Watercooled GPU but HOT backplate

[eidairaman1]

Not all backplates serve to provide just card rigidness but a thermal solution too.

Im presuming this card is a totally non reference design

 

[thesmokingman]

You have an air pocket in your cpu block, which leads me to think you don't have enough pump. Also remove the flow meter, that's a waste of precious flow and removing will recover some head pressure from the pump. That block... what to say... its a big sink for the vrms, but it defeats the point when the sink is covered by a big sheet of plexi. I don't get wtf they were going for there.

 

[Sasqui]

Condensation in the res means the fluid is getting pretty warm, and that means the components are getting hot. The heat on the backplate could be from the GPU.

I mentioned this previously, all the heat is being first dumped in the res, then to the radiator. It really should be the other way around.

At least that is what I see from those pics.

 

[Carsomyr]

You have an air pocket in your cpu block, which leads me to think you don't have enough pump. Also remove the flow meter, that's a waste of precious flow and removing will recover some head pressure from the pump. That block... what to say... its a big sink for the vrms, but it defeats the point when the sink is covered by a big sheet of plexi. I don't get wtf they were going for there.

About the order in which liquid is circulating. I guess I could invert the way water circulates. By having the tube go from the bottom of pump to components then rad... But Boy o Boy would that involve a lot of work. Do you how many hours it took to get those Bends on PETG hard tubes 16mm? (those are thick tubes) If I had to redo the whole thing just to get rid of some condensation... might as well just fill up the res (which I don't mind doing)

Plus my temps aren't that high. I mean, the backplate is hot yes, but the core temps seem super good:
CPU: usually maxes out around 55c-58c. that's at 5.1ghz i7 7700k delidded with 1.31vcore
GPU: Seems to hover around 26-30 for idle, 36-41c under full load. I'm guessing on a hot summer day might go to 44-48c. When I was EXTREME testing stressing the card yesterday, after a couple of hours of continuous stress testing, I think it reached 44-45c on the GPU core... But that's not real world probable scenario. The card's never gonna get tortured that way ever lol

<<< ABOUT PUMP >>>

See embedded picture below:

but again... I was under the impression that the d5 vario was plenty of power for this loop.
quoting someone who posted in this thread earlier here:

"And it's plenty, i have a D5 working for two Machines at the same time ( looped ) and moving 6 1⁄2 liter of water at 2800rpm without problems."

 

[thesmokingman]

but again... I was under the impression that the d5 vario was plenty of power for this loop.
quoting someone who posted in this thread earlier here:

"And it's plenty, i have a D5 working for two Machines at the same time ( looped ) and moving 6 1⁄2 liter of water at 2800rpm without problems."

Look at your cpu and wasted head pressure with the flow meter. If you have enough pump, you wouldn't have air pockets. D5 pumps are not known for high head pressure regardless.

Since you added more to your post re; the pump. The D5 vario isn't a super strong pump, it is a high flow pump can push lots of water when there is little restriction, but its very weak when there is restriction. That's why when ppl use D5 pumps then don't load it with a lot of high restriction blocks or go parallel as much as possible. There are also ppl who believe D5 pumps can push any load regardless of the facts. Shrugs...

 

[infrared]

Sorry smoking man, but a D5 is absolutely fine for 1x cpu, 1x gpu, 1x rad. The pump would hardly even notice a flow meter. A dual pump setup would be totally overkill and a waste of money.

Seriously, I don't think there's even a problem here, it's perfectly fine for the back of the card to get hot, I'm pretty sure the back of my reference 1080Ti gets a lot hotter! Time to stop worrying and enjoy your creation


edit: I've got a temp sensor in the reservoir on both my wc'd rigs, I tweak the fan curve to not let the coolant exceed 40c. 40c coolant = 48c GPU load temp for me, so yours looks like it's 'in the ball park'. I'd expect your coolant to be cooler than that though, since you've got less components and more rad area. For comparison my 6700k rig has 2x 300W GPU's, CPU & VRM on only 360mm of rad area (1x 120mm, 1x 240mm) so it's fairly under-rad'd tbh. That's all on 1x D5 as well, definitely no issues with flow rate!

 

[trog100]

I've got a thermaltake core p5 case with a zotac arcticstorm 1080 ti graphics card. I noticed the temps on the GPU are absolutly awesome. However the backplate gets so hot that even the water in the reservoir next to the gpu seems a bit affected (condensation)...

