Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
PoizenJam
Dec 2, 2006

Damn!!!
It's PoizenJam!!!
I whinged about whether this is better suited to the build thread or this thread, but I decided to post here. I'm pretty new to water cooling solutions- bought my first AIO a little over a year ago for a new build. I plan to do a custom loop in the near future, but right now I have an O11D XL and 2 AIOs, one for a 12900k and one for a 3080. I'm curious about a technical thermodynamics question for how best to utilise these AIOs and cool my system.

Right now, my case is oriented as such:

FRONT/SIDE: 3 x 120mm fans (INTAKE)
TOP: 360mm CPU AIO (EXHAUST, PUSH CONFIGURATION)
BOTTOM: 240mm GPU AIO (INTAKE, PULL CONFIGURATION)
REAR: 1 x 120mm fan (EXHAUST)

It occurs to me that this is sub-optimal, since the heated air coming over the GPU AIO RAD on the bottom intake will inevitably go through the CPU AIO RAD at the top exhaust (albeit mixed with some fresh air from the side intakes). This is meaningful, since the 12900k struggles to avoid throttling if OC'd + the GPU is also running hot. After a bit of research online, it seems the recommendation for dual-AIO systems is to have both running as either intake or exhaust, to avoid the issue of sending heated air from one to the other. The thing is, I am not sure which is the best option here. Rather than ask you guys to tell me how to build the system, here are two alternatives I figured may work:

AIO INTAKES
FRONT/SIDE: CPU AIO (INTAKE, PULL CONFIGURATION)
TOP: 3 x 120mm fans (EXHAUST)
BOTTOM: 240mm GPU AIO (INTAKE, PULL CONFIGURATION)
REAR: 1 x 120mm fan (EXHAUST)

AIO EXHAUSTS
FRONT/SIDE: 240mm GPU AIO (EXHAUST, PUSH CONFIGURATION)
TOP: 360mm CPU AIO (EXHAUST, PUSH CONFIGURATION)
BOTTOM: 3 x 120mm fans (INTAKE)
REAR: 1 x 120mm fan (EXHAUST/INTAKE?)

I figure the AIO INTAKE configuration maximizes cool airflow over the RADS but dumps a lot of hot air into the case. The AIO EXHAUST confgiuration keeps the ambient temperature in the case cooler, but will run slightly warmer air over both RADS and may have issues with airflow (3/4 intake vs. 5/6 exhaust, depending on rear fan). I am interested especially because these principals will help instruct how I want to set up the eventual custom loop- I have room for 3 x 360mm RADS (2 x regular, 1 x slim; top, bottom, and side respectively) and a proper understanding of the thermodynamics involved here will help inform that build.

I will consider alternative suggestions to these, but the pipe length on the AIOs + the pump orientation may be a limiting factor. Pump for CPU is on the block, Pump for the GPU is in the RAD.

PoizenJam fucked around with this message at 17:59 on Jul 7, 2022

Adbot
ADBOT LOVES YOU

VelociBacon
Dec 8, 2009

Sorry I'm on my phone but I would try to get the GPU AIO in exhaust config on the top and the cpu AIO to intake config on the front. Use the bottom fans for intake and keep the exhaust fan. Scale the intake fans on the bottom and that rear exhaust fan to a thermal probe for case temp and scale the AIO rad fans to their respective coolant temps if you have that information accessible. Leave the pump speeds on the aios around 50% imo if you don't feel like playing with that.

I suggest finding the audible floor of your fans and setting all of them to just below that level at baseline.

My reasoning for the positioning is the tdp of the GPU and therefore the energy it is turning into heat is better dispersed directly out of the case than the cpu heat load, which is less.

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
Man there's a lot of shade being thrown at ek over at TPU comments, is everyone angry at ek or is everyone jealous that they can't afford ek products? They can't really be that bad, they have a good reputation I thought. Some people swear they wouldn't buy anything else.

https://www.techpowerup.com/294095/ekwb-lays-off-25-of-workforce-blames-lower-watercooling-sales

PoizenJam
Dec 2, 2006

Damn!!!
It's PoizenJam!!!

