my semi-custom loop for 3900x system

D

Darth Ender

Gawd
Joined
Oct 11, 2018
Messages
543
I have a thermaltake water 3.0 240mm aio and an arctic liquid freezer 240 that had pump power failure about 6-8 months into it's life.

The thermaltake has woefully insufficient flow rate to handle the 3900x at full load. Temps rise to 80C range and as soon as load drops, so does temps, instantly.

This tells me that the radiator is more than adequate to dissipate the heat, but the flow rate is not.

My plan is to add a second pump (modtek silence) and use the other 240mm radiator attached to the front so the loop goes like this:

cpu block + pump -> orig radiator at top (exhaust setup) -> 2nd pump -> 2nd radiator (intake setup) -> back to cpu block

The extra parts and all should be coming tomorrow and I should be able to test the setup before the weekend. The extra radiator might be overkill, but I have it sitting around doing nothing so why not.


I've been looking at review sites and each one seems to have drastically different numbers that make it impossible to take those numbers and compare to yourself or other sites. Frostytech has test numbers that make every liquid cooler look like they're amazing with single digit or low teens above ambient with 125+ watt loads on the cpu blocks. play3r.net and whatever else we can find these days since hardocp is dead have completely different numbers (40c over ambient, etc).

I'm seeing load temps of around 60C over ambient, and maybe 3-4C over ambient at idle with the thermaltake system. 50-55C over ambient when I undervolted the 3900x.

meh, just curious if other real-world system uses find themselves closer to what i'm seeing or if what I'm seeing is exceptional. The thermaltake is less than a year old. I'll be doing a full cleaning and breakdown so I'll see what the state of this aio is after such a short time of use and see how the cpu block looks before replacing all the liquid with some new stuff i'll be mixing up.


edit:

So parts list is as follows so far:

pump1 + radiator 1 = Thermaltake Water 3.0 240mm aio

pump2 = modtek 17w 12v - 253.7gph , pwm controlled

radiator 2 = 240mm radiator from Arctic Liquid Freezer 240 aio

Hoses will stick to the aio sizes of 1/2" od 1/4"id

liquid additive = primochill liquid utopia.


basically this is the original build with my old 1800x and the arctic cooler before the pump died


The new 3900x is essentially all the same except the arctic has been replaced by the thermaltake, new mobo and new cpu and no stupid wire tie on the gpu corner :)

The second radiator will go in the front against those intake fans. The second pump will likely be mounted to the back motherboard tray area. Sitting between the two radiators.
 
Last edited:
3900X runs hot under full load with all the cores. I have a custom loop and at full load it's at around 75C on the chip and at 32C at idle and the room is 24C. Your numbers dont look unreasonable for the cooling you have. Also those sites comparing water blocks are usually comparing the water temperature over ambient difference not the cpu temp to ambient.
 
That would be odd unless all you are measuring is the radiator or all the blocks are identical. A crappy water block can absorb less heat and yield low water temps and appear like a very good block and better radiator .

Aio's don't fall into that category.

In any case,. I dont doubt the 3900x runs really hot under full load. Just means you need better cooling. :)
 
why? the weakest part of an aio system is the pump and I'm basically adding a pump that doesn't need the aio's at all to be more than adequate. The aio pump is just going to be there for moral support.

everything else would be kinda wasteful to replace. blocks aren't going to be significantly better performing since they're all basically the same ... radiators are radiators and I'm not running pure water so I dont care about additional galvanic corrosion any more than basically every commercial water cooling setup sold does. None of it is being done or setup for eye candy, the hoses are all back, there's no see thru blocks or anything. No led's on the water block.

I can't help but feel like with a pump upgrade, an aio system is going to be indistinguishable from any other similar radiator sized system in terms of cpu cooling performance.
 
There's a world of difference between quality rads and blocks vs AIOs. Cost being one of those differences of course.
 
Meh, as far as build quality, you'd be pretty hard pressed to find 80 dollars worth of difference that matters.
https://www.xtremerigs.net/2015/02/11/radiator-round-2015/5/

Good cross section of a bunch of radiators from various sources ...at normal fan speeds, the best and worst dissipate more than 2x the wattage than a 3900x puts out. So unless you're undersized, you should be fine and as long as you're flowing 1gpm and able to push air thru it, it'll do the job and more.

blocks are blocks. Most adopt the same micro-fin cross flow design but pretty much any design ends up performing within a couple degrees of eachother. Look at this little roundoup . https://www.techpowerup.com/review/aqua-computer-cuplex-kryos-next-cpu-waterblock/7.html

The truth is water block choice is more aesthetic than anything else. AIO's adopt the same basic designs as these have. Exotic materials and carefully designed water pathways all dont appear to really add up to any meaningful difference in performance.

Certainly not going to perform 60+ bucks better than the water block that i already have with the aio.

So spend about 200-250 bucks on replacing the radiators and water block for something that will perform no better or dont spend that money. I could spend the money, i just dont see why i would. The one aspect of water cooling that is still annoying would exist with both systems, maintenance, and frankensteining the aio's will accomplish the goals desired just as much as scratch building.
 
Darth Ender said:
This tells me that the radiator is more than adequate to dissipate the heat, but the flow rate is not.
Click to expand...

Nope. If the temp drops quickly, the rad is working, but it also means you have plenty of flow.

Your issue is the block on your CPU
 
I'd be surprised if the block factored in because that would indicate that the surface area must be significantly lower than any normal water block and that seems unlikely without heavy corrosion/blockages.

Possible, but unlikely. I'll find out soon enough though when I open it up this weekend.

More likely is that the flow rate is low, so the water comes in as cold as it can get but isn't moving fast enough to avoid taking on far more heat than would be efficient given the radiator size and all. Low flow rate is common in aio's and less than 60gph on a 240 where the volume of water is vertical is not outside of possibility.

The other option is that basically no water block can move the heat well enough to give significantly better temps. That could also be the case given that we're not really seeing non-chilled setups being talked about left and right where the 3900x is running with max temps in the 60's during all core benchmarks @ > 4.2Ghz.
 
Darth Ender said:
Meh, as far as build quality, you'd be pretty hard pressed to find 80 dollars worth of difference that matters.
https://www.xtremerigs.net/2015/02/11/radiator-round-2015/5/

Good cross section of a bunch of radiators from various sources ...at normal fan speeds, the best and worst dissipate more than 2x the wattage than a 3900x puts out. So unless you're undersized, you should be fine and as long as you're flowing 1gpm and able to push air thru it, it'll do the job and more.

blocks are blocks. Most adopt the same micro-fin cross flow design but pretty much any design ends up performing within a couple degrees of eachother. Look at this little roundoup . https://www.techpowerup.com/review/aqua-computer-cuplex-kryos-next-cpu-waterblock/7.html

The truth is water block choice is more aesthetic than anything else. AIO's adopt the same basic designs as these have. Exotic materials and carefully designed water pathways all dont appear to really add up to any meaningful difference in performance.

Certainly not going to perform 60+ bucks better than the water block that i already have with the aio.

So spend about 200-250 bucks on replacing the radiators and water block for something that will perform no better or dont spend that money. I could spend the money, i just dont see why i would. The one aspect of water cooling that is still annoying would exist with both systems, maintenance, and frankensteining the aio's will accomplish the goals desired just as much as scratch building.
Click to expand...

Dude, you gotta stop with the opinion/assumptions passed off as fact. A real loop will cost around 300-400 for starters depending on how many blocks/rads. Each part is unique, blocks are not blocks, pumps are not just pumps. The bits you get in an aio tend to be cheap junk.
 
thesmokingman said:
Dude, you gotta stop with the opinion/assumptions passed off as fact. A real loop will cost around 300-400 for starters depending on how many blocks/rads. Each part is unique, blocks are not blocks, pumps are not just pumps. The bits you get in an aio tend to be cheap junk.
Click to expand...

i know how much diy loops cost, I've built them before. And 10 years ago the blocks were all within a couple degrees of eachother and today they're still within a couple degrees of eachother. That's reflected in pretty much every single comparison test you can find.

Nobody said pumps were pumps, that's the specific single thing I said was the major variable.

Radiators aren't all the same, sure, but even the crappier ones (referring to 360mm sized ones) are more than 2x sufficient to dissipate the heat produced from a cpu like the 3900x (120 watts) at a reasonable fan air flow and pump rate.

I do not doubt the pumps in aio's are cheap and the fluid is mostly anti-freeze instead of water. But where we differ is in how much better quality you think your overpriced radiators and water blocks are to what amounts to a major savings in cost due to mass production / order quantities afforded to the major brands. Are you paying for quality or are you paying for eye candy and low production quantities ? I'll find out this weekend when I open mine up and see. Though judging from the insides of other brands that I've seen, it's no different than pretty much any other block you can get out there being sold for 60+ bucks.
 
Those 3900s running at 4.2 are putting out way more than 120w of heat. As for those charts, it's worth mentioning those are all still custom loop parts. It'd be nice if they took apart some aio systems to include.

I do have an old h110 lying around. I'm feeling curious enough to pop the top off the block to check it out. As for the rad, it’s thinner, has fewer channels, has a simpler fin design, and has narrower inlets for more constricted flow. No doubt to keep up pressure with the weak pump.

I do wish I had a 3900 to play around with, maybe I can sneak one past the wife? Lol.
 
The arctic cooler water 240mm aio Pump power died about 6-8 months into use.

Seems to use straight anti-freeze ...about half a cup to 1 cup total volume

copper block is covered by a rubber gasket seal that controls the flow inlet and outlets and seals to the pump cover. Looks like the gasket is glued to the copper plate. I'd imagine the max pressure a setup like this can take is not very much. How much is too much? dont know. but I am wondering if getting close to 1gpm is even possible if my current aio employs a similar setup

Looks like water inlets are on either side of the center channel where they then go through the fins to the outlets on both sides of the outside area of the fins. Gunk in the middle caused by cooking the anti-freeze when the pump died (1800x survived just fine)

They purposely stripped all the screws holding the copper plate in but luckily i have micro-broken screw removers just for this stupid kind of occasion.



















 
Last edited:
All in all though, very much the same design as diy water blocks. The fins even look to be fairly wide and tighter than what you'd find in a hundred dollar ekwb velocity. Though, seems fairly clear that while that hundred dollar block is cut from a single piece, this looks to be two piece. Either welded or soldered (probably soldered) together. whether those things cancel eachother out...meh.. dont know.

The primary downside is the relatively weak gasket seal and extremely small volume of liquid involved. I am actually hesitant now to add my secondary pump since at the time, this arctic system was said to have a fairly strong pump for an aio and if my thermaltake employs a similar setup, even at 50% power the new pump may be too strong and blow a gasket seal.
 
Darth Ender said:
All in all though, very much the same design as diy water blocks. The fins even look to be fairly wide and tighter than what you'd find in a hundred dollar ekwb velocity. Though, seems fairly clear that while that hundred dollar block is cut from a single piece, this looks to be two piece. Either welded or soldered (probably soldered) together. whether those things cancel eachother out...meh.. dont know.

The primary downside is the relatively weak gasket seal and extremely small volume of liquid involved. I am actually hesitant now to add my secondary pump since at the time, this arctic system was said to have a fairly strong pump for an aio and if my thermaltake employs a similar setup, even at 50% power the new pump may be too strong and blow a gasket seal.
Click to expand...

Since Asetek makes most of the AIO's, its very likely the pump is the same or very similar.
 
after seeing the method used to connect the pump to the water block plate, my main concern is that the volume of liquid being restricted to such tiny amount and this restriction being secured by just a silicone gasket glued to the plate and pressure fit to the pump will lead to a massive buildup of water pressure if I connect my pump to the system. Even with a second radiator. A gasket blowing out is a real possibility. Would not be surprised though if the flow restriction is so great that 1gpm is not even possible thru the block. It was literally so small in the other pump setup that liquid was happy to just stay in it even when turned upside down. Had to shake it to get the little bit of liquid to drain out.

Very likely I'll be moving to a ekwb velocity water block or similar. Still plan on using the radiators I have though. All will depend on how testing goes today when i hook up the new pump and see how the flow rate behaves thru them.
 
Darth Ender said:
after seeing the method used to connect the pump to the water block plate, my main concern is that the volume of liquid being restricted to such tiny amount and this restriction being secured by just a silicone gasket glued to the plate and pressure fit to the pump will lead to a massive buildup of water pressure if I connect my pump to the system. Even with a second radiator. A gasket blowing out is a real possibility. Would not be surprised though if the flow restriction is so great that 1gpm is not even possible thru the block. It was literally so small in the other pump setup that liquid was happy to just stay in it even when turned upside down. Had to shake it to get the little bit of liquid to drain out.

Very likely I'll be moving to a ekwb velocity water block or similar. Still plan on using the radiators I have though. All will depend on how testing goes today when i hook up the new pump and see how the flow rate behaves thru them.
Click to expand...

Those radiators are going to be made of aluminum. I know you said you weren't concerned about corrosion but you should be. Those AIO's mix metals, but they are completely sealed units and steps are taken to try and prevent corrosion. That doesn't even work allot of times. Liqtech coolers are a prime example of that. They had a high failure rate on some models due to that very issue.
 
They don't run water. That much is obvious. All glycol/antifreeze

Absolutely no corrosion in the system and it has been sitting stagnant for a year in my closet.

Water flow tests look good with the new pump and aio radiator from the Arctic system.

This radiator is definitely larger than my thermaltake's though in terms of thickness.

Still debating if I want to cut apart that working aio though since I'm fairly decided already on not using the water block like I had originally planned.

Just no way to flow 1gpm and not leak due to pressure buildup with the tiny channels it has all made up of flexible silicone.
 
I know they don't run water. Similarly, you can't completely seal the system once opened. There is a big difference between a sealed AIO and AIO parts in an open loop. Corrosion via mixed metals is a very real problem. There is plenty of documentation and images out there of that very problem. Using those shitty radiators just isn't worth it.
 
ive refurbed a couple corsair aios, still work great. just cleaned the clogged barb and refilled with af/dw mix. we used to use jankier setups than this. keep going op, I want to see what you come up with.
 
what makes a radiator shitty is being too thin, not having dense air fins and maybe having some inefficient water channel setup. The material it's made out of doesn't make it shitty. It just means you have to pay attention to what else you're adding just like if you had decided to buy all copper setups. The thing is, it's easy to buy an expensive but poorly made copper radiator because they still want to make a profit and they're using an expensive material. So they usually skimp out on air fin density etc.

In any case, i've posted comparison tests above that have shown even large differences in quality of radiator still are more than 2x the capability of heat dissipation than what i'll be putting into them. You're concerned over differences that dont matter.

My primary issue is that I have to run basically antifreeze in my system to eliminate any corrosive behavior with the coolant. If i buy a standalone radiator so I dont have to cannibalize the thermaltake setup, it'll almost certainly be copper since that's all microcenter stocks. But if my liquid is mostly glycol or similar compounds, it shouldn't matter any more than it matters with aio setups.. which appears to be not at all from what i've seen in taking one apart.

hrm.. decisions.

edit. also, I wouldn't be making an open loop. I never make open loops. Always fully sealed closed loop systems that get no maintenance. Previously these have been in all copper setups.
 
Last edited:
So didn't see a point in cutting up the tt aio just to use the radiator since I'd not be using the block as originally planned after seeing how the other aio is sealed around the block and the crazy flow restrictions.

Basically this is now just a custom loop with a stolen radiator from an old aio.

I got a ekwb 360mm copper slim 3 way radiator and a ekwb velocity water block. I got the slim instead of the pe because I didn't think the pe would fit in my case with fans.

The donor aluminum radiator is pe sized and has much denser air fins so I couldn't justify wasting it. And spending another 70 bucks.

My coolant will be a dexcool compat antifreeze. Should be good for the life of the computer. It is rated for 10 years operating at 190 degrees, it should be fine forever.
 
Running some benchmarks on the thermaltake water3 now that it's had about
a week to settle in on the 3900x. Then i'm putting this new dual radiator custom loop in.
 
Some images of the new setup

1 360mm ekwb slim copper with 3 120mm fans

1 240mm asetek thick high density alu with 2 120mm fans

Modtek silent pro pump

Nickel ebwb water block

50/50 prestone with cor-guard!

Had to eye up the hoses and leave some room to install and remove it but I could probably stand to lose about 6-10 inches total. Oh well.

Couple side by side radiator pics to show just how much more dense the alu radiator is.

 
the embedded image thing is a carousel widget. Hover over the right side and an arrow appears to switch the images.

here's an initial benchmark run. Flow rate has increased massively over the aio. Cooling capacity has increased over 200%. Yet, there's been no significant decrease in temps I'll try the test again after a few days to see if any settling down occurs.

But it looks like what i'm seeing is the limits of what copper can do in terms of moving dense localized heat to water. The only way to cool better than even mid ranged aio's can do is by chilling to below ambient with some form of active cooling. 200-300 dollar custom has no hope of doing better than a properly functioning 80-100 dollar aio. At least that's the only conclusion i can see.

https://openbenchmarking.org/result/1908180-HV-1908170HV38
 
True, if you're just cooling the CPU. Most people with a custom loop cool the video card as well.

Then again the largest draw of custom loops is for the hobby side of it, at least for me that was true. I'd be hard pressed to find anyone who says the overall performance of a custom loop scales with the cost in a linear graph. My whole argument was quality of custom parts > AIO parts.


Even that radiator pic. The AIO rad might have a higher fin density, but that means nothing if the water can't transfer enough heat to the fins. The EKWB has way more channels for more water to contact fins. Might have larger inlets also. My HWlabs GTR radiator for instance only has 16 fpi, yet they're stacked so tightly that you have to put 2000+ rpm fans on or push/pull to make it shine. Fin density is like CPU clock speed, it's only a facet of performance.
 
The comparison between radiators there wasn't apples to apples either. Ekwb was a 3 way 360mm. I should see if I can find a pic to see how many channels the 240mm has since they are all 2 way like the aio is. I did look at both at microcenter though and fin density was the same for all the ekwb's

Inlet Port size is definitely larger.

I restricted my system to 6-7mm though to keep the amount of liquid in tubes down and velocity up.

The pump is running full tilt in this system and I'm not disappointed in dropping the extra 200-250 on making a custom loop system that doesn't really do much better than an aio I already had. I wouldn't have been satisfied until I knew for sure it couldn't improve the temp situation.
 
Water cooling elitism is that perfect mixture of sad and funny. People literally, in real life, becoming indignant for others deigning to mention AIOs in the same paragraph as the holy custom loop, going so far as to write several long posts to convey the strange idea that a $400 loop is better than a $79 AIO.
 
And it turns out, there exists setups where it certainly is not.

If I was doing a new system and was convinced by people that I had to do a custom loop to get good cooling out of the zen2 I'd be pretty pissed. A 100 dollar aio will do it just as good it seems.
 
  • Like
Reactions: viivo
like this
I wonder if a threadripper block wouldnt show better results with these chips? Being that most current blocks have channels that are centered covering a much smaller area vs tr blocks that are much larger in every way. I know that it wasnt the case for the 2000 cpus but it may be for the 3000 series. Its worth a try if someone has a tr block lying around.
 
hititnquitit said:
I wonder if a threadripper block wouldnt show better results with these chips? Being that most current blocks have channels that are centered covering a much smaller area vs tr blocks that are much larger in every way. I know that it wasnt the case for the 2000 cpus but it may be for the 3000 series. Its worth a try if someone has a tr block lying around.
Click to expand...

I do, but the problem is the mounting hardware is very different. The TR pattern is considerably larger.
 
I doubt it would matter. It isn't like the cores and io chip are off on the sides. They are just slightly off-center.

Something far more heat conductive than copper is needed to spread the heat out before the water factors in.
 
I'm confused - why don't you ditch the crappy AIO rad? You've already got a decent loop - reasonable rad, reasonable pump, reasonable block - but then you're going to run less efficient fluid through tiny tubes (and still risk corrosion) just to add a 240mm rad and save $30-40? If you've got "normal" fittings and tubing laying around you should test out the loop with distilled water (obviously without the aluminum 240 rad.) I'm betting the 360 by itself is capable of similar performance to the setup you listed (and I wouldn't be shocked if it actually performed better.)
 
it flowed rather well in the flow test i did with the pump just connected to that radiator. It's not a slim 240 like most aio's would have. Radiators (even 240) are closer to 55-65 bucks. Aluminum != crappy. When we're well inside overkill territory like we are with most radiators, the material it's made out of matters less than how well the thing is made. The aluminum radiator i have has a fin density that looks to be twice as dense as the copper alternatives being sold and it's as thick as a PE style. There's only one drawback to it that I had, galvanic corrosion concerns.

Switching to anti-freeze to use aluminum isn't worth not including the radiator, since I was almost certainly not going to use nearly pure water anyway since I do not take apart my water cooling loops and do maintenance on them every 6 months. I build them sealed and do not want to take them apart ever again. So non-water (or very little) was all but guaranteed. So galvanic corrosion is not really a concern. The pump will die before this liquid begins failing.


I'll have another test run of the same benchmark i ran before and after coming today as the system has had about 5 days to settle in. Temp wise, I dont think i'm missing out on much. We do not see people talking about all core full load temps in the 50's and 60's regardless of liquid or loop types.

I suspect my new custom loop will be a single handful of degrees different from my aio at most. And i suspect that will be about on par with all other top end cooling systems for people who are not chilling or running their computer in < room temps (I test mine around 74-77F always)


edit: the other consquence of having a secondary radiator is that the pump is never pumping hot water and the cpu never receives slightly warmed water.

Pretty much the only reason I'd have for evolving the system to something new now is if a heatpipe enabled water block comes out ....or I decide to water cool some future big navi video card.
 
Last edited:
YOu keep mentioning the fin density but keep glossing completely over superior the tube density in the copper one which would work better IMHO. Slightly lower fin density but more tubes for water to flow = more water flow in contact with air and less air resistance as well.
 
  • Like
Reactions: Dan_D
like this
Is there a site that benchmarks this stuff?

Would be pretty easy to bench a radiator, for instance.

Same pump, same fan... measure temp in vs temp out

Air fin and tube density are opposite sides of the same coin - it takes total exposed surface area for both mediums. Just looking at pics of the two rads I have no idea which would perform better
 
You must log in or register to reply here.
Back
Top