can we talk about water chillers please?!

newls1

Supreme [H]ardness
Joined
Sep 8, 2003
Messages
4,607
Currently, my 3950x is pretty tough to keep cool with my highend water cooling (EK 60mm 420 rad in push/pull fan config, D5, heatkiller Pro IV copper block cpu only loop) with CCD0 @ 4.525GHz and CCD1 @ 4.325GHz using 1.350v die and package temps will be VERY HARD to cool during summer times here in Atlanta Ga area. Absolutely always wanted to experiment with a waterchiller, more specifically the units performance-pcs sells. They seem priced right and have good reputations for being constant duty units (even though this pc will never be run for very long durations, its more of just a fun pc to play with) in the below links are the chillers im looking in to. Having a hard time deciding if 400watts will be enough or 790watts for extra insurance. Also, my current water cooling setup is 2 separate loops. 1 for cpu and 1 for gpu, but when I go chiller, do i combine both cpu and gpu and have the gpu reep the benefits of this type of cooling, or just keep it simple at first and make it cpu only at first...? I know people say to remove the rad from the chilled loop, but is this still the case? I mean it makes sense, might actually "ADD" heat into the loop, so i understand the theory behind it, but if i was to leave it in as a fail safe, is that bad?

I understand the whole condensation thing, i used to phase change cool with a "chilly1" phase changer back in the day with my FX55/FX57 overclocks, but these chillers I can set right to the dew point to minimize condensation, but of course i will do extra preventative measures to prevent hardware failure. Need your thoughts and opinions on this possible future cooling upgrade (pretty huge upgrade) hopefully someone has experience with waterchillers and can tell me yay or nay? Thank you

https://www.performance-pcs.com/wate...r-hc-300a.html

https://www.performance-pcs.com/wate...r-hc-500a.html
 
I don't have any experience with chillers, so I apologize for the derailment, but I've been following your 3950x temperature woes and something in this post caught my eye:

How hard would it be for you to temporarily combine your loops into one as a test case? Running separate loops is almost never thermally superior to a single loop, given the same radiator area. Very few use cases heavily stress the CPU and GPU at the same time. When you're running a CPU-heavy application in a dual loop scenario, your GPU is monopolizing a large portion of your cooling system that it isn't even using. With a single loop, both components have access to all your rad area, and that is worth more than the drawback of them sharing.
 
I don't have any experience with chillers, so I apologize for the derailment, but I've been following your 3950x temperature woes and something in this post caught my eye:

How hard would it be for you to temporarily combine your loops into one as a test case? Running separate loops is almost never thermally superior to a single loop, given the same radiator area. Very few use cases heavily stress the CPU and GPU at the same time. When you're running a CPU-heavy application in a dual loop scenario, your GPU is monopolizing a large portion of your cooling system that it isn't even using. With a single loop, both components have access to all your rad area, and that is worth more than the drawback of them sharing.
how in the world can you say running a extremely hot gpu and a even hotter cpu in 1 loop would be the better case?? I am completely lost here. Having dual independant loops obviously each with there own huge rad, pump, res, etc.. is thermally the best way to go. Why would I want to dump the gpu heat in with a hot ass cpu and even remotely thing that would be the better option. Im not trying to come off as rude, but dude i dont follow your tactics at all
 
how in the world can you say running a extremely hot gpu and a even hotter cpu in 1 loop would be the better case?? I am completely lost here. Having dual independant loops obviously each with there own huge rad, pump, res, etc.. is thermally the best way to go. Why would I want to dump the gpu heat in with a hot ass cpu and even remotely thing that would be the better option. Im not trying to come off as rude, but dude i dont follow your tactics at all
If you're gpu is running cooler than your cpu, it would benefit the cpu while hurting your gpu temps.

I'm pretty sure his suggestion was just for testing purposes only. You have the gear, dual loop would be ideal, but if you hadn't tried everything and it's easy to test..? Now if it's a full breakdown to test, than no, not worth it at all.

That said there are many ways you can chill it. There are a bigger selection of aquarium chillers that will probably do the same thing, are mostly cheaper, and can be scaled down to not run below room temp enough to condense at your block.

You can also look into adding a tec to the loop. (not under the cpu block) They're wildly inefficient, but can cool your loop without a compressor/big box outside of your case.

Those are not specific purchase suggestions. Just ideas to consider
/look into.
 
