Jump to content
Corsair Community

Upgraded Watercooling Loop to a Dual Radiator Setup, but temps are now hotter?


Recommended Posts

Hi everyone,

I hope I'm posting in the right place. I need some help regarding my most recent PC build.

 

Previously my watercooling setup consisted of a single EK SE 420x28mm radiator in the top of my Phanteks Enthoo Luxe case with 3 ML140 Pro fans, an EK XRES 140 D5 Revo Pump/Res combo, EK Supremacy RGB CPU block, Phanteks Glacier GPU Block and soft tubing. I was cooling my 8700k (with conductonaut liquid metal between the IHS and die) at 4.8Ghz 1.232V load voltage and my Strix GTX1080Ti in it's out of the box OC mode. Here is a picture of the old system:

 

JZX5s2n.jpg?1

 

For my new build I moved over the core components into the Lian Li PC-011 Dynamic case, however I switched from my old ROG Maximus X Hero to a ROG Maximus X Code, extended the reservoir, used rigid tubing and this time I'm using 2 radiators; an EK XE 360x60mm in the top and an XSPC EX360 360x35mm in the back as an intake. Both radiators are fitted with ML120 Pro RGB fans. Here is a picture of the new system:

 

4BFxokr.jpg?1

 

While I am very pleased with the build aesthetically, it seems for some reason my temperatures are HIGHER than before...

 

Previously, after applying liquid metal to the CPU, in AIDA64 FPU Stress Test after about 15mins the CPU temp would be leveled out at 41-42C with the cores in the low to mid 50s. In 3DMark Time Spy Stress Test the GPU would stay around 50C. Both tests on my old system were running the fans at 1500ish RPM and the water pump at full speed.

 

Now, in my new setup, in AIDA64 FPU Stress Test after 15mins the CPU temp is at 46C and still slowly rising. In 3DMark Time Spy Stress Test the GPU is hitting 55C and hovering around there. Both tests on the new system were running the fans at 1500-1600RPM and the water pump at full speed.

 

Here are the AIDA64 screenshots.

 

Old system:

 

mxmKtcV.png

 

New System:

 

CTExE7n.png

 

To me this makes absolutely no sense? I should have much better cooling performance now that I have far more total radiator surface area and thickness. The only other thing I changed was that I used Kryonaut thermal paste this time instead of Hydronaut, but Kryonaut is supposed to be better. I know I applied it properly as well.

 

There are still some little bubbles in the CPU block and sometimes I can hear what sounds like maybe little bubbles moving around in the top radiator (EK XE 360), but can those things really make that much of a difference?

 

I really need some help figuring this out. I am so happy with how the build turned out aesthetically but it's really bringing me down that all that work is leading to higher temps.. :(

Link to comment
Share on other sites

Readjust your CPU Block - Maybe it doesnt fits right. Or the screws are not tight enough...

 

Hey thanks for the reply.

 

I have done this and all the thumb screws are secure. I am confident that there is no issue with the mounting of the CPU block.

Link to comment
Share on other sites

I don't think you have a contact or any other hardware issue. If anything, the core temps are even tighter together and you did a good job. Whenever you see a slow and steady climb of temperature sensors, you are probably witnessing a general increase in local temperature. This could be solely a coolant temp rise (and thus fan or cooling system issue), but I suspect it is case ambient which then in turn affects everything in it.

 

The differences in two sets of graphs are clear. Original run is nearly flat with a very slight rise over the 16 minutes. Would it stay flat for 3 hours? That's another story and not relevant for now. In 011 graph you can see a slow rise of in the 0.5-1.0C per minute range across the board - cores, package, and drives. Of particular interest is the motherboard sensor on the Code and its clear increase in the 011 system. You argue the Code has it in a different place than the Hero (it doesn't by the way), but I think the clear indication is your interior case temperature is rising at a noticeable rate.

 

There are probably several combining factors leading to the differences between the Enthoo and the 011. 1) Case construction and materials; 2) Your change in layout, fan and radiator positioning; 3) Most notably the air pathway design of the 011. I have had an eye on this case for some time and I naturally drawn to non-standard designs. It is interesting, but I have always been unsure how it would fare in a less than ideal configuration (or even when optimized). It is also possible there was a small room temperature difference between tests, which you would also need to factor in. I used to tag all my AIDA bench tests with room temp data and time of day, mostly so I could labor over them for a ridiculously long amount of time. The room temp helps keep me from making huge assumptions about differences in performance, when that 2C difference was all atmosphere.

