save me the effort and/or prevent the reassembly not fixing the issue.

Cheers

With that water block does it matter which ports you go in and out of?
Some you can use whichever ports even ones on same side ie 2 top or 2 bottom
Some blocks it may matter which you use same as cpu blocks in/out can matter
Though not mounted flat on the gpu die would be my first thought

wookiee87 · 29 Dec 2019 at 14:21

Mcnumpty2323 said:
With that water block does it matter which ports you go in and out of?
Some you can use whichever ports even ones on same side ie 2 top or 2 bottom
Some blocks it may matter which you use same as cpu blocks in/out can matter
Though not mounted flat on the gpu die would be my first thought

with that block it dosnt really matter which ports you use, having used one my self i had it one either side oposte each other and both on the same side and noticed no problems with temprature, sure a couple of degrees diffrence between configurations but noting huge

Mcnumpty2323 · 29 Dec 2019 at 14:24

wookiee87 said:
with that block it dosnt really matter which ports you use, having used one my self i had it one either side oposte each other and both on the same side and noticed no problems with temprature, sure a couple of degrees diffrence between configurations but noting huge

Thanks that rules out 1 thing then

MikeTimbers · 1 Jan 2020 at 19:26

No way it should be at 70C. What clockspeed (GHz not FPS) is it hitting when at load and at that 70C temp?

GilesGuthrie · 1 Jan 2020 at 19:39

Alright, so I got some time, drained the loop, and disassembled the card. The issue revealed itself fairly clearly. There’s a definite NSFW pic of it here: https://imgur.com/gallery/tnEpoXB

So kids, don’t do the “pea-sized blob of TIM” method. Instead, go for the “spread a thin layer with a credit card” method. We’re back up to 120FPS, with the GPU regularly boosting over 1850MHz at 65 degrees.

Thanks again for your advice everyone.

wookiee87 · 1 Jan 2020 at 19:56

GilesGuthrie said:
https://imgur.com/gallery/tnEpoXB

ouch i'm suprised the gpu's working at all, but still you new temps at 65 degrees are still a bit high for watercooling what rad's have you got?

tyler_jrb · 1 Jan 2020 at 20:17

GilesGuthrie said:
Alright, so I got some time, drained the loop, and disassembled the card. The issue revealed itself fairly clearly. There’s a definite NSFW pic of it here: https://imgur.com/gallery/tnEpoXB

So kids, don’t do the “pea-sized blob of TIM” method. Instead, go for the “spread a thin layer with a credit card” method. We’re back up to 120FPS, with the GPU regularly boosting over 1850MHz at 65 degrees.

Thanks again for your advice everyone.

Still personally wayyy hot, mine hits max 45-50c after hours of gaming at solid 2075-2100mhz. It looks to me like the paste hasn't even attempted to spread. Have you used the correct screws for the core? They need to be the shortest screws supplied and its easy to get mixed up with the crappy bag of randoms EKWB supplies. If they aren't the correct screws then the thread comes to a stop before fully tightening the pcb to the block. Resulting in poor temps.

The fact that the standoffs have come out with your pcb likely suggests the above.

In terms of paste, I always just put a line straight down the middle of the die and 2 smaller lines in the middle to form a + pattern.