Anyone doing watercooling these days?
- Thread starter Cool Modine
- Start date
So would you guys go for 2 D5 or 3 D5 pumps?
Currently running 2, and I get about 1.1 GPM at full blast in a system with a 360 HWL GTS XFLOW + 480 Alphacool Nexxos Slim rad, Optimus AM4 block and Optimus 3080 FTW3 block. Using a bunch of Koolance QD3H quick disconnects as well, and some 90° fittings. I typically run it at where I get 0.8GPM.
Am replacing the Alphacool with 2x Heatkiller 480-L rads, and maybe add 2-3 more QDC sets (just makes for much easier maintenance when using soft-tubing). So am adding more rad space and more QDCs which are quite restrictive.
Currently running 2, and I get about 1.1 GPM at full blast in a system with a 360 HWL GTS XFLOW + 480 Alphacool Nexxos Slim rad, Optimus AM4 block and Optimus 3080 FTW3 block. Using a bunch of Koolance QD3H quick disconnects as well, and some 90° fittings. I typically run it at where I get 0.8GPM.
Am replacing the Alphacool with 2x Heatkiller 480-L rads, and maybe add 2-3 more QDC sets (just makes for much easier maintenance when using soft-tubing). So am adding more rad space and more QDCs which are quite restrictive.
I prefer minimal number of components in this case to reduce failure points so two pumps would be my preference before having to add a third. That said I don't know the actual details your proposed layout so no idea how hard it'd be to add a third pump after the fact?
I'm moving things over to a new V3000 Plus case, but to be fair, I have been scratching my head about how I would fit a 3rd pump. Currently, one is mounted to the reservoir, and another one separately. I can get a dual-top, but those take up quite a bit more space and might be more challenging to mount while still having my build look clean (one of my aims of this migration is to make it all look much cleaner - current build is a bit messy).
Currently planning on moving the 2nd pump to the basement of the case, out of sight, and only having the reservoir + pump visible up top.
Currently planning on moving the 2nd pump to the basement of the case, out of sight, and only having the reservoir + pump visible up top.
2 D5's is already overkill unless you need/want redundancy or you're running them a ridiculously low speeds, and in the latter case you're better off just isolating the pump better if it's due to noise. A properly decoupled D5 doesn't make much sound on a 12V supply at 100% duty cycle.
Why do you think you need that many? I'd say implement the system with 1 first and see how it runs before you go for more. If you plan on replacing rads anyway, look at going with crossflow models, they're less restrictive generally.
I'm already running 2 right now and will keep that as I migrate over to a new case and add another rad. Running 2 for redundancy and having enough flow rate.
The only reason I was considering it is because I'll be disassembling everything anyway, to move it over to a new case. Once it's all rebuilt I will not be changing anything anytime soon. My current build has been running for about 3 years without any changes.
I don't think adding another failure point would be a good idea unless you REALLY need the extra flow-rate. Moving to less restrictive rads would give you some benefit, though. And there's different sizes for QDs as well, some of them have wider IDs than others.
When you say maintenance, are you referring to replacing coolant, or replacing components, or something else? If the QDCs are so restrictive (what is their ID out of interest?) I'd question their value since you're leaving your system to it for years at a time. You may as well just have a single pump (which could be mounted entirely out of sight?) and the ability to drain the system down and swap it out for a working one, in the unlikely event that it failed.
There should be plenty of space to implement a decent drain system in that monster of a case. Probably off one of the rads in the bottom of the case, assuming they are being populated.
I kinda want to do my next rebuild in a smaller case. It'll probably be something like a LL O11D-Mini, where I can fit two 360 rads with a ITX mobo.
The alternative is upgrading the video card in my current system, but with only a single 360 rad running a 5950X and 2070S right now, it's stretched to upgrade, and with the FLT in the front of the Meshify 2 (non-XL), I don't have a good spot for another rad.
Probably end up leaving this system as a backup as is, and just do a total refresh.
The alternative is upgrading the video card in my current system, but with only a single 360 rad running a 5950X and 2070S right now, it's stretched to upgrade, and with the FLT in the front of the Meshify 2 (non-XL), I don't have a good spot for another rad.
Probably end up leaving this system as a backup as is, and just do a total refresh.
I'm redoing my loop, moving it all inside a V3000 case. Using 2x 480 rads (Heatkiller L), and a 360 HWL GTS XFLOW rad. And then I have my Optimus AM4 and 3080 FTW3 blocks.
Using a dual pump setup and trying to figure out the best routing here. This is option 1, with the yellow rectangle representing the 360 XFLOW rad.
Option 2:
I overall might prefer the aesthetics of option 1, but a potential downside is that the flow would need to go from one of the bottom ports on the front mounted rad all the way up to the top mounted rad, working against gravity and potentially impacting flow-rate.
For option 2, I don't like the look of a hanging tube going from the CPU block to the top rad, something about that irks me. And it'll get very busy with tubing down by the second pump, not sure I'll have enough space (as I do need the HDD cage that sits behind the 2nd pump. But otherwise, flow-wise it might be slightly better?
I otherwise specifically want the rads to come after both blocks. I'm using an Ultitube res which has a built-in filter, and I want any potential gunk coming out of the rads to be caught by the filter, and thus not go into the blocks. If I didn't care about that, I could simplify the routing a bit more by going from the 2nd pump into the 360 rad, and from there go into the blocks and then other rads.
Currently leaning towards option 1.
Using a dual pump setup and trying to figure out the best routing here. This is option 1, with the yellow rectangle representing the 360 XFLOW rad.
Option 2:
I overall might prefer the aesthetics of option 1, but a potential downside is that the flow would need to go from one of the bottom ports on the front mounted rad all the way up to the top mounted rad, working against gravity and potentially impacting flow-rate.
For option 2, I don't like the look of a hanging tube going from the CPU block to the top rad, something about that irks me. And it'll get very busy with tubing down by the second pump, not sure I'll have enough space (as I do need the HDD cage that sits behind the 2nd pump. But otherwise, flow-wise it might be slightly better?
I otherwise specifically want the rads to come after both blocks. I'm using an Ultitube res which has a built-in filter, and I want any potential gunk coming out of the rads to be caught by the filter, and thus not go into the blocks. If I didn't care about that, I could simplify the routing a bit more by going from the 2nd pump into the 360 rad, and from there go into the blocks and then other rads.
Currently leaning towards option 1.
Yeah, that's what I ended up doing.
Now as things always go with these types of things, I need 2 more fittings (quick disconnects to be precise) before I can wrap it up. Those will take a few days to get here, ugh. Guess I'll complete all the other wiring and stuff in the meantime.
Yeah they're big boys. But they make water-cooled life so easy. They are quite restrictive though, hence the need for 2 pumps.
I can see the appeal of QDs. But I think a properly implemented drain system would yield better results over all.
For me the #1 appeal of having QDs is I can for example disconnect my CPU and GPU waterblocks without needing to drain the whole loop first.
For example, last year I upgraded from a 3900X to a 5950X, and all I had to do was disconnect my CPU block, pop in the new CPU, add fresh thermal paste and put the same block back in place. Easy peasy, and took less than 15 mins. While draining the whole loop and all that stuff would have taken me over an hour to get it all back up and running.
I otherwise do have a drain system in place as well, in the bottom there's a valve I can open to drain out most of the loop. It's at the lowest point of the loop.
That's fair. Last time I messed with QDs in any detail was the late 90s/early oughts, and I used Colder industrial ones. They worked very well, but they added a lot of bulk. Maybe consider going with some offset rotary fittings on the MB block so you can get rid of those 90 degree elbows.
Last edited:
Alright, my transition over to a V3000 Plus case is done. Very roomy case to work in, very pleasant.
Yep, that's a super nice build. I know the fashion is for SFF builds but as someone with fairly large hands I always prefer to build in something with a bit of room if I can!
Looks like your board has liquid cooled VRMs - how are the temps given you're not putting any liquid through them?
Joining the dots between several of your prior posts, I'm thinking the reason you might need two pumps for this look is the filter in the res and the QDs. Assuming you're flushing out all of your parts before use, I don't honestly think you the filter (and I suspect the second pump is also not strictly needed, although I get the desire for redundancy). But clearly it all works just fine as is
Looks like your board has liquid cooled VRMs - how are the temps given you're not putting any liquid through them?
Joining the dots between several of your prior posts, I'm thinking the reason you might need two pumps for this look is the filter in the res and the QDs. Assuming you're flushing out all of your parts before use, I don't honestly think you the filter (and I suspect the second pump is also not strictly needed, although I get the desire for redundancy). But clearly it all works just fine as is
You're using x2 D5 pumps, running at well over 4k rpm, in a circulation setup (ie you're not lifting the fluid from A to B, you're just moving it in a circuit) and you've got multiple points of restriction in that circuit (QDs, filter) - gravity is pretty much irrelevant in this scenario!
People vastly overestimate the actual needed flowrate in loops and/or the pressure overcoming abiilties of modern watercooling pumps(D5 or DDC). We're not blasting 6+ GPM at a flat piece of copper using repurposed aquarium pumps anymore, or using said aquarium pumps with an insanely restrictive micro-jetplate design.
120-240L/h(0.5-1GPM) is all you really need to aim for unless you're running at the ragged edge of component temperatures.
120-240L/h(0.5-1GPM) is all you really need to aim for unless you're running at the ragged edge of component temperatures.
Yep - my D5 Vario is running at setting #2 / ~2200rpm, cooling a 5900X, a 3080Ti, and a chipset block via two radiators (straight flow 360 and 280) and piped with 10mm ID soft tubing and hose barbs.
I did experiment with higher flow rates, they didn't help (and this is a loop that can handle 500W at ~1000rpm rad fan speed).
I did experiment with higher flow rates, they didn't help (and this is a loop that can handle 500W at ~1000rpm rad fan speed).
I need to get a different flow meter (or put back in my old meter). I went from an Aquacomputer High Flow that was typically showing a flow-rate of 170 lph to an Aquacomputer MPS Flow that's now showing 45 lph. I did add 3 more QDCs to the loop, and another radiator, but I wouldn't expect it to affect my flow-rate that much. Temps are fine though, getting about 3C DeltaT when doing light tasks with fans spinning around 800 RPM, and 7-8C DeltaT when playing a game.
I wouldn't expect it to either, given that you have two D5 pumps in there
As long as you have "enough" flow across the heat exchange surfaces (inside water blocks and rads), the limiting factor is getting the heat out of the silicon and then out of the coolant. For CPUs we're mainly limited by the heatspreader, even though it's "only" a thin piece of copper and it's soldered to the die. ideally we'd move to bare die cooling but that seems unlikely for a user installable part. It's a die protector as much as it is a heat spreader, after all.
For GPUs, there's a certain benefit to be gained from using liquid metal TIM, and after that you're limited by the thermal conductivity of copper. I suppose they could start machining GPU blocks out of sterling silver but I suspect people don't want their full cover block to cost more than the GPU
Getting heat out of the coolant is again straighforward in that you can scale radiator area, depth and airflow to suit the application and noise tolerance. Functionally you can usually get to a point where the radiators aren't really (that) limiting either - it comes back to conduction out of the silicon.
As mentioned above, "enough" flow isn't actually very much, although clearly performance does tank if you have an issue that causes it to fall below that low level. So lots of loops are probably running more flow than they need, which won't do any harm, it just isn't really gaining you anything other than wasting a handful of watts (which ends up in the coolant and being dissapated by the radiators) and wearing the pump out a bit quicker.
None of this is a criticism of your build per se! I understand you have your reasons to want QDCs and dual pumps, they don't align with how I would build but there's a reason we call these setups custom loops
I just thought it was worth exploring in more detail since as lithographies have shrunk and power densities (as well as absolute power outputs) have increased, the bottlenecks have changed a bit.Historically silver blocks have been sold at various times.
The pressure vs. delta-T plot for a water block really depends on the design of the waterblock. You can make a block that will perform better at higher pressure/flow or blocks that will work well at lower pressure but may not scale to higher pressure.
The pressure vs. delta-T plot for a water block really depends on the design of the waterblock. You can make a block that will perform better at higher pressure/flow or blocks that will work well at lower pressure but may not scale to higher pressure.
I'm looking at upgrading my system. Need another rad to support a video card upgrade in the future it seems. With the other items in my current case (A FD Meshify 2 non-XL), there isn't a good spot.
Lian Li O11D Evo, here I come I suspect. That will let me use the existing FLT240 I have on the side mount, and add another 360 rad at the bottom, with the existing one mounted at the top.
Lian Li O11D Evo, here I come I suspect. That will let me use the existing FLT240 I have on the side mount, and add another 360 rad at the bottom, with the existing one mounted at the top.
Even setting cost aside, looks like water blocks made out of silver would not be meaningfully better than copper
View: https://youtu.be/zbWyZSuoh2w?t=804
(Der Bauer YouTube link)
View: https://youtu.be/zbWyZSuoh2w?t=804
(Der Bauer YouTube link)
45 is pretty low, especially with 2 D5's. I assume the MPS is the middle break in the tube running down the front rad, in which case you're meeting the 5cm recommendation. Is it the right sized model(200 I'm guessing) and you're not overflowing the sensor perhaps?
It's the MPS 400 actually, which should still be fine given the flow-rates I was getting before. I might pop back in my old flow meter at some point (all the newer ones are out of stock everywhere) to see what the flow-rate is like.