As an eBay Associate, HardForum may earn from qualifying purchases.
found the same chiller on amazon for 60$ cheaper then performance pcs, and ive messed with TEC's years back... no way in hell would i consider that route again. guess my only question now is this: should i run 2 d5's in series for this chilled loop, or will 1 d5 be ok
 
how in the world can you say running a extremely hot gpu and a even hotter cpu in 1 loop would be the better case?? I am completely lost here. Having dual independant loops obviously each with there own huge rad, pump, res, etc.. is thermally the best way to go. Why would I want to dump the gpu heat in with a hot ass cpu and even remotely thing that would be the better option. Im not trying to come off as rude, but dude i dont follow your tactics at all
I say that because very seldom if ever are you hammering both components at full load.

Take gaming as an example. Your GPU will be working hard, and in most cases (some games are exceptions) your CPU will be very lightly loaded. In a system with split loops, your CPU is maybe utilizing 20% of the thermal capacity available to it, and the GPU doesn't have access to the other 80% - it's wasted rad for that use case. In a single loop, your GPU has the entire sum of the thermal capacity that's not being used by your CPU.

There are use cases for which dual loops make sense, but they are very niche and not common at all. Rendering with GPU acceleration comes to mind, probably some deep learning and physics simulation.
 
how in the world can you say running a extremely hot gpu and a even hotter cpu in 1 loop would be the better case?? I am completely lost here. Having dual independant loops obviously each with there own huge rad, pump, res, etc.. is thermally the best way to go. Why would I want to dump the gpu heat in with a hot ass cpu and even remotely thing that would be the better option. Im not trying to come off as rude, but dude i dont follow your tactics at all

In addition to what was said, at a high enough flow rate the difference between inlet and outlet temps are minimal. Assuming you have a 1 GPM flow rate, the difference between inlet and outlet temps of a 400 watt unit is literally 1.5 C. That means if your GPU was directly after your CPU, it would see water that is 1.5 C higher than what came into the CPU with the CPU running at 100%. 1.5 C is hardly anything, and is not worth separating the loops and not having the extra radiators for cooling when load is uneven.

The math is as follows: 400 (watts, heat input) x 60 (convert to joules per minute) / 4.186 (specific heat of water) / 3785 (flow rate of water in grams per minute).
 
i appreciate the reedback, but im just going to cool cpu. gotta keep it simple at first.
 
Is it really worth it just for a cpu? You going to run into condension issues.
 
Check out tangoseal setup he has rad overkill and doesn't go over 70c on tr3.
Thermal gradient helps when silicon transfer rate is the limiting factor.

I have used chillers. But not for pcs lol. Cw3500 and similar are cheap and good, anything that can cool a large laser will work. You can set temp and they have flow alarms and level indicators etc. Pretty quiet too, surprisingly so. But I was using them around air cooled lasers with delta's in them lol. You'll need a few gallons of distilled water too.
You'll see them being sold for crappy old co2 laser rigs ( no offence to co2 users but fibre is your God ;)).

Please keep us updated as I want to test the same thing, also use liberal amounts of dielectric grease..
 
newls1 , I think I might have come off the wrong way in my post.

If you want to use a chiller because you want to use a chiller, I don't intend to dissuade you - it's a cool project and I'm curious about your results.

If you want to use a chiller because you feel like you need to use a chiller to hit the temps you want, I don't think that's necessarily the case. I believe you can get some more performance out of your existing setup - that's all. My 3900x barely breaches 60°c under stress tests in the same loop with my 1080ti while it's also being stressed, and the two are sharing a 480 and 420.
 
newls1 , I think I might have come off the wrong way in my post.

If you want to use a chiller because you want to use a chiller, I don't intend to dissuade you - it's a cool project and I'm curious about your results.

If you want to use a chiller because you feel like you need to use a chiller to hit the temps you want, I don't think that's necessarily the case. I believe you can get some more performance out of your existing setup - that's all. My 3900x barely breaches 60°c under stress tests in the same loop with my 1080ti while it's also being stressed, and the two are sharing a 480 and 420.

appreciate your feedback sir... Im still way up in the air about this project. Yes, this pc is just a "toy to play with" and if you go back years and years ago, in the 04/05 days i had 1st hand experience with a real phase change unit made by "chilly1" back on the old xtremesystems.org forums.. im familiar with condensation build up and all the BS envolved with sub ambient cooling, and THIS will not go below ambient given i set temps on chiller as per dew points. Im still thinking about this idea as its not very expensive and would be SUPPPPEEEERRR FUUUUUUN to setup, but right now im trying to find the absolute best cpu block for this 3950x and am leaning toward "Optimus PC" as their new block looks very promising. that heatplate and fins actually cover the entire surface areas of all chiplets and IO die... super nice.
 
