Upgrading to Custom Loop

[tomdean]

Please review this and advise me on poor choices, better components, etc.

My goal is to experiment with OC and to run tasks at 4.5 ~ 5 GHz. The
machine will be mostly 90-100% loaded. For now, there will be a
low-end graphics card. This may change if I add CUDA
processing. Then I will add a 2nd loop and maybe a 2nd radiator.

I have a Bequiet Dark Base 900 case with all the drive mountings removed.
Front, 3 140mm fans intake
Top, 2 140mm fans exhaust
Rear Top, 1 140mm fan exhaust
Gigabyte TRX40 AORUS Xtreme
AMD Ryzen Threadripper 3970X

The radiator will be mounted on top with 2 140mm fans exhaust. Push-pull?

The bottom of the case is open with a light filter. So, air
flow will be in the front and bottom, out at the top and back. The
case has vents on the sides.

Primary components:
Thermaltake Pacific PR22-D5 Pump/Reservoir Combo
EKWB EK-Quantum Velocity CPU Water Block, AMD Threadripper sTR4 DRGB, Nickel/plexi
Alphacool NexXxos UT60 Radiator 140mmx2 Brass/Copper Black

Connectivity
PrimoFlex Advanced LRT Tubing - 1/2in. ID X 3/4in. OD Clear
PrimoChill Anti-Kink Coils 3/4" OD - Perl
Barb Fitting for ID 13mm (1/2in) *Black*, G 1/4 BSPP Brass

In case...
Alphacool HF G1/4" 10mm Male to Male Extender Fitting
EKWB EK-AF G1/4" 12mm Female to Female Extender Fitting

Since I am EE/CS/MA,
Thermaltake Pacific TF2 G1/4 Copper Core Temperature and Flow Indicator CL-W275-CU00SW
2x Phobya G1/4" Inline Temperature Sensor, 50cm, Nickel

The fans, cooling and some in case temperatures will be monitored by
a net connected RPi.

If anyone is interested, I can post a schematic in a few days.

 

[EarthDog]

I'd imagine you to be a lot closer to 4.5 GHz with ambient water cooling. 280W CPU at stock and you're trying to run all c/t at 4.5 GHz+ = a lot of Watts (heat!).

I would go with 3x140mm radiator for adequate headroom and quiet operation. Since the UT60 is 'only' 10 FPI, I'm not sure how much push/pull fans will help... but I'd still do it (do you have the room above the mobo?) if you want to stick with 2x140mm and get 'better + quieter than air' temps. then when you decide to add a GPU, add a 2x140mm.

I'd also mount either radiator in the front for now. That case is, from your experience in your other thread about temps, not great with airflow. So I'd want the rad to get the coolest air possible. The internals can take the few degree difference, but, since you're (apparently) after better temps than the other thread (and especially because you want a 2x140mm rad), I'd want the coolest air possible (from direct intake).

Tubing is garden hose size, lol. You don't need to go that big if you don't want to. There are few benefits to do so. Maybe go 3/8 with 1/2 OD. I just think they look unruly at much more than 1/2".

 

[tomdean]

This machine is in a remote location with solid sides and a 18-22C ambient temp.
Noise is not really a factor. I can't hear it anymore anyway!

The water cooling systems I have experience with (at least those with no cooling tower) used chilled water and had 8~10C delta-T across the heat exchanger on the load side.

How much delta-T, comparing ambient air to radiator exhaust air, can I expect at high power? I saw some graphs, but, can not find them now. What do you see at full load?

 

[EarthDog]

2x140mm's worth of rad isn't a lot for the loads it will be under...

Delta-T varies wildly due to several factors... who knows what the ambient air to exhaust will be. Def. a few C warmer inside the case.

When you say 'high-power' it would be more helpful to know the wattage that means. We talking 400W? 450W? Either way, a 2x140mm seems like too little rad for the load. A general rule of thumb is ~100W for every 120mm of rad for optimal results (quiet and better temps).

 

[tomdean]

Surely, someone has measured the exhaust temp from a radiator and the ambient temp?

Hardware labs says their GTR 280 (2x140mm, 45mm thick) has a capacity of 1300W. The radiator I selected is 60mm . Should do as well?? Remember, noise is not a factor.