Is it normal for backplates to get that hot?
I thought that the waterblock touched the components that could heat up leading to a cooler temperature for all parts of the card.

What are your thoughts about this and suggestions?
Aside from a million case fans :|

with some cards it is.. water cooling wont change much or it might makes thing worse.. i had a pair of 980 TI cards in sli.. lets just say for example that the temps read 70 C.. the entire card including the backplate would heat up to 70 C.. it took two or three minutes for the cards to heat up or cool down but that is what happened.. i now have a pair of 1070 cards which use much less power but have a similar coolers and the entire card does not heat up like the 980 TI does..

a million case fans wont help but case side fans blowing cool air directly at the cards will.. it was the only way i could tame my pair of power hungry 980 TIs..

trog

 

[EarthDog]

You frame capped and neutered your cards.

 

[trog100]

You frame capped and neutered your cards.

i did as part of my heat control system.. but having swapped them for a pair of less power hungry 1070 cards with very similar performance and coolers i no longer bother.. there is no longer the need..

trog

ps.. but i must add i cant see any noticeable gain from not running the 75 fps fame cap.. playing battlefield 1 at a capped 75 fps looks just the same to me as running it at an uncapped 120 fps he he..

i run fraps otherwise i would not have a clue what frame rates i am running.. it might just be down to g-sync but i do wonder why i cant see the differences others claim to see..

 

[Carsomyr]

Techpoweru

Sorry smoking man, but a D5 is absolutely fine for 1x cpu, 1x gpu, 1x rad. The pump would hardly even notice a flow meter. A dual pump setup would be totally overkill and a waste of money.

Seriously, I don't think there's even a problem here, it's perfectly fine for the back of the card to get hot, I'm pretty sure the back of my reference 1080Ti gets a lot hotter! Time to stop worrying and enjoy your creation


edit: I've got a temp sensor in the reservoir on both my wc'd rigs, I tweak the fan curve to not let the coolant exceed 40c. 40c coolant = 48c GPU load temp for me, so yours looks like it's 'in the ball park'. I'd expect your coolant to be cooler than that though, since you've got less components and more rad area. For comparison my 6700k rig has 2x 300W GPU's, CPU & VRM on only 360mm of rad area (1x 120mm, 1x 240mm) so it's fairly under-rad'd tbh. That's all on 1x D5 as well, definitely no issues with flow rate!

LOL tech power up the place you go to for advice where everybody disagrees and where you'll find you're a genius and a Moron at the same time lol!

So.... Here's the bottom line guys:

Indeed the backplate gets very very hot. I confirm, very hot indeed. I can place my hand on it and wait like 5 seconds before I can't take it anymore...

The heat is a bit higher compared to my previous card which I still have, the gtx 1080 ti aorus xtreme... My reason for changing was that the card was choking and not getting enough air because of the front tempered glass window...

That didn't happen with my previous 1080 but I take it the 1080 ti obviously requires more cooling and tends to heat up more... Anyhow I didn't want my front glass off all the time. (even though that seemed to fix the problem of the aorus xtreme... Temps were just fine without the window on...)

but the case was ugly and cost me a lot and I was like nope... Let's get a water cooled one. Zotac had the looks I was looking for so I went for it.

Final result:

Look at the zotac design here:
https://www.vortez.net/news_story/z...rce_gtx_1080_arcticstorm_with_waterblock.html

You will find that they are blatantly lying. Indeed no coverage of the vrm section with their water block. I should have known better. In fact my very first idea was to keep the aorus and install an ek waterblock on it... Which I didn't because I got seduced by the looks and the fact I wouldn't have to install the water block myself on a gpu (I don't have experience with gpu disassembly plus water block installation)

Boy o boy was I wrong!

Here on techpowerup and on the internet as well you get people that claim as the absolute truth that backplates high heat is normal and then others say it's not...

Well I made up my mind after what happened yesterday evening:

Again the backplate was being very hot. I was looking at the space I had for fans helping the plate and there was just no room... Then I played ark and started getting a few artifacts Here and there on a very very mild gpu core overclock with the memory not overclocked. For me that was the confirmation that thesmokingman is right...

Plus the water condensation which I wasn't getting before when I didn't include the gpu in the loop

Plus the heat is so huge on the backplate... Even the room temperature becomes a bit uncomfortable for my liking.

I talked to Newegg. I'll get a full refund on the zotac and I ordered the ek backplate and water block got the aorus xtreme édition!

Stay tuned for interesting comparison results. We'll be able to see the difference with the backplates and the aorus xtreme is the only card I know of with a power target limit of 150% it will he interesting to see what kind of OC I can get on it once it's watercooled!

As for the components order and routing on the build I'll just have to keep it the say it is. Don't feel like spending another 20 hours on tube bending

 

[EarthDog]

i did as part of my heat control system.. but having swapped them for a pair of less power hungry 1070 cards with very similar performance and coolers i no longer bother.. there is no longer the need..

trog

ps.. but i must add i cant see any noticeable gain from not running the 75 fps fame cap.. playing battlefield 1 at a capped 75 fps looks just the same to me as running it at an uncapped 120 fps he he..

i run fraps otherwise i would not have a clue what frame rates i am running.. it might just be down to g-sync but i do wonder why i cant see the differences others claim to see..

Perhaps next time a form of fiscal responsibility will take over and you get one card. A 1080Ti would have been fine at your res/Hz and still cheaper than 2 1070s as well as less hassle, but, Ive led that horse to water. At least you aren't neutering your cards intentionally. Baby steps.

 

[Shambles1980]

You should embed a pic of that in this thread. Truly awesome looking rig. Aside from the VRM possibly not being cooled enough, you probably should have put the radiator in front of the res in your loop, it appears to be the other way around. All the heat is hitting the res before being cooled off.

That does not really matter as the system will eventually arrive at a nominal temperature.. does not matter where the rad(s) is/are in the loop.

its a bit counter intuitive. but lets imagine you have 3 rads.
You go
res>pump>rad>cpu>rad>gpu>rad>res
everything will still be the same temp as if you went.
res>pump>rad>rad>rad>cpu>gpu>res

Also the water flow thing isnt a MASSIVE concern either. if the water is moving fast it absorbs less heat whilst its actually in the block, where if its moving slow it absorbs more. it evens out to the same amount of heat dissipation achieved in the same amount of time..
The water flow would need to be RIDICULOUSLY low for it to be a factor.

More water in the system can help keep the system cooler for longer due to differential heat transfer (the colder the water is the more heat it can carry in a shorter space of time, and the more water there is the more mass there is to heat up so it takes longer) but even then its still going to reach an equilibrium given enough time.
and over all its going to end up at a nominal state just as with the other examples above.

P.s

as for the back plate..
Well that's just normal

 

[trog100]

i can see why some say its normal for backplates to get hot and some say its not..

in an identical set up the back plates on my 980 TI cards ran near to what the sensor read out told me.. and 75 C is fucking hot to the touch.. not burning hot but too hot to comfortably keep your fingers on.. i put a heat gun on mine to check actual temps on different parts of the cards..

but with my current 1070 cards the entire card does not get hot and the backplates run much cooler.. which is why i said its normal for some cards but not others..

so both answers are in fact correct..

i did the side swap from 980 ti cards to 1070 cards for mining purposes.. no other reason.. the 980 TI cards ran too hot and used too much power for 24/7 mining..

nice cool cards and nice much cooler backplates.. with both cards running roughly the same temps..

trog

Perhaps next time a form of fiscal responsibility will take over and you get one card. A 1080Ti would have been fine at your res/Hz and still cheaper than 2 1070s as well as less hassle, but, Ive led that horse to water. At least you aren't neutering your cards intentionally. Baby steps.

the only way to find out is suck it and see.. when i bought my 980 cards they were top of the line.. i did think such cards in sli might cause heat problems they did.. but i got around the heat problems in my own way.. the cards did a fine job as gaming cards for a couple of years..

they have been changed as i have said before because they not not much use for mining.. and believe it or not a pair of 1070 cards (because the scaling for mining is 100%) piss all over a single 1080 TI for mining purposes.. the small extra cost is more than justified..

one day dude if you watch what i do and learn why i have done it.. you may learn something useful.. he he..

i aint too sure about that fiscal responsibly you mention though.. i often save money on some things so that i can waste it on others thing.. quite what that makes me i havnt a clue

trog

 

[EarthDog]

one day dude if you watch what i do and learn why i have done it.. you may learn something useful.. he he..

LOL, no thanks. Overbuying and neutering just isn't in my blood. I spend less money and waste less time buying appropriately for my needs.

Mining saved your arse. Cheers bud.

 

[Sasqui]

That does not really matter as the system will eventually arrive at a nominal temperature.. does not matter where the rad(s) is/are in the loop.