VelociBacon posted:

Sorry I'm on my phone but I would try to get the GPU AIO in exhaust config on the top and the cpu AIO to intake config on the front. Use the bottom fans for intake and keep the exhaust fan. Scale the intake fans on the bottom and that rear exhaust fan to a thermal probe for case temp and scale the AIO rad fans to their respective coolant temps if you have that information accessible. Leave the pump speeds on the aios around 50% imo if you don't feel like playing with that.

I suggest finding the audible floor of your fans and setting all of them to just below that level at baseline.

My reasoning for the positioning is the tdp of the GPU and therefore the energy it is turning into heat is better dispersed directly out of the case than the cpu heat load, which is less.

I agree with parts of this in principle, but placing the GPU at the top of the case is not (physically) possible with my current setup, and would also place the pump at the extreme top of the GPU AIO loop (i.e. the pump is in the RAD, not the water block). But also, isn't dumping the warmed air from the CPU RAD into the way-warmer GPU RAD going to just decrease the effectiveness of that GPU AIO?

As for the fan control you suggest, I use FanControl. I have done something conceptually similar to what you suggest already. I link the RAD fans to Temp of their respective device; the case fans to the averaged temp across several internal components. I also link the AIO pump speeds to the time-averaged temp of their devices, but in those cases I've used a step function (i.e. they'll typically kick up to 100% pump speed under heavy load where the sound is masked by the fans anyway).

PoizenJam fucked around with this message at 20:18 on Jul 7, 2022

SpartanIvy
May 18, 2007
Hair Elf

Nolgthorn posted:

Man there's a lot of shade being thrown at ek over at TPU comments, is everyone angry at ek or is everyone jealous that they can't afford ek products? They can't really be that bad, they have a good reputation I thought. Some people swear they wouldn't buy anything else.

https://www.techpowerup.com/294095/ekwb-lays-off-25-of-workforce-blames-lower-watercooling-sales

From my surface level engagement of the water cooling community it's that they can't afford them but spin it like they don't want to buy them in the first place.

EKWB makes mostly high quality and very nice looking hardware but its functionality is usually identical to competitors that are down to half the price for some things.

It's a premium option, for sure, and the have-nots get real mad about it existing.

When I put together my system with a mix of parts, the EKWB parts and fittings ~just worked~ which was nice. If I had to do it all again I'd probably go all in with EKWB.

VelociBacon
Dec 8, 2009

PoizenJam posted:

I agree with parts of this in principle, but placing the GPU at the top of the case is not (physically) possible with my current setup, and would also place the pump at the extreme top of the GPU AIO loop (i.e. the pump is in the RAD, not the water block). But also, isn't dumping the warmed air from the CPU RAD into the way-warmer GPU RAD going to just decrease the effectiveness of that GPU AIO?

As for the fan control you suggest, I use FanControl. I have done something conceptually similar to what you suggest already. I link the RAD fans to Temp of their respective device; the case fans to the averaged temp across several internal components. I also link the AIO pump speeds to the time-averaged temp of their devices, but in those cases I've used a step function (i.e. they'll typically kick up to 100% pump speed under heavy load where the sound is masked by the fans anyway).

Ah sorry I missed the part about the pump being on the rad. And yeah it's not ideal but with two radiators and that configuration you can't do all of them as exhausts. I tried! See my posts earlier in the thread.

I'm not convinced that pump speed helps cooling and I believe there is actually a performance falloff at both ends - too slow and too fast. You might want to try not scaling your pump speed and see if you get better results. I'm nearly certain that 100% pump speeds are worse than 80% for example and that somewhere is a sweet spot from 50-80%.

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
I'm hesitating a purchase for 3 days now all because this one 80 cent item in my cart is out of stock https://www.ekwb.com/shop/ek-hdc-fitting-14mm-o-ring-6pcs I probably don't need spare o-rings for my first build right?

VelociBacon
Dec 8, 2009

Nolgthorn posted:

I'm hesitating a purchase for 3 days now all because this one 80 cent item in my cart is out of stock https://www.ekwb.com/shop/ek-hdc-fitting-14mm-o-ring-6pcs I probably don't need spare o-rings for my first build right?

I didn't buy any. An o ring is one of very few items I would be happy getting from a hardware store with a real one in hand to compare and make sure it's the same dimensions.

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
I mean I'm not going to damage an o-ring, all I have to do is not damage one. How hard could it be?