 

Based on your description and what I can see, the 011 set-up has a top mounted 360mm as exhaust and the other radiator is "XSPC EX360 360x35mm in the back as an intake". I think you mean the vertical and 90 degree rotated position toward the "front" of the glass and the "back" (in a traditional sense) is the passively vented side on the left that you really can't do much with, except tack on a little 120mm. I would try and figure out is if the exhaust temp coming off the intake 360 and see how that compares to what happens to the motherboard temp sensor. I am expecting them to go step for step, +1C = +1C. You can use any tools you can from extra temp probes to a thermal gun to water in/out sensors if you have any. If you compare this to the single radiator Enthoo build, the top was exhausting all your heat directly out of the top with both active unrestricted exhaust and and clean air intake.

 

I don't think you have done anything wrong and at this moment I don't have an alternative design or clear change recommendation, but I will think on it. I really sympathize with you. I had a little whiny pump and 120mm AIO on my Titan X. While thermally effective, the irritation at idle finally reached a critical mass and I yanked it out and went back to a custom loop -- on the GPU only. $800 later I reduced my idle noise and GPU temps by 10C. Of course 50C vs 40C on the GPU wasn't very relevant. To make this work in my Air 740 with dual 280mm radiators, I needed to run the case in reverse. My back 140mm fan is intake, along with 2x140mm bottom. The CPU cooler exhaust out the top. The GPU HW Labs 280 out the front. It is fairly efficient and the quietest "at load" system I have ever built. However, I have been driving myself nuts for months over the difference in internal case temperature. My prior set-ups in both this case and in a 540 were so unbelievable efficient. Case internal temp change was +1C on most days. Maybe +2C with sustained GPU and CPU loads. Now I am running what feels like +4-5C higher. I believe it to be down to me going against the natural flow of the case. Where as I had substantial free intake air plus a clear pathway through exhaust and rear venting, now any extra intake case flow is trapped until I can exhaust it through either radiator. Really it's all ridiculous. I have 8700K at 5.0 GHz that can't break 60C and GPU running 2100 MHz under 40C and I can barely hear my fans over the ceiling fan and AC. Nevertheless, I am having the same feelings as you --- what have I done.

 

Anyway, I am happy to keep working on this as much as you like. I can't even tell you the number of actual hours I have spent trying to certify my own state of affairs with the reverse flow system. See if you can figure out how much the total cycle rate is affecting this temp rise. The quick and easy way? Run it again with the door off.

Link to comment
Share on other sites

I don't think you have a contact or any other hardware issue. If anything, the core temps are even tighter together and you did a good job. Whenever you see a slow and steady climb of temperature sensors, you are probably witnessing a general increase in local temperature. This could be solely a coolant temp rise (and thus fan or cooling system issue), but I suspect it is case ambient which then in turn affects everything in it.

 

The differences in two sets of graphs are clear. Original run is nearly flat with a very slight rise over the 16 minutes. Would it stay flat for 3 hours? That's another story and not relevant for now. In 011 graph you can see a slow rise of in the 0.5-1.0C per minute range across the board - cores, package, and drives. Of particular interest is the motherboard sensor on the Code and its clear increase in the 011 system. You argue the Code has it in a different place than the Hero (it doesn't by the way), but I think the clear indication is your interior case temperature is rising at a noticeable rate.

 

There are probably several combining factors leading to the differences between the Enthoo and the 011. 1) Case construction and materials; 2) Your change in layout, fan and radiator positioning; 3) Most notably the air pathway design of the 011. I have had an eye on this case for some time and I naturally drawn to non-standard designs. It is interesting, but I have always been unsure how it would fare in a less than ideal configuration (or even when optimized). It is also possible there was a small room temperature difference between tests, which you would also need to factor in. I used to tag all my AIDA bench tests with room temp data and time of day, mostly so I could labor over them for a ridiculously long amount of time. The room temp helps keep me from making huge assumptions about differences in performance, when that 2C difference was all atmosphere.