45 would be below the supposed measurement range of the 400, which might not be helping matters. And depending on what the "zero" reading is, that could throw things off even further as it can trim even more from the measurement range. But without a secondary source it'll be hard to track down the error if it exists.
The high flow 2 has gotten quite a bit of praise but, as you say, out of stock.
aquacomputer seems to suffer these stock shortages across large sections of their product lines every now and then and, at the risk of suffering from a first world problem, mildly aggravating to happen right now. I want to pickup an aquaero 6 after discovering the Commander Pro has a ~7V floor for DC control and uses stupidly undersized mosfets.
The high flow 2 has gotten quite a bit of praise but, as you say, out of stock.
aquacomputer seems to suffer these stock shortages across large sections of their product lines every now and then and, at the risk of suffering from a first world problem, mildly aggravating to happen right now. I want to pickup an aquaero 6 after discovering the Commander Pro has a ~7V floor for DC control and uses stupidly undersized mosfets.
Do you happen to know if the High Flow Next is basically the same as the High Flow 2, but with a display and RGB? As I just saw on the aquacomputer site that the Next is back in stock (it wasn't last week when I checked). As I might just get that instead.
It's the same internals sensor wise, the NEXT just includes a microcontroller so can operate somewhat independantly.
I"m looking at building a new system and doing a custom cooling rig.
Aliexpress has radiators for about 20% the cost of Corsair 360mm radiator
A quick sample ...
But, ... it's aliexpress. It's chinese made, made in china. Yes I am very aware that the Corsair made radiators are also made in China (and quite possibly in the same factory)
16x 5 star reviews, but many from Russia.
Is this just junk? The radiator that Corsair rejected? Or likely to be identical to more branded items.
Thoughts on other fittings from the same store on Aliexpress?
Or is it more nuanced, like "yes for simple things like a radiator, no for a pump".
Aliexpress has radiators for about 20% the cost of Corsair 360mm radiator
A quick sample ...
But, ... it's aliexpress. It's chinese made, made in china. Yes I am very aware that the Corsair made radiators are also made in China (and quite possibly in the same factory)
16x 5 star reviews, but many from Russia.
Is this just junk? The radiator that Corsair rejected? Or likely to be identical to more branded items.
Thoughts on other fittings from the same store on Aliexpress?
Or is it more nuanced, like "yes for simple things like a radiator, no for a pump".
Chinese made doesn't automatically mean "bad" any more, unless you told the factory you just wanted it to be as cheap as possible! Part of the reason they're cheap is because they're made of aluminium, rather than copper. If you're using copper blocks in your loop (and you will be) - you're gonna have a hard time avoiding galvanic corrosion, IMO. You well might be able to do it with a coolant containing a higher level of anti-corrosion agent than usual, but it's not an experiment I'd want to do with my own hardware!
EKWB for a time did sell a complete set of aluminium parts for custom loops, including radiators, blocks, fittings, etc, targetting a lower price point due to the material cost savings. IIRC is was called Fluid Gaming or something similar. Anyway, I think they stopped producing it some years ago.
In general one of the things that gets dropped with "suspiciously cheap stuff from AliExpress" is QC, so most of the units might be fine, but you have a higher than average chance of getting a lemon.
These don't look terrible FWIW https://www.aliexpress.com/item/1005004570871579.html - Barrow is a known and reasonably well respected brand in the WC world.
EKWB for a time did sell a complete set of aluminium parts for custom loops, including radiators, blocks, fittings, etc, targetting a lower price point due to the material cost savings. IIRC is was called Fluid Gaming or something similar. Anyway, I think they stopped producing it some years ago.
In general one of the things that gets dropped with "suspiciously cheap stuff from AliExpress" is QC, so most of the units might be fine, but you have a higher than average chance of getting a lemon.
These don't look terrible FWIW https://www.aliexpress.com/item/1005004570871579.html - Barrow is a known and reasonably well respected brand in the WC world.
Thing about building a custom loop is that most of the parts, chosen well, will last for aaaages. The fly in the ointment is the GPU block which is expensive and custom made to that GPU. You can also save a lot (as well as gaining reliability) by using hose barbs and spring steel hose clamps, rather than the fancy looking screw down versions the manufacturers like to sell you. Couple that with EPDM or Norprene tubing and it's not going to leak, leach plasticisers into the coolant, or go weird and sticky like the PVC stuff can do.
Doesn't stop it being a pretty significant investment at the front end, though
Doesn't stop it being a pretty significant investment at the front end, though