[SOLVED] Is custom cooling for only the GPU worth it?

yaggaz

Distinguished
Nov 17, 2013
86
2
18,535
0
I am an eVga fanboy all the way so with the 3080s coming out I'm going to go either with their Hybrid or Hydro Copper and if the latter it will be my first crack at custom water cooling. My CPU temps are perfectly fine with Dark Rock Slim so I'm wondering, is custom cooling worth it for just one piece, or should you only go custom for both CPU and GPU? Thanks
 
Last edited:

rubix_1011

Contributing Writer
Moderator
Most of liquid cooling really depends on what on what you are trying to accomplish. I've always watercooled and will likely continue to watercool my main PC builds. I like the challenge and re-use the majority of my components for as long as I can. It's expensive up front, but can be less so over time due to correct maintenance and upkeep on pumps, radiators, fittings, etc.

If you are about absolute cooling above all else, watercooling is a good everyday, bang-for-the-buck approach before you get into stuff like realm of water chillers and Peltier units, which still are an advanced form of watercooling. It is still a price premium over just using the factory installed cooler, and cutting corners for cost savings is not recommended until you have a very firm idea of what is advisable to get creative with and what might not be (like pumps...always make sure you have a good pump).

If this is about trying something new and liquid cooling seems neat, then using an AIO with a mounting plate is a decent gateway into this arena. You aren't worrying about coolant and fittings - everything is pre-installed and filled. Its a more plug and play approach. It also starts you down the path of making sure your new cooling solution will fit the GPU you've purchased....often a card is bought and then GPU blocks are shopped - often the opposite way it should be. You want to make sure your cooling solution has a block or component that will fit specific, known SKUs of graphics cards, and then search out those graphics cards.
 
Reactions: yaggaz

rubix_1011

Contributing Writer
Moderator
I've watercooled for about 19 years. Currently have a RTX 2080 and it never sees above 45C at most loads...maybe close to 50C under prolonged Folding @ Home, but only on really warm days. Idle temps are relatively normal - 23-25C, depending on room temp.

Of course, this is a full watercooling loop with CPU also cooled. Dual 240mm EK PE rads and D5 pump.

You'll see a lot better temps on direct die cooling (like a GPU) than you will with an indirect method like IHS and solder/compound between die and IHS on CPU.
 

Phaaze88

Glorious
Ambassador
I'm sure it is addictive.
Too bad so many people blindly dive in without doing the research, or they do it with the wrong focus in mind: thermals.

The maintenance part is what I believe makes or breaks most of 'em.
 

yaggaz

Distinguished
Nov 17, 2013
86
2
18,535
0
So I am currently struggling with my RTX 2080 Super. When I first noticed it throttling at 80c, without ANY OC whatsoever, I got some Arctic MX4 and redid the paste. Thermals dropped to 60 to 65c on Average. Then over time it keep sneaking up and up until it was throttling at 80c again.

So got a H55 AIO. It causes Temps to be averaging 50c to 60c depending on the game, then over time, slowly creeps up again, now back to struggling with throttling at 80c again. W.T.F???? Other advice I received was to go back to more dense paste, but I'm wondering if I just got a bad die that pulls away from the heat sinks on the cooler and AIO?
 

Phaaze88

Glorious
Ambassador
Yeah that's what scares me. I'm not lazy but I am a panicker lol. I was having a panic attack just putting on the AIO, terrified I was gonna kill my card.
I was referring to custom loops. Most hybrid coolers require no real maintenance; they're easy mode.
-flushing the loop
-can't just use any cleaning solutions. For example, some cleaning solutions cannot be used with aluminum parts = corrosion. Also, some cleaning solutions don't mix well with certain coolants either, if they happen to mix, such as EK Cryofuel + Mayhems Blitz.
-refilling and leak testing
-if using pastel coolants - not dyed coolants - they clump up really quickly, and because of that, need to be replaced and refilled often, some as early as a week!
Just a few steps of just how involved the user needs to be regarding custom loops.

As for that gpu: EVGA's been using shorter(length) but thicker heatsinks compared to most of the competition.
Perhaps one or some of the screws securing the heatsink to the PCB were actually stripped? That would cause the heatsink to 'hang off' or not sit flush with the gpu/Vram/VRM.
That's what sounds like is happening in your situation.
If you actually "got a bad die that pulls away from the heat sinks on the cooler and AIO", the gpu would die if that actually happened, right?
That's why I think there's a mating issue somewhere.
 

rubix_1011

Contributing Writer
Moderator
So I am currently struggling with my RTX 2080 Super. When I first noticed it throttling at 80c, without ANY OC whatsoever, I got some Arctic MX4 and redid the paste. Thermals dropped to 60 to 65c on Average. Then over time it keep sneaking up and up until it was throttling at 80c again.

So got a H55 AIO. It causes Temps to be averaging 50c to 60c depending on the game, then over time, slowly creeps up again, now back to struggling with throttling at 80c again. W.T.F???? Other advice I received was to go back to more dense paste, but I'm wondering if I just got a bad die that pulls away from the heat sinks on the cooler and AIO?

Well, the H55 is only a 120mm AIO, the 2080 is a 215w+ TDP part, but I doubt it would struggle to not keep it at least in partial check. Wondering if there is something going on with the mount of the cooling block to the GPU die. It likely is not a paste issue (always seems that's the assumption), its likely a contact issue, but you'll check paste in the process.

Also wondering if there is any issue with the radiator placement - can you post an image of how you have this configured?
 

grimfox

Distinguished
Jun 2, 2009
1,015
44
19,690
159
@yaggaz If you just want lower temps, water cooling is probably not for you. There are other things you can do to manage temps that don't require as much work or money as water cooling. If you want to water cool because you want to water cool then we are here to help.

Water cooling is a hobby within a hobby. I compare it to the people that like to paint their own miniatures for DND or warhammer etc. They like playing the game, but also like doing something different that's related to the game. The skill sets don't really overlap. The same is true of water cooling.
 
Reactions: Phaaze88

rubix_1011

Contributing Writer
Moderator
Most of liquid cooling really depends on what on what you are trying to accomplish. I've always watercooled and will likely continue to watercool my main PC builds. I like the challenge and re-use the majority of my components for as long as I can. It's expensive up front, but can be less so over time due to correct maintenance and upkeep on pumps, radiators, fittings, etc.

If you are about absolute cooling above all else, watercooling is a good everyday, bang-for-the-buck approach before you get into stuff like realm of water chillers and Peltier units, which still are an advanced form of watercooling. It is still a price premium over just using the factory installed cooler, and cutting corners for cost savings is not recommended until you have a very firm idea of what is advisable to get creative with and what might not be (like pumps...always make sure you have a good pump).

If this is about trying something new and liquid cooling seems neat, then using an AIO with a mounting plate is a decent gateway into this arena. You aren't worrying about coolant and fittings - everything is pre-installed and filled. Its a more plug and play approach. It also starts you down the path of making sure your new cooling solution will fit the GPU you've purchased....often a card is bought and then GPU blocks are shopped - often the opposite way it should be. You want to make sure your cooling solution has a block or component that will fit specific, known SKUs of graphics cards, and then search out those graphics cards.
 
Reactions: yaggaz

yaggaz

Distinguished
Nov 17, 2013
86
2
18,535
0
Also wondering if there is any issue with the radiator placement - can you post an image of how you have this configured?
Certainly and thanks for advice. As you can see, the dark rock slim blocks rear positioning. There is no floor or wall mounting and tubes aren't long enough to reach the ceiling. I even had it sitting outside the case for a few hours while gaming with a mock, taped cardboard wall sealing the internal temps of the case, but the thermals inside the case remain delightful, except the GPU still which was pushing toward 80c still, all the while the radiator being external with air conditioned room temps outside:

 

Phaaze88

Glorious
Ambassador
I think I know where grimfox was coming from with that statement - poor choice of words though.



Truth be told, that H55 shouldn't actually be enough for that card.
-120mm hybrid coolers are around the level of a Hyper 212 Evo or worse, so around 120-150w of heat?
-A gpu's TDP or max power draw is attributed to the entire package, not the gpu core alone, but the core on it own does account for like 3/4ths or 2/3rds of that power anyway
-If the gpu is bouncing up and down on load, the H55 should actually be fine
-If you're somehow managing to keep the gpu pegged at high loads throughout it's use, then yes, I can see that unit struggling. The gpu core would be capable of some 200w+, and a 120mm hybrid cooler can't deal with that long term.

Something else that I just remembered that separates my setup from yours: Did you use thermal pads and heatsinks for your memory? I know the Kraken G12 kit doesn't come with any.
-Nvidia's gpus currently report thermals as an average of 3: the core, the Vram, and VRM.
-The GDDR5X on my 1080Ti doesn't need those pads and sinks at all; my card sustains thermals under 40C even under sustained load. Different story with GDDR6 - and likely 6X on the 30 series - on the 20 series card you have, as it has higher operating temps by default.

So, if you didn't put pads and heatsinks on your memory, then the reason the gpu averages are so 'bad' is because the Vram is running hot.
 
Reactions: yaggaz

rubix_1011

Contributing Writer
Moderator
Yep, I wouldn't recommend a 120mm AIO for a GPU...at least a 240mm should be considered.

One of the primary problems with the general public looking to AIOs and watercooling is the misconception of 'just having a radiator' or 'just having liquid cooling' makes it 'good liquid cooling'....you still have to account the thermal loads being added to the cooling loop. Liquid cooling isn't magic, planning and correct application are still required to make the solution work as expected for the application and time/money spent.

There can't be an expectation that a 120mm AIO on a GPU will perform like a full, custom watercooling loop with dual 360mm rads and a high flowing D5 pump pusing 1gpm.
 
The maintenance part is what I believe makes or breaks most of 'em.
Phaaze is quite correct. Maintenance is high on the list. You have to clean and dump them twice a year, and sometimes disassemble the block every two years to clean the cold plate fins with a toothbrush.

Graphics cards produce the most heat in your system, not the CPU. If you want the quietest setup and best overclocking/boost possible (albeit small gains), then full block water cooling with a decent sized radiator has no substitutes.
 
Last edited:
Reactions: yaggaz

yaggaz

Distinguished
Nov 17, 2013
86
2
18,535
0
Something else that I just remembered that separates my setup from yours: Did you use thermal pads and heatsinks for your memory? I know the Kraken G12 kit doesn't come with any.
-Nvidia's gpus currently report thermals as an average of 3: the core, the Vram, and VRM.
I remember doing research on this and I came across a couple stories where the heatsinks over time came loose, fell into the fan and were thrown about the delicate circuitry, so I shied away from that and instead put a silent wings in the Kraken to cool the VRM and RAM and put it up as fast as I could stand it.

Note that I have an evga Super Black, I think I read that only certain versions of their card had multiple temp sensors on them and the Black wasn't one? But if it is the average then it sounds like you are onto something. But if it IS the average, then why after reapplying paste do the temps plummet, stay that way for a few weeks then start climbing again?
 

yaggaz

Distinguished
Nov 17, 2013
86
2
18,535
0
By the way the evga Hybrids for the 30xx is a 240mm radiator, so going by the advice here it should be more than enough for a non overclocked card? Wondering if a 320mm with the Hydro Copper custom would be overkill then?
 

Phaaze88

Glorious
Ambassador
I remember doing research on this and I came across a couple stories where the heatsinks over time came loose, fell into the fan and were thrown about the delicate circuitry, so I shied away from that and instead put a silent wings in the Kraken to cool the VRM and RAM and put it up as fast as I could stand it.
I did similar with my Kraken G12, but with an NF-A9. But as I understand it, that's just not enough for GDDR6 memory.

Note that I have an evga Super Black, I think I read that only certain versions of their card had multiple temp sensors on them and the Black wasn't one? But if it is the average then it sounds like you are onto something. But if it IS the average, then why after reapplying paste do the temps plummet, stay that way for a few weeks then start climbing again?
I keep forgetting that EVGA has something like that.

Thermal paste is just a bridge; it's primary role isn't cooling. It makes me SMH when I see people suggest others use top end, expensive, thermal paste when they're having high temperature problems.
If the paste is doing it's job correctly, then the heatsink/radiator/pump/fans can do theirs. When it is not, then the others can't either. There are 3 instances where the paste can't do it's job:
-Incorrect application. For example, the dot in the middle doesn't work as effectively on Ryzen 3000 because of the multiple dies around the sides of the actual cpu.
-Dries up, which can take a few years or more with known brands.
-Cooler not mounted correctly.

Now, why would it stay that way for a time, and then start climbing again?
Sounds like something is coming loose over time, and that's not the paste's fault...

By the way the evga Hybrids for the 30xx is a 240mm radiator, so going by the advice here it should be more than enough for a non overclocked card? Wondering if a 320mm with the Hydro Copper custom would be overkill then?
2080 Super Black Edition vbios: https://www.techpowerup.com/vgabios/212660/evga-rtx2080super-8192-190701
Board power limit
Target: 250.0 W
Limit: 292.0 W
^Again, that's across the entire package, and the gpu core makes up the bulk of that. You also have to remember that the card will try to boost higher on it's own if the available headroom is present.

https://www.techpowerup.com/vgabios/191285/gigabyte-gtx1080ti-11264-170323-1
That's my 1080Ti.
The latest version of Gpu-Z shows power draw from the core, and I've seen it pull quite close to the 250w board power target... again, that's by itself, even with the small overclock I have on it.

Unlike gpu heatsinks and full cover water blocks that handle the entire package, the hybrid coolers isolate the gpu core from the rest.
The H55 can manage, but only if the core is allowed to throttle down. If whatever you're doing constantly has it sitting at high usage for extended periods of time... I think we've already covered that 120mm hybrids can't really handle 200w+ of heat...
 
Last edited:
WORTH is a personal thing.
Reading the question, my reaction is if you need to ask the question, then it is not worth it.

If the objective is better performance, my advice is to buy a stronger card in the first place.
 

neojack

Prominent
Apr 4, 2019
383
61
790
9
i didn't really read everything on this thread.
BUT THIS PICTURE IS FRIGHTENING

ok simple rule : in a watercooling system, the pump, should always be fee with water by gravity.
it means the resevoir should be placed on top, and the pump bellow.

in a AIO, the radiator acts like a reservoir.

yes you waterblock and pump is now full of air.

doesn't work well ? no kidding

here is a full video of gamer nexus on the subjet. please share !
View: https://www.youtube.com/watch?v=BbGomv195sk
 

madmatt30

Titan
Ambassador
i didn't really read everything on this thread.
BUT THIS PICTURE IS FRIGHTENING

ok simple rule : in a watercooling system, the pump, should always be fee with water by gravity.
it means the resevoir should be placed on top, and the pump bellow.

in a AIO, the radiator acts like a reservoir.

yes you waterblock and pump is now full of air.

doesn't work well ? no kidding

here is a full video of gamer nexus on the subjet. please share !
View: https://www.youtube.com/watch?v=BbGomv195sk
You're misinterpreting the video, the rad is fine where it is, I'd go a slot higher but that's just me because I'd want it central for aesthetics.

Im just not sure of the split airflow from a front intake and a pull fan on the rear of the radiator.
Unless these are synced/matched properly you're not getting uniform airflow through the rad.

I just don't think the h55 is quite good enough to be running on that gpu.
 
Reactions: digitalgriffin

Karadjgne

Titan
Ambassador
When it comes to pastes, there's several criteria that differentiates them. Arctic silver 5 has a finite drying time of approx. 200 heat cycles. Noctua, thermal grizzly and a few others are more oily than pasty, low viscosity, so are far easier to spread and get a good spread. Mx-4 is more pasty, thicker viscosity.

When it comes to cpu coolers, a thinner viscosity paste does better. It's in the nature of the IHS to base contact. Gpus are direct die, no IHS, and have a mirror-like surface. Low viscosity pastes do not do so well there, the thicker and tackier pastes do far better, making Mx-4 one of the best use gpu pastes around.

There's a difference between air and liquid. The metal in aircoolers transfers heat extremely fast, taking seconds to transfer heat from the cpu to the fins and gone. Put a pan on a stove and see how long it takes to burn your fingers. Liquids pass thermals extremely slowly, takes a good half an hour at load to acclimate the coolant, put water in that same pan, see how long it takes to come to boil.

So you are going to see time gaps where gpu temps are OK, but an hour later, same game, they aren't.

As for split airflow, it won't make a difference. Not enough static pressure from the intake fan. The rad fan is still doing the lions share of the work, the created low pressure pulling in the intake air, not the other way around.
 
Reactions: yaggaz
You're misinterpreting the video, the rad is fine where it is, I'd go a slot higher but that's just me because I'd want it central for aesthetics.

Im just not sure of the split airflow from a front intake and a pull fan on the rear of the radiator.
Unless these are synced/matched properly you're not getting uniform airflow through the rad.

I just don't think the h55 is quite good enough to be running on that gpu.
He is quite correct. You are misinterpreting the video.

H55 is a crap cooler. I owned two, I hated them both. The plate surface on two were absolutely uneven and made a clean mate near impossible with a ton of screw down force (which had me snap a screw head)
 

ASK THE COMMUNITY

TRENDING THREADS