 

Based on your description and what I can see, the 011 set-up has a top mounted 360mm as exhaust and the other radiator is "XSPC EX360 360x35mm in the back as an intake". I think you mean the vertical and 90 degree rotated position toward the "front" of the glass and the "back" (in a traditional sense) is the passively vented side on the left that you really can't do much with, except tack on a little 120mm. I would try and figure out is if the exhaust temp coming off the intake 360 and see how that compares to what happens to the motherboard temp sensor. I am expecting them to go step for step, +1C = +1C. You can use any tools you can from extra temp probes to a thermal gun to water in/out sensors if you have any. If you compare this to the single radiator Enthoo build, the top was exhausting all your heat directly out of the top with both active unrestricted exhaust and and clean air intake.

 

I don't think you have done anything wrong and at this moment I don't have an alternative design or clear change recommendation, but I will think on it. I really sympathize with you. I had a little whiny pump and 120mm AIO on my Titan X. While thermally effective, the irritation at idle finally reached a critical mass and I yanked it out and went back to a custom loop -- on the GPU only. $800 later I reduced my idle noise and GPU temps by 10C. Of course 50C vs 40C on the GPU wasn't very relevant. To make this work in my Air 740 with dual 280mm radiators, I needed to run the case in reverse. My back 140mm fan is intake, along with 2x140mm bottom. The CPU cooler exhaust out the top. The GPU HW Labs 280 out the front. It is fairly efficient and the quietest "at load" system I have ever built. However, I have been driving myself nuts for months over the difference in internal case temperature. My prior set-ups in both this case and in a 540 were so unbelievable efficient. Case internal temp change was +1C on most days. Maybe +2C with sustained GPU and CPU loads. Now I am running what feels like +4-5C higher. I believe it to be down to me going against the natural flow of the case. Where as I had substantial free intake air plus a clear pathway through exhaust and rear venting, now any extra intake case flow is trapped until I can exhaust it through either radiator. Really it's all ridiculous. I have 8700K at 5.0 GHz that can't break 60C and GPU running 2100 MHz under 40C and I can barely hear my fans over the ceiling fan and AC. Nevertheless, I am having the same feelings as you --- what have I done.

 

Anyway, I am happy to keep working on this as much as you like. I can't even tell you the number of actual hours I have spent trying to certify my own state of affairs with the reverse flow system. See if you can figure out how much the total cycle rate is affecting this temp rise. The quick and easy way? Run it again with the door off.

 

Hey c-attack. Thanks for taking the time to write all this for me. I appreciate it. Although I don't recall saying anything about the temperature sensors on the Code being any different than the Hero?

 

You must have an awesome 8700k and cooling system. Mine needs 1.344V under load to sustain 5Ghz without an AVX offset and it will reach a little over 70C core temps (which is still fine but I'd be over the moon if they were still under 60 like yours).

 

I've posted on many other forums and a lot of people suggested switching all my fans to exhaust to avoid pulling hot air from one radiator through another. It really got to me and I stayed up until 4AM last night tearing the loop apart just to get access to the fans and turn them around..

 

Before I did that I ran this test with no dust filters in the case at a room temperature of about 18.5C. During this test the water temperature reached 25.5C:

 

1KbsosY.png

 

This morning I ran the test again with no dust filters and the fans all exhausting air at a room temperature of about 16.8C. During this test the water temperature reached 22.9C:

 

JliumAN.png

 

When taking into account the room temperature difference there is hardly any difference in the tests at all. This is incredibly frustrating after staying up until 4AM to make the change.. :(

 

I really hope there's just some air in there that needs to come out, but I left the pump running all night and tilted the case every which way to try and get the air out. I can see there are a few air bubbles still in the CPU block but not really any different from before.

 

Maybe I need to run the test for longer or under combined loads to see any difference? I don't know.. I really wish I took a room temperature measurement when I ran my stress test on my old system as that could be the culprit I guess..

 

I'll try a few other stress tests and look for improvement throughout the day but man.. this sucks.. I'm so stressed out. :(

Link to comment
Share on other sites

After some discussion on another forum I'm starting to think that the difference between my temps today and the temps on my prior system probably comes down to being at the point of diminishing returns for the CPU and margin of error/room temperature differences. I haven't run any long sustained loads on either system which I really should have done in the first place rather than these attempted controlled 15 minute tests. I'll try running some games and maybe some more overclocking and see how I go. The stress from staying up so late last night just to turn 3 fans around has been getting to me this morning..

 

Thanks for your help everyone.

Link to comment
Share on other sites

"Although I don't recall saying anything about the temperature sensors on the Code being any different than the Hero?"

 

You know both boards and so do I. That really wasn't for you. It was more of a preemptive argument for anyone else following along. Since I was basing a substantial part of my supposition on that measurement, it would make sense to be rock solid sure about it and that was my suggestion. However, I see you this taking over your mind and jumped in all the way with a reconfiguration. I understand. Every other day I wake up and try to talk myself in or out of redoing my loop another way, all for 2-3C of ambient case temp.

 

My 8700K is delidded, so that does give me a strong advantage at 1.30v. Running 15 minute AIDA tests is completely appropriate for what we are doing. That is exactly what I do as well. It's enough time for the coolant to level off and make hardware and voltage bases determinations. Once you get into 1 hour plus testing, the case design and component nature really have a stronger bias. Since those are things you cannot change, I usually don't test them, at least that way. 1 hour CPU only stress test is only useful if it resembles how you will use the system. For some people doing encoding or similar things, maybe. For me, the gaming is a more practical way to test long term performance and more fun than watching your graph for 1 hour.

 

It looks like the motherboard temp climb is slightly reduced, at least compared to the original test data up top. I am now wondering about your coolant temperatures. Going on the assumption that all slow changes in CPU (package) temp values are due to the rise in coolant temperature, your original run was +6-7C. I looks like 39C after it settles post 100% load on before climbing up to 46C at the 15 min mark. That is about what I would get at 5.0/1.30 on my H115i Pro (280mm). Not entirely favorable at 280mm vs 2x360. However, on both of you new tests, the post load climb looks more like +3C and maybe even +2 on the morning run. That is outstanding and appears to match the +3C (38-41C) from the original test in the old system. I have never run two 360s in a combo loop so I can't tell what is the expected delta for that exact configuration. However, +2-3 is about as low as I see anyone go without resorting to external mounting or standalone cooling devices. While you may not be feeling great about this, it is a 50% reduction of your coolant delta and a pretty substantial step. When I originally saw your first graph from the Enthoo system, I was immediately struck by how flat the CPU temp was down the line. +2C in the first 2-3 minutes, then rock solid -- the sign of a very efficient system. Your CPU package temp from this morning is just as flat. To me, that makes any differences remaining attributable to ambient temp differences and whatever small differences are inescapable with the glass box construction. I bet you could get another 1-2C if you went with an open frame chassis, but obviously that's not the point. There was always going to be some trade-off going this style of case. It looks fantastic.

Link to comment
Share on other sites

"Although I don't recall saying anything about the temperature sensors on the Code being any different than the Hero?"

 

You know both boards and so do I. That really wasn't for you. It was more of a preemptive argument for anyone else following along. Since I was basing a substantial part of my supposition on that measurement, it would make sense to be rock solid sure about it and that was my suggestion. However, I see you this taking over your mind and jumped in all the way with a reconfiguration. I understand. Every other day I wake up and try to talk myself in or out of redoing my loop another way, all for 2-3C of ambient case temp.

 

My 8700K is delidded, so that does give me a strong advantage at 1.30v. Running 15 minute AIDA tests is completely appropriate for what we are doing. That is exactly what I do as well. It's enough time for the coolant to level off and make hardware and voltage bases determinations. Once you get into 1 hour plus testing, the case design and component nature really have a stronger bias. Since those are things you cannot change, I usually don't test them, at least that way. 1 hour CPU only stress test is only useful if it resembles how you will use the system. For some people doing encoding or similar things, maybe. For me, the gaming is a more practical way to test long term performance and more fun than watching your graph for 1 hour.