 

[EarthDog]

Surely. However your system is different than others...so knowing mine (others) with different loads, rad, airflow, and ambient temps wouldn't apply to yours. It's warmer...how much varies with chassis, airflow, rad, etc...

1300W is a ton for 280mm worth of rad.... I wonder what the delta is on that... I've run 500W on a 120mm rad... but temps hit the 90s on the device (gpu), and the air blowing out of that radiator was bordering hot/wouldn't want that dumped inside my case if possible.

 

[tomdean]

I WILL NOT have the radiator exhaust air going into my case. Why add heat when the goal is to remove heat??
Someone, Please post ambient air temp and radiator exhaust air temp.

 

[EarthDog]

I WILL NOT have the radiator exhaust air going into my case. Why add heat when the goal is to remove heat??

Maybe I'm confused. I thought you're looking to make the cpu temps cooler? To help with that endeavor, ideally you want the coolest air reaching the radiator... which means air not warmed by the internals and exhausting through the rad. That said, in a chassis with good airflow, this isn't typically an issue. But you said yourself in the other thread that chassis doesn't have good airflow. In the end, 'case heated airflow' isnt going to be a lot warmer anyway (a couple of C, id imagine....again this varies wildly on chassis, flow, etc...you have a robust vrm on that board so warmed air shouldn't be an issue internally.

But if you're looking for the best cpu temps, obviously getting the radiator the coolest air possible will do that.

So, for ****s and giggles - in a completely different chassis, different cpu, with different airflow and a 3x120 rad, when the loop is saturated, I see a difference of ~4C from intake temp to the other side of the rad. Your mileage will vary. Dumping 400W into a 2x140mm id imagine the temp difference to be more. How much, we can't answer.

 

[tomdean]

I have greatly modified the air flow in the case from the previous thread. I added 2x140mm exhaust fans to the top of the case. Now, the case has GOOD air flow.
Please read the next sentence carefully.

I do not intend and will not install a radiator so the exhaust air from the radiator goes into the computer case. The intake air to the radiator will be ambient.

I have no experience with an air-cooled heat exchanger (radiator).

 

[tomdean]

... when the loop is saturated, I see a difference of ~4C from intake temp to the other side of the rad. Your mileage will vary. Dumping 400W into a 2x140mm id imagine the temp difference to be more. How much, we can't answer.

What do you mean by "when the loop is saturated?

 

[EarthDog]

Please read the next sentence carefully.

You too....... please understand I'm not trying to force your hand to do anything. My goal here is to get facts available so you can make the most informed decision you can. But it's shared so you have a firm foundation to stand on. What you do with it, is on you!

Understand if you say your goal is X, but the steps to get there won't do so or does so in an inefficient way, we'll say something... which is why you had the advice to add more fans with your air cooler (while still having more intake CFM than exhaust).

I do not intend and will not install a radiator so the exhaust air from the radiator goes into the computer case. The intake air to the radiator will be ambient.

How are you going to do that while mounting the radiator on top? You'd have to have those fans blowing IN the chassis, not exhausting out of the chassis as stated in your first post. That would leave one exhaust (rear) right? You'd have slightly warmed air hitting the rad. It won't be much (though it could be if you get a higher-powered GPU (even water-cooled the back of the GPU dumps a lot of heat into a chassis), but it not be ambient (at least, I've never been able to achieve that and had chassis that were better for airflow than the DB 900.

I have no experience with an air-cooled heat exchanger (radiator).

What does this mean? Relevance? Help me out here, lol!

What do you mean by "when the loop is saturated?

The loop takes time to reach equilibrium to the load that is being served and the rate at which it can dissipate the load. When testing for custom water cooling temps, you want to run for AT LEAST 45 mins (I like an hour or so) so the water temperature has a chance to warm up to its steady-state.

 

[tomdean]

Where did I say I was going to mount the radiator so its exhaust air went INTO the chassis? Never happen on one of my systems.
A radiator, simply, is a heat exchanger. In this case, it removes heat from the cooling water and adds to the air flowing through the radiator.

 

[EarthDog]

Where did I say I was going to mount the radiator so its exhaust air went INTO the chassis? Never happen on one of my systems.

You didn't, but you said you wanted ambient air to go across a top-mounted rad, right(?)...there's only so many ways that can happen.

The radiator will be mounted on top with 2 140mm fans exhaust. Push-pull?
...
The intake air to the radiator will be ambient.