It does if he doesn't want condensation in the reservoir, not to mention extending the pump life. The radiator should be the first in line after all the heat generation (CPU/GPU)

 

[EarthDog]

There is condensation in my res, PERIOD. It doesn't need to be hot. Its there cold. Since its in the inside of the res, that would be a product of evaporation though, no?

AS far as loop order, its been proven (Skinee labs) that a properly radded and flowing loop, temperatures internally do not vary by more than 1-2C. The only really relevant point for loop order is making sure the res is before the pump so it's fed properly.

EDIT: Flow rates were also covered. There is a sweet spot (1-1.5 GPM). Diminishing returns more than that... I do not recall below it...

 

[newtekie1]

LOL tech power up the place you go to for advice where everybody disagrees and where you'll find you're a genius and a Moron at the same time lol!

This should literally be the moto for TPU!

Indeed the backplate gets very very hot. I confirm, very hot indeed. I can place my hand on it and wait like 5 seconds before I can't take it anymore...

If it doesn't burn you instantly, it isn't that hot. Seriously.

You will find that they are blatantly lying. Indeed no coverage of the vrm section with their water block.

That isn't true. There is VRM coverage on the block.

Plus the water condensation which I wasn't getting before when I didn't include the gpu in the loop

You are going to get this no matter what once you add the GPU to the loop. You are dumping a ton more heat into the fluid, and it is getting hotter because of that. This is how physics works, condensation is caused by warm water evaporating and then condensing on cooler surfaces. Warm surrounding air around the res would only cause the condensation to happen less. Condensation goes from warm wet air to cool surface. The air around the outside of the res is cooler, making the walls of the res cooler, the air inside the res is warmer and full of moisture. Hence, the moisture condenses on the walls of the res.

The hot back plate has nothing to do with this.

AS far as loop order, its been proven (Skinee labs) that a properly radded and flowing loop, temperatures internally do not vary by more than 1-2C. The only really relevant point for loop order is making sure the res is before the pump so it's fed properly.

Exactly this. The fluid is going to warm up to an equilibrium. The temp of the fluid is not going to vary by more than a few °C from beginning to end.

 

[Sasqui]

AS far as loop order, its been proven (Skinee labs) that a properly radded and flowing loop, temperatures internally do not vary by more than 1-2C. The only really relevant point for loop order is making sure the res is before the pump so it's fed properly.

You [Skinee labs] are saying that warm water coming into a radiator comes out the same temperature or only within 1-2c of the incoming fluid? I'd like a link to that, sir.

making sure the res is before the pump so it's fed properly.

Yes, that I agree with.

 

[thesmokingman]

This should literally be the moto for TPU!



If it doesn't burn you instantly, it isn't that hot. Seriously.



That isn't true. There is VRM coverage on the block.



You are going to get this no matter what once you add the GPU to the loop. You are dumping a ton more heat into the fluid, and it is getting hotter because of that. This is how physics works, condensation is caused by warm water evaporating and then condensing on cooler surfaces. Warm surrounding air around the res would only cause the condensation to happen less. Condensation goes from warm wet air to cool surface. The air around the outside of the res is cooler, making the walls of the res cooler, the air inside the res is warmer and full of moisture. Hence, the moisture condenses on the walls of the res.

The hot back plate has nothing to do with this.



Exactly this. The fluid is going to warm up to an equilibrium. The temp of the fluid is not going to vary by more than a few °C from beginning to end.

You're missing the OP's point. In this case they are using full coverage as in a literal term, yea its literally covered but its not literally cooled. This is crazy, how many posts deep and no one has noticed the vrm heatsink is covered in a sheet of plexi. The next guy who says it doesn't need air is gonna get one... lol. Even swiftech stopped making blocks like this in principle, as in their fullbody heatsinks. If you look on their product page, even their testing results bear this. A heatsink, even a fullbody heatsink is not enough left to its own devices. Now do one worse and take that fullbody heatsink and cover it in a sheet of plexi, and see how that works. That's what we have here with this Zotac block.

 

[Ferrum Master]

Now do one worse and take that fullbody heatsink and cover it in a sheet of plexi, and see how that works. That's what we have here with this Zotac block.

I told that from the start... heat doesn't disappear... also adding fans would not help... there are no fins... surface area is so small it will not act as a proper finned heatsink, thermodynamics are thermodynamics. It is a design derp for sure, Zotac should be spanked. Okay the same block design worked for 1080, but not the Ti.

 

[FireFox]

quoting someone who posted in this thread earlier here:

"And it's plenty, i have a D5 working for two Machines at the same time ( looped ) and moving 6 1⁄2 liter of water at 2800rpm without problems."

Yeap, it was me.

If you have enough pump, you wouldn't have air pockets. D5 pumps are not known for high head pressure regardless.

Since you added more to your post re; the pump. The D5 vario isn't a super strong pump, it is a high flow pump can push lots of water when there is little restriction, but its very weak when there is restriction. That's why when ppl use D5 pumps then don't load it with a lot of high restriction blocks or go parallel as much as possible. There are also ppl who believe D5 pumps can push any load regardless of the facts. Shrugs...

Sorry Bud but i totally disagree with you, why?

This:

Sorry smoking man, but a D5 is absolutely fine for 1x cpu, 1x gpu, 1x rad. The pump would hardly even notice a flow meter.

Exactly.

 

[newtekie1]

You're missing the OP's point. In this case they are using full coverage as in a literal term, yea its literally covered but its not literally cooled.

Yes, it is cooled. I get what you are saying, but the grooves are for looks, not for cooling. The VRM heat is going into the block, the heat from the block is being transferred to the fluid by the fluid paths. There is more than enough surface area along the fluid path to pull the VRM heat out of the block.

If the fluid wasn't pulling the VRM heat out of the block, the block would be getting super hot to the touch. But it isn't.

You [Skinee labs] are saying that warm water coming into a radiator comes out the same temperature or only within 1-2c of the incoming fluid? I'd like a link to that, sir.

I can't exactly tell you the temperature of the water when it leaves the radiator, but I can tell you that after it leaves the radiator it only goes up 1 or 2°C before getting back to the res.

There is a very good article over on overclockers.net that goes into the math, but it basically boils down to(no pun intended) every 50w of heat the water has to absorb, raises the temperature of the water 0.19°C. Assuming 1gpm of water flow(pretty reasonable for a D5). This is because water has a pretty high thermal capacity.

Even if you assume some pretty high heat output for the parts, the water temp rise is only a couple degrees. If you assume, 150w for the 7700k, and 300w for the 1080Ti, and then another 50w for the D5 pump, you are only at 500w. That's only 1.9°C of fluid temperature increase from when it leaves the radiator to when it goes into the reservoir.

 

[EarthDog]

That's also assuming perfect transfer of the load. It's also a consideration of how long the water is in the rad. My biggest loop with 2 rads, a pump, res, cpu, and dual GPU block didn't have a Liter of water and around 7 feet of tubing the rads were double pass. If you are rocking 1.5 GPM its not sitting in the rad for long. It cycles ~6 times a minute/every ~10 seconds through the loop...or ~1-3 seconds it's in the rads.

I couldn't dig the skinne labs link up, but many many threads referring to it by name. Here is something from EK though: https://www.ekwb.com/blog/does-loop-order-matter/

'Optimal thermal performance' isn't defined, however skinee/thermodynamics...1-2C...

 

[Carsomyr]

Yes, it is cooled. I get what you are saying, but the grooves are for looks, not for cooling. The VRM heat is going into the block, the heat from the block is being transferred to the fluid by the fluid paths. There is more than enough surface area along the fluid path to pull the VRM heat out of the block.

If the fluid wasn't pulling the VRM heat out of the block, the block would be getting super hot to the touch. But it isn't.



I can't exactly tell you the temperature of the water when it leaves the radiator, but I can tell you that after it leaves the radiator it only goes up 1 or 2°C before getting back to the res.

There is a very good article over on overclockers.net that goes into the math, but it basically boils down to(no pun intended) every 50w of heat the water has to absorb, raises the temperature of the water 0.19°C. Assuming 1gpm of water flow(pretty reasonable for a D5). This is because water has a pretty high thermal capacity.

Even if you assume some pretty high heat output for the parts, the water temp rise is only a couple degrees. If you assume, 150w for the 7700k, and 300w for the 1080Ti, and then another 50w for the D5 pump, you are only at 500w. That's only 1.9°C of fluid temperature increase from when it leaves the radiator to when it goes into the reservoir.

Look at the image below and explain to me how the VRM is being efficiently cooled. Somehow, the image strikes a different impression in my mind:

 

[EarthDog]

I dont think anyone can tell squat from that image.....