 

It looks like the motherboard temp climb is slightly reduced, at least compared to the original test data up top. I am now wondering about your coolant temperatures. Going on the assumption that all slow changes in CPU (package) temp values are due to the rise in coolant temperature, your original run was +6-7C. I looks like 39C after it settles post 100% load on before climbing up to 46C at the 15 min mark. That is about what I would get at 5.0/1.30 on my H115i Pro (280mm). Not entirely favorable at 280mm vs 2x360. However, on both of you new tests, the post load climb looks more like +3C and maybe even +2 on the morning run. That is outstanding and appears to match the +3C (38-41C) from the original test in the old system. I have never run two 360s in a combo loop so I can't tell what is the expected delta for that exact configuration. However, +2-3 is about as low as I see anyone go without resorting to external mounting or standalone cooling devices. While you may not be feeling great about this, it is a 50% reduction of your coolant delta and a pretty substantial step. When I originally saw your first graph from the Enthoo system, I was immediately struck by how flat the CPU temp was down the line. +2C in the first 2-3 minutes, then rock solid -- the sign of a very efficient system. Your CPU package temp from this morning is just as flat. To me, that makes any differences remaining attributable to ambient temp differences and whatever small differences are inescapable with the glass box construction. I bet you could get another 1-2C if you went with an open frame chassis, but obviously that's not the point. There was always going to be some trade-off going this style of case. It looks fantastic.

 

Ah I understand what you meant now. Non problem. My CPU is still has the IHS but I've delided/relidded it with conductonaut liquid metal inbetween the die and the IHS. It has been that way since my old system.

 

Your cooling sounds incredible for a 280mm cooler! Maybe the fact that I am cooling my Strix 1080ti in the same loop has something to do with it, or maybe it's hotter here in Australia haha.

 

Yeah I do agree the graph is a lot flatter since changing the fan layout, where as before it was on a slow climb likely due to increasing temps inside my case. In my old system I had a constant feed of cool air into the (large) Phanteks case where as this one is much more compact and relying on either warm air intake through the slim rad or negative pressure intake through the vented floor and rear of the case.

 

I have done some gaming and my GPU temperatures have dropped a bit now which is good. Perhaps combined loads like this are where I will see more benefit from this dual radiator setup.

 

Last night I played around with some more overclocking and I'm currently at 5Ghz (AVX-1) @ 1.30-1.31V load voltage (1.350 in BIOS with LLC5) and the core temps did not go higher than around 62C during an hour of Prime95 26.6 small FFTs test. Planning on running AIDA64 and RealBench today.

 

Thanks again for your insight.

Link to comment
Share on other sites

I'm willing to bet it's likely an issue of the case perhaps recycling the air around it. It seems to me that since it's in a corner, a lot of that air being exhausted is finding its way behind the case since the back radiator is trying to pull air from somewhere, other than that I can't think of a reason that a case with seemingly way better airflow would raise temps.

 

Try moving the case over a couple feet temporarily from the corner and see if it helps.

Link to comment
Share on other sites

I'm willing to bet it's likely an issue of the case perhaps recycling the air around it. It seems to me that since it's in a corner, a lot of that air being exhausted is finding its way behind the case since the back radiator is trying to pull air from somewhere, other than that I can't think of a reason that a case with seemingly way better airflow would raise temps.

 

Try moving the case over a couple feet temporarily from the corner and see if it helps.

 

The fans are all set to exhaust now as I mentioned in one of the comments, so air is being drawn in through the vented floor and rear of the case via negative pressure. I also taped over the extra vented slots in the top to prevent air from the top radiator from being drawn back in. I have moved the PC as far over as I can from the wall and there's probably about 4.5-5 inches space.

 

Something I've noticed while running stress tests and playing around with the pump speeds is that when I reduce the pump speed to say 50% (2450RPM) from 100% (4900RPM), the water temperature reading drops a couple or so degrees very fast, but the component temperature (CPU/GPU) rises fairly quickly by a few degrees, and vice versa, when I turn the pump speed up back up to 100% the water temp reading goes back up a few degrees very quiclky but the component temps come down again.

 

At first I thought the drop in water temp was due to the pump putting it's own heat into the system at high speeds but as soon as I saw the corresponding changes in component temperatures I thought this couldn't be right.