...if the rad is mounted up top, you could only accomplish that by the air flowing top to bottom (out of the case to in the case...which you've made clear multiple times you don't want to do, lol). If you have the air go from bottom to top (in the case to out) then you aren't getting ambient air to the rad. However, you're going to try to pump what I'd guess to be 400W from the CPU so the VRMs on the mobo will run warm/hot and exhaust mostly out the top of the case through the top-mounted radiator. The GPU is a factor as well, regardless if it's low power. When you get a bigger GPU, it will radiate more heat (even if it's water cooled) up through the top/rear. So any air coming from inside the case will be warmer than ambient by a couple of C. How warm varies on many factors. What's better or worse needs to be tested for your specific configuration.

Edit: To simplify, you have to choose the lesser of evils... which, it feels like in your case, is lower CPU temps and overclocking. If you want the lowest CPU temps, you need to get as close to ambient/ambient air temp to the rad. This is typically done by placing the rad in front. Indeed it exhausts warmed air into the chassis (again as I know you don't want to do, lol), but this is, generally, the way to best CPU temps. Your mobo and low powered GPU won't have any issue with this. The bigger the rad, the lower the temperature coming out of the fins (though the wattage load remains the same).

The other option is to mount it up top and let the rad take the slightly warmed case air through it and out the top. This keeps things in the case cooler, but raises CPU temps (by the difference of ambient air to the air coming across the rad).

The third option is as above, mount the rad up top with fans blowing in the case... but the rad location is right above the VRMs, so I would revert to mounting one up front so at least it has a chance to mingle with cooler air. You can mount fans on the bottom as intake too (right?), that would help a front mount scenario (again, I know you don't want that... just putting all options and their outcomes on the table for ya. )

EDIt2: Apologies if I'm being particularly dense here. If there's some config I'm not thinking about, please, gently (lol), let me know what I'm missing.

 

[tomdean]

How about mounting the radiator above the case, aligned with the left size, as viewed from the front, with intake on the left, exhaust on the right, merging with exhaust air from the case.

I really do not have any restrictions on mounting the radiator. This system is not for beauty. No fancy LED's. No colored water.

Another subject: I have been reading about silver kill coils. Almost as many opinions as posts.
I plan on running distilled water. If I have to , I can change it more often. Maybe monthly drain half and fill half? Drain and fill at the same time to avoid air pockets?

 

[EarthDog]

Sure that will work. An external mount isn't typically pretty, but, functional!

 

[tomdean]

New question: I am planning a test stand for the cooling system. The part I am hung up on is how to get measured power into the CPU block? Has anyone done this?

 

[EarthDog]

I'm not sure what you're asking here, but I'll take a stab at it..

...you can get CPU power readings from software. Hwinfo64 shows that (and a shedload more), and Coretemp likely works too. Radiator placement shouldn't matter.

....or are you saying you're going to build it first outside of the case and want to know how to put a heatload on it to simulate a cpu??? You can use an adjustable hot plate to give you an idea but unless its the same size as the CPU IHS, the data you get won't be terribly accurate (good enough for sure) and translate to your cpu.

 

[tomdean]

I have a 40A power supply that will put out 450W. I can PWM that to get any power level I want. So, with this I can get an accurate measure of power.
I can machine a copper block with a flat surface to match the CPU block.
The problem is how to transfer power to the flat surface. Construct an insulated box, put a heating element inside. Put the flat block on top. screw down the CPU block. After some settling time, may be within a few %. Do this in steps from 100W to 400W+-,depending on temps.
1. Set a power level - this can be to < 1%. P=I*E, measured with 16-bit ADC.
2. measure temps until stable - need dead air around the test stand. - with 10-bit ADC.
3. collect data. water flow and temps. Air flow and temps. Humidity.
repeat.
Stand back, in case it catches on fire or explodes!

 

[EarthDog]

Sounds like a lot... to what end, I'm not sure(ends justify the means?). For me, I'd just put 3x120's worth of radiator on it for that CPU and go. You're making this way more complicated than it (building a custom loop) needs to be... but if that's what floats your boat, we're here!

 

[tomdean]

How do people normally mount radiators? I think I have to make brackets that fits under the fan screw heads. I want to allow for push-pull fans. The radiator came with 30mm screws, 35mm screws,and, really short screws. I think these later may have been intended for mounting,