money is DEFINITELY an object! Recommend a chiller sub 500$ please!
 
there are no blocks design specifically for 3950x, yet and is your google broken?
 
attitude? ok. can you not google what youre looking for? is that better?
how about stay out of this thread if you cant help or keep this thread on tract.. no one needs your attitude and tuff guy behind a keyboard mentality
 
i was commenting on your search for a block and i wasnt giving you attitude
but since you insist, maybe calm down with the people are mean to me mentality.
 
money is DEFINITELY an object! Recommend a chiller sub 500$ please!

Did you read my post earlier?

Cw-3000 is sub 200...
Might be just enough to do ambient but do your calculations/homework. They make bigger models too.

Here is the one I get to play with again next year;
upload_2019-12-12_3-9-22.png


https://www.teyuchiller.com/air-coo...000-9l-water-tank-110v-200v-50hz-60hz_p6.html


That one is 189 USD on amazon but IIRC are TEC designs though so pretty inefficient and low cooling capacity. But better than nothing to test with in an existing loop especially if you are only looking for ambient or close.. The CW-5000 can be temp-set, has a proper compressor and is probably more what you are after (800W capacity) but they are just over 500 USD. There is a knockoff B&H CW-5000 if you want to try your luck which is around 480.. I'd stick with S&A they are pretty reliable and use good jap/korean compressors.
 
Last edited:
Only want to add, radiators will add heat if water temperature in the radiators drop below ambient air temp.

I would think loop order here could be of high importance depending on the outlet water temp of the chiller you pick.

For flow rates, I found this article: https://www.ukgamingcomputers.co.uk/blog/does-pump-flow-rate-affect-water-cooling/

Definitely makes a difference despite the D5 being known for putting more heat into a loop. You might reach a point of diminishing returns with one, but if you're going all out I'd say series them up.
 
I have the ek dual d5 xtop on order with another d5 pump coming as well, so dual d5's will be a sure thing. This project changed a bit however. ill post up pics and project path when parts get here
 
I have the ek dual d5 xtop on order with another d5 pump coming as well, so dual d5's will be a sure thing. This project changed a bit however. ill post up pics and project path when parts get here

You could have just gone with a single 35x ya know.
 
You could have just gone with a single 35x ya know.

I'm not familiar with that one, but if the specs I see are correct they do 1050 L\h and 4.4m of head pressure. A D5 is 1500 and 3.9m, but two D5s in series will double the head pressure(7.8m) and have the same flow(1500 L\h).
 
I'm not familiar with that one, but if the specs I see are correct they do 1050 L\h and 4.4m of head pressure. A D5 is 1500 and 3.9m, but two D5s in series will double the head pressure(7.8m) and have the same flow(1500 L\h).

Those specs are meaningless in a lot of ways. We don't care what the flow is as long as it can hit 1.5gpm for the resistance in said loop. Thus the real measure of dynamic head pressure is within the range of flow we care about 1gpm-1.5gpm. In this range two D5 Strongs at max speed 24v is only slightly stronger than a single 35x.
 
A typical industrial evaporative chiller has a coefficient of performance of 4.0 or so, which means for every 100 watts used, it can cool 400 watts. The 790w chiller therefore has a CoP of around 2, probably due to scale. If the chiller is sitting in the same room as the computer, you're now outputting 50% more heat overall.
 
A typical industrial evaporative chiller has a coefficient of performance of 4.0 or so, which means for every 100 watts used, it can cool 400 watts. The 790w chiller therefore has a CoP of around 2, probably due to scale. If the chiller is sitting in the same room as the computer, you're now outputting 50% more heat overall.

Yep, I have around 1000w of TECs here that I just can't ever justify using due to their horrible inefficiency. Not in any project, believe me, I've tried to justify their use.
 
  • Like
Reactions: spine
like this
In my opinion chiller is the only sub ambient device which can be used on daily basis. You can set the themperature on the level which you want to avoid the condensation (it also depends on humidity).

I'd recommend also use a bigger reservoir (at least couple of liters) and placing the chiller outside the room if possible. With big reservoir the chiller will rarely turn of.
 
Back
Top