VelociBacon
Dec 8, 2009

Nolgthorn posted:

I mean I'm not going to damage an o-ring, all I have to do is not damage one. How hard could it be?

Honestly I used soft tubing and I don't think I ever even saw an o ring.

Theophany
Jul 22, 2014

SUCCHIAMI IL MIO CAZZO DA DIETRO, RANA RAGAZZO



2022 FIA Formula 1 WDC

Nolgthorn posted:

I'm hesitating a purchase for 3 days now all because this one 80 cent item in my cart is out of stock https://www.ekwb.com/shop/ek-hdc-fitting-14mm-o-ring-6pcs I probably don't need spare o-rings for my first build right?

I made a real hash job when I did hardline tubing and didn't need any spare o-rings. Just make sure you sufficiently deburr your tubes before fitting them.

I would give my opinion on the anti-EKWB comments but I feel like I'm solidly out there as a big fan of their products and my opinion should be treated with equal weight. As SpartanIvy said, their stuff just works. Is it any more performant than their main competitors given the higher price? Debatable, I'd even argue not justifiably so on that metric alone. Does their stuff have a far more premium feel? I definitely think so. The worst part of the internet is every dickhead and their dog can make comments and I see a minority of comments moaning about negative user experiences with EKWB and far more being pissy about the price of it, which I kind of understand for prospective customers outside of Europe where they probably eat a bunch of taxes on them.

That being said I don't post on motoring forums making GBS threads on a recall Bentley issued for a car I can't afford. :shrug:

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
My experience with a lot of companies is that when something goes out of stock they never know when it'll be in stock so I'm going to go ahead and make the purchase without spare parts. Oh baby Jesus. Who bought all the o-rings, are they clumsy or something?

VelociBacon
Dec 8, 2009

Nolgthorn posted:

My experience with a lot of companies is that when something goes out of stock they never know when it'll be in stock so I'm going to go ahead and make the purchase without spare parts. Oh baby Jesus. Who bought all the o-rings, are they clumsy or something?

The spare parts you SHOULD buy are a few fittings, maybe 1 extra 90 and an extra 45 in addition to an extra straight fitting. Also, figure out how much coolant you'll need in your loop because I needed more than I ordered. Whoops.

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
I got two extra 90's, I'm only planning to use the two 45's. I shouldn't need more than a liter, it's just a 200 height pumpres and a reasonably short loop with 14mm tubes... I have a week off work coming up and my hope is that I can be back after taking my computer apart.

I bought extra thermal paste so in a pinch I'd be able to stick the air cooler back on.

Nolgthorn fucked around with this message at 23:53 on Jul 7, 2022

Indiana_Krom
Jun 18, 2007
Net Slacker

IMO if you have two radiators they should either both be on intake or both be on exhaust, I experimented with my own two radiators and both on intake kept my coolant about 1-2C lower at load compared to various intake/exhaust/mixed configurations. I settled on both as intake because my case has filters for the locations where I have my radiators mounted so I could minimize dust intake by making my case positive pressure and drawing in all the air passing over my radiators from outside. As a result my case has 5 fans on intake (280MM + 420MM radiators) with only a single fan in the back on exhaust that barely needs to exist at all because the positive pressure is so high. But it works and quite well at that, when my system is at load you can easily feel the heat blowing out the single exhaust fan in the back.

VelociBacon
Dec 8, 2009

Indiana_Krom posted:

IMO if you have two radiators they should either both be on intake or both be on exhaust, I experimented with my own two radiators and both on intake kept my coolant about 1-2C lower at load compared to various intake/exhaust/mixed configurations. I settled on both as intake because my case has filters for the locations where I have my radiators mounted so I could minimize dust intake by making my case positive pressure and drawing in all the air passing over my radiators from outside. As a result my case has 5 fans on intake (280MM + 420MM radiators) with only a single fan in the back on exhaust that barely needs to exist at all because the positive pressure is so high. But it works and quite well at that, when my system is at load you can easily feel the heat blowing out the single exhaust fan in the back.

Have you looked at your temps for other components in the PC? When I had both radiators drawing air in my HDDs were getting into the 50C+ area. Yes I have a few local HDDs in addition to a variety of other storage.

Indiana_Krom
Jun 18, 2007
Net Slacker
I have 3 local HDDs which usually hang around in the 35-45C range, it helps that I have them set to spin down after only 3 minutes of inactivity so for the most part they just don't spin or get hot. Most of my file activity happens on a couple 1 TB SSDs that hang around 30C. Airflow from one of my radiators goes right over the HDDs which is more than sufficient to keep them at a safe/optimal operating temperature. And basically everything else is serviced by the radiators, my motherboards VRM is overbuilt with LN2 cooling in mind so it is totally fine with just the airflow from the overhead radiator going over its heatsinks and then getting exhausted out the back.

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
Some info about what I'll be doing in a couple weeks.

I'm going to watercool a Threadripper 3970x, the primary reason is that I'm tired of my computer doing it's best impression of a jet engine. I tried to make it as practical as possible but I found that watercooling is still somewhat impractical honestly. The parts are too expensive, there's maintenance involved, there's too much to learn, this is a totally new sphere of knowledge well beyond what the average consumer is going to take on. There's got to be some kind of shift in the industry coming up, maybe that's all AIO's, maybe that's completely self contained computers that you can't tinker with at all. I dunno.

I also found cases are terrible for watercooling. Even ones that devote a lot of information in their manual to watercooling, I am using a Meshify 2 and I want the radiator on the top. There's not any space up there! It'll fit, but barely and only because I happened to buy 33mm tall ram.

EK-Quantum Magnitude sTRX4 D-RGB - Nickel + Acetal
EK-Quantum Kinetic TBE 200 D5 PWM D-RGB - Acetal
EK-Quantum Surface P360M - Black

I'm using 14mm acrylic and I'm going to attach the pump directly to the bottom of my case using a 120mm vertical bracket, there's 4 helpful little holes there. Currently the case has two 140mm intake fans on the front and a 140mm out. I'm thinking to remove part of the floor of the case in order to move the fan from the back of the case to the front, in order to maintain positive pressure, but I'll cross that bridge possibly in the build.

I've also bought a Phobya inline temperature sensor. I dunno how this is going to work but apparently I stick it in the loop somewhere and plug it into the motherboard at the appropriate location. It'll monitor my coolant and then I can use that to control the fan speed. This is going to cause me a headache probably, I can feel it coming. I bet not having anything plugged into the cpu fan header for example is going to make my computer look at me cross eyed.

But that's what I'm doing, I feel like I read way too much and I feel like a lot can go wrong even though I'm keeping it relatively simple.

Theophany
Jul 22, 2014

SUCCHIAMI IL MIO CAZZO DA DIETRO, RANA RAGAZZO



2022 FIA Formula 1 WDC
Plug the pump into the CPU fan header? I would imagine without something plugged into that header the system either won't POST or BIOS will constantly moan at you.

For short, bursty loads it will be quieter and just as efficient to have the pump speed increase as you'll be passing a greater volume of coolant over heat-generating components, then you can set the fans to spin up to a higher RPM once you hit a certain temperature that it will usually only reach under sustained loads. For example, up to 70'c on my 3900X the fans remain at around 650RPM whilst the pump gradually increases in speed, this means that the fans only ever become noticeably audible when there's a heavy sustained load like a video encode or large file export.

Coolant temperature sensors are a good failsafe if you don't have a flow meter, but I wouldn't rely on it for setting fan curves to.

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
I'm not opposed to disregarding the temp sensor entirely, should I use one? I'm pretty sure my computer will scream at me if the cpu gets hot regardless so I'll know there's a problem.

Indiana_Krom
Jun 18, 2007
Net Slacker
Setting your fan curves off the coolant temperature is exactly how you want to run a water cooler. The important bit is to set the ramp up appropriately; allow the coolant to reach an operating temperature before you start to increase the fan speeds so at normal room temperatures roughly 36C is a reasonable point to begin the ramp. Basically anything less than 10C over ambient is too aggressive and will just make the fans noisy for little benefit, coolers won't seriously dissipate heat until they are a fair amount warmer than the surrounding air and how efficiently they do so depends on how big the temperature difference is. My own fans "curve" is a flat 35% (~500 RPM, completely silent) until the coolant reaches 36C and then it is a straight line to 100% at 50C, the thermal mass of the coolant keeps the fan speeds from rising or falling rapidly and generally even with 500W of load the coolant settles at about 40C with the fans only reaching 50-60%.

Also the vast majority of BIOS implementations have a "CPU fan warning" option that disables the check on the CPU fan and can run perfectly fine without anything plugged in to that header.

Indiana_Krom fucked around with this message at 12:06 on Jul 8, 2022

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
Helpful information thank you.

PoizenJam
Dec 2, 2006

Damn!!!
It's PoizenJam!!!

Indiana_Krom posted:

IMO if you have two radiators they should either both be on intake or both be on exhaust, I experimented with my own two radiators and both on intake kept my coolant about 1-2C lower at load compared to various intake/exhaust/mixed configurations. I settled on both as intake because my case has filters for the locations where I have my radiators mounted so I could minimize dust intake by making my case positive pressure and drawing in all the air passing over my radiators from outside. As a result my case has 5 fans on intake (280MM + 420MM radiators) with only a single fan in the back on exhaust that barely needs to exist at all because the positive pressure is so high. But it works and quite well at that, when my system is at load you can easily feel the heat blowing out the single exhaust fan in the back.

So last night I decided to test the dual-intake config. Rather than move the AIO RAD as initially planned, I simply flipped the fans on my CPU AIO RAD into an intake pull config, then flipped the 3x120mm FRONT/SIDE fans to be exhaust, for the following config:

FRONT/SIDE: 3 x 120mm case fans (EXHAUST)
TOP: 360mm CPU AIO RAD + FANS (INTAKE, PULL CONFIGURATION)
BOTTOM: 240mm GPU AIO RAD + FANS (INTAKE, PULL CONFIGURATION)
REAR: 1 x 120mm case fan (EXHAUST)

This is slightly sub-optimal (compared to moving the rad to the FRONT-SIDE position), if only because I am fighting convection with those top intake fans. However, I noticed an immediate decrease of 6-10C on the CPU at idle, depending on the load of the GPU. At load, though, there doesn't appear to be a huge difference compared to my original config (GPU AIO Intake, CPU AIO Exhaust). The temperatures of interior components in the case (MOBO, PCH, SSD) increased across the board as both the CPU and GPU are now dumping hot air in the case. I'm not particularly worried, since MOBO is still safe and NVME run just fine at the temps. The PCH seems to run hotter in this config though... About 82C. Which is concerning, but maybe not a problem.

I could flip the FRONT/SIDE case fans back to INTAKE for extreme positive pressure (8 x 120mm intakes total vs. 1 x 120mm exhaust), and to add some fresh, non-heated air to the mobo. But I always thought a balanced or slightly positive pressure build was best?

Edit: So after analyzing this further, it seems that this configuration is worse in pretty much every way except for the CPU idle temps. GPU is ~2-5C warmer in most use conditions, and my PCH temps climbed up to >90C after a soak in period. In my old config (GPU AIO INTAKE + CPU AIO EXHAUST), those temps were low 70s. Ram feels pretty hot too. Best guess is dumping so much hot air into the case is particularly bad after a 'soak' period. I'm going to try an exhaust configuration next.

PoizenJam fucked around with this message at 16:34 on Jul 8, 2022

AutismVaccine
Feb 26, 2017


SPECIAL NEEDS
SQUAD

SpartanIvy posted:

From my surface level engagement of the water cooling community it's that they can't afford them but spin it like they don't want to buy them in the first place.

EKWB makes mostly high quality and very nice looking hardware but its functionality is usually identical to competitors that are down to half the price for some things.

It's a premium option, for sure, and the have-nots get real mad about it existing.

When I put together my system with a mix of parts, the EKWB parts and fittings ~just worked~ which was nice. If I had to do it all again I'd probably go all in with EKWB.

Imo it is complicated, cause

-they dont really have the best blocks and are expensive
-not many radiators to choose from (no good low fpi stuff)
-if you use no rotatable fittings they really dont matter and i dont see the value in a quantum fitting for 8€ a piece

IMO if you are serious about watercooling it is really hard to buy only EK stuff

AutismVaccine
Feb 26, 2017


SPECIAL NEEDS
SQUAD

PoizenJam posted:

MOBO is still safe and NVME run just fine at the temps. The PCH seems to run hotter in this config though... About 82C.

just a small question: Is your NVMe ssd installed at the right slot? On modern Mobos the ssd has a direct link to the CPU and the PCH shouldnt really do anything (or you have lots of other stuff installed?)

edit: my chipset is at 60 at almost no airflow

AutismVaccine fucked around with this message at 18:47 on Jul 8, 2022

Theophany
Jul 22, 2014

SUCCHIAMI IL MIO CAZZO DA DIETRO, RANA RAGAZZO



2022 FIA Formula 1 WDC

AutismVaccine posted:

Imo it is complicated, cause

-they dont really have the best blocks and are expensive
-not many radiators to choose from (no good low fpi stuff)
-if you use no rotatable fittings they really dont matter and i dont see the value in a quantum fitting for 8€ a piece

IMO if you are serious about watercooling it is really hard to buy only EK stuff

I was with you until the no true scotsman poo poo at the end. You can take something seriously and not feel compelled to min/max if your needs are straightforward.

AutismVaccine
Feb 26, 2017


SPECIAL NEEDS
SQUAD

Charles Leclerc posted:

I was with you until the no true scotsman poo poo at the end. You can take something seriously and not feel compelled to min/max if your needs are straightforward.

yeah agreed mate, i just saw the min max side. No hate against all EK setups ofc.

Theophany
Jul 22, 2014

SUCCHIAMI IL MIO CAZZO DA DIETRO, RANA RAGAZZO



2022 FIA Formula 1 WDC
No sweat, I do have a box full of those 45 and 90 degree swivel fittings because I'm an absolute sucker for them lol

PoizenJam
Dec 2, 2006

Damn!!!
It's PoizenJam!!!

AutismVaccine posted:

just a small question: Is your NVMe ssd installed at the right slot? On modern Mobos the ssd has a direct link to the CPU and the PCH shouldnt really do anything (or you have lots of other stuff installed?)

edit: my chipset is at 60 at almost no airflow

I'm running:

Lian Li O11D XL Case
ASUS ROG z690 Hero
Intel 12900k
ASUS RTX 3080 (Uses the
2 x 32GB DDR6 @ 5600Mhz
3 x NVME 1TB SSD (1x PCIE4 x4 via Processor, 1x PCIE4 x4 via Chipser, 1x PCIE3 x4 via Chipset)
2 x SATA SSD (Rear drive mounts)
1 x PCI USB HUB (provides additional 5 x USB + 2 x USB C ports; installed on the Z690 Chipset's expansion slot)
FULL PARTS LIST FOR THE CURIOUS

I also have most of my USB ports populated by things like capture cards, audio interfaces, midi controllers, etc. I use this build for gaming, streaming, work, AI/Big Data, and audio/visual production; so just about every use case you can imagine. So there's a lot of hot stuff inside my case and I really want to get this right- especially since it's in my room with my audio recording equipment and I like it to be as quiet as possible.

None of the 'all AIO intake' configs seem promising in this particular use case, and it's likely that every config short of a custom loop with shared CPU/GPU coolant + RADS is going to involve some sacrifice here.

VelociBacon
Dec 8, 2009

PoizenJam posted:

So last night I decided to test the dual-intake config. Rather than move the AIO RAD as initially planned, I simply flipped the fans on my CPU AIO RAD into an intake pull config, then flipped the 3x120mm FRONT/SIDE fans to be exhaust, for the following config:

FRONT/SIDE: 3 x 120mm case fans (EXHAUST)
TOP: 360mm CPU AIO RAD + FANS (INTAKE, PULL CONFIGURATION)
BOTTOM: 240mm GPU AIO RAD + FANS (INTAKE, PULL CONFIGURATION)
REAR: 1 x 120mm case fan (EXHAUST)

This is slightly sub-optimal (compared to moving the rad to the FRONT-SIDE position), if only because I am fighting convection with those top intake fans. However, I noticed an immediate decrease of 6-10C on the CPU at idle, depending on the load of the GPU. At load, though, there doesn't appear to be a huge difference compared to my original config (GPU AIO Intake, CPU AIO Exhaust). The temperatures of interior components in the case (MOBO, PCH, SSD) increased across the board as both the CPU and GPU are now dumping hot air in the case. I'm not particularly worried, since MOBO is still safe and NVME run just fine at the temps. The PCH seems to run hotter in this config though... About 82C. Which is concerning, but maybe not a problem.

I could flip the FRONT/SIDE case fans back to INTAKE for extreme positive pressure (8 x 120mm intakes total vs. 1 x 120mm exhaust), and to add some fresh, non-heated air to the mobo. But I always thought a balanced or slightly positive pressure build was best?

Edit: So after analyzing this further, it seems that this configuration is worse in pretty much every way except for the CPU idle temps. GPU is ~2-5C warmer in most use conditions, and my PCH temps climbed up to >90C after a soak in period. In my old config (GPU AIO INTAKE + CPU AIO EXHAUST), those temps were low 70s. Ram feels pretty hot too. Best guess is dumping so much hot air into the case is particularly bad after a 'soak' period. I'm going to try an exhaust configuration next.

Are you running 100% pump speed though? I still maintain that's a mistake and I wouldn't test anything else while that variable is set. Sorry if I'm misremembering.

PoizenJam
Dec 2, 2006

Damn!!!
It's PoizenJam!!!

VelociBacon posted:

Are you running 100% pump speed though? I still maintain that's a mistake and I wouldn't test anything else while that variable is set. Sorry if I'm misremembering.

I did a lot of experimenting with this a while back, and I am less concerned with that aspect of my setup currently.

However, I have confirmed that the cooling performance of both AIOs improves *very slightly* as the pump speed increases in both cases. About 2-3C for 50% vs 100% in both cases. The trend is linear in both cases, and I did not identify a sweet spot for temperature between the minimum and maximum. What I did notice was that the pump ramping up/down made the noise more noticeable, and that there were some speeds that were louder than others (CPU AIO was loudest at 65-80%, for instance). Changing pump speed a lot is also potentially bad.

So what to do if pump speed improves cooling but you don't want it ramping it up and down all the time? I used FanControl.

I set the pump speed to track the average temperature of the CPU and GPU, respectively. I combine that with a step function, ramping the pump speed up or down at critical thresholds, always setting it as high as it can go without being louder than the fans at the same temp. This kept the pump speeds mostly stable rather than constantly ramping up or down, but still allowed me to increase it as load demanded.


Anyway, an update on my dual AIO setup... I spent almost all of today playing around with different orientations. Almost all of them came upsides or downsides, pushing the temps of the CPU, GPU, and other components up or down a degree or two, usually trading off. Some came with big downsides (all intake baked the internals pretty bad! 600W of heat at max load!). I ended up reverting to my original layout, because the airflow + Pressure setup was best in that case, even if the CPU temps were a bit higher. Still throttling on Cinebench R23 occasionally. But this was true in every configuration I tried though. So I probably won't get that completely under control unless use a less aggressive overclock or move to a custom loop. I think I lost the silicon lottery (bottom 2.5% according to SP), and my voltages are kind of high.

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
Hey dudes my first water-cooled loop, even though I haven't finished all of the runs yet and rgb isn't plugged in I couldn't wait to show it off.

Nolgthorn fucked around with this message at 15:45 on Jul 12, 2022

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense

Charles Leclerc posted:

Plug the pump into the CPU fan header? I would imagine without something plugged into that header the system either won't POST or BIOS will constantly moan at you.

For short, bursty loads it will be quieter and just as efficient to have the pump speed increase as you'll be passing a greater volume of coolant over heat-generating components, then you can set the fans to spin up to a higher RPM once you hit a certain temperature that it will usually only reach under sustained loads. For example, up to 70'c on my 3900X the fans remain at around 650RPM whilst the pump gradually increases in speed, this means that the fans only ever become noticeably audible when there's a heavy sustained load like a video encode or large file export.

Coolant temperature sensors are a good failsafe if you don't have a flow meter, but I wouldn't rely on it for setting fan curves to.

I'm realizing I don't think my motherboard has a header for the temperature sensor.

Motherboard manual.pdf

What should I do? I think I need a specialized fan controller that accepts a temperature reading, but we're entering a realm of customization with this thing that makes me uncomfortable. What would I need to buy?


E:

It's going in the loop either way because it makes this 45 clear the fan.

Nolgthorn fucked around with this message at 20:18 on Jul 18, 2022

VelociBacon
Dec 8, 2009

Nolgthorn posted:

I'm realizing I don't think my motherboard has a header for the temperature sensor.

Motherboard manual.pdf

What should I do? I think I need a specialized fan controller that accepts a temperature reading, but we're entering a realm of customization with this thing that makes me uncomfortable. What would I need to buy?


E:

It's going in the loop either way because it makes this 45 clear the fan.



It's so lovely and 2022 that your mobo has multiple RBG points but no header for temperature probes. I tried googling around to see if there was a PCI-E or USB solution but couldn't find one. I would agree that you don't really want to go to a third party temp monitor/fan speed controller if you can help it.

Maybe there's a non-2 pin style inline temp probes you can use. Not sure? Hoping someone else here knows of an elegant solution.

e: Is this a NEW build or are you adapting water cooling to your existing system? If it's new I'd honestly return that mobo and buy one with temp headers. If you can't return it, I'd sell it at a small loss and buy a mobo with the right headers for what you're doing here.

If it's a system you're already using that does make it more of a headache.

VelociBacon fucked around with this message at 20:34 on Jul 18, 2022

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
Nah nobody here has threadripper, it was difficult to even get the mobo shipped. I've been using it about a year.

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
Seems like what I can find is Corsair iCUE. That uses proprietary fan headers so I'd have to buy new fans, I don't really want to do that.

E:

Aqua Computer Quadro - There's this thing, maybe I could hook it up and use the software, maybe?

Nolgthorn fucked around with this message at 21:08 on Jul 18, 2022

Warmachine
Jan 30, 2012



Nolgthorn posted:

Seems like what I can find is Corsair iCUE. That uses proprietary fan headers so I'd have to buy new fans, I don't really want to do that.

E:

Aqua Computer Quadro - There's this thing, maybe I could hook it up and use the software, maybe?

Yes. the Aqua Computer Quadro/Octo are probably your best bet. The controller software is very nice, and you don't need to run said software for the device to function once configured unless you're doing some more advanced virtual sensor stuff. But expect to pay out the nose for them. They're made in one place and pretty much no one stocks them in the US.

Fun fact: The Octo provides 2-amp PWM headers, which is beyond the spec of most motherboard headers and allowed me to safely run a DDC pump off the header alone with no need for a separate molex connection. :science:

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense

Warmachine posted:

Fun fact: The Octo provides 2-amp PWM headers, which is beyond the spec of most motherboard headers and allowed me to safely run a DDC pump off the header alone with no need for a separate molex connection. :science:

You are meddling with powers I can not possibly comprehend. Luckily I'm in Europe and I can buy one for 65 bones shipped. I think I'm going to get one... whatever, all this stuff cost so much money already. I mostly don't want to do it because I'm fighting against the system.

I'm giving my computer more points of failure and more things to pay attention to. I wanted to just use the motherboard and motherboard software, but it's lacking.

Nolgthorn
Jan 30, 2001

The pendulum of the mind alternates between sense and nonsense
After buying that and a rgb hub and a combo square-ruler tool, and the heat gun and the fans and the fittings and the dang distilled water and on and on I'm tapped out. If I have to spend 1 more dollar I will toss it all out!

Peek at my assembled water pump.




Top inlet, 45 degree out, drain port, on a vertical bracket anchored to the bottom of the case.

Adbot
ADBOT LOVES YOU

movax
Aug 30, 2008

Warmachine posted:

Yes. the Aqua Computer Quadro/Octo are probably your best bet. The controller software is very nice, and you don't need to run said software for the device to function once configured unless you're doing some more advanced virtual sensor stuff. But expect to pay out the nose for them. They're made in one place and pretty much no one stocks them in the US.

Fun fact: The Octo provides 2-amp PWM headers, which is beyond the spec of most motherboard headers and allowed me to safely run a DDC pump off the header alone with no need for a separate molex connection. :science:

The OCTO has been pretty sweet so far, though I have not had time to scratch the surface in configuring an actual control loop -- I just have the fans set to static speeds (Noctua A12s at ~65%) and my 3960X and 1080 do quite well. I could probably optimize so it gets quieter / cooler under high loads, but I max at 71 C under Prime95 and hit my PPT/CPU/EDC limits first with a mild overclock.

Pleasantly surprised at its ability to get data from other things in the system + appear to be pretty stable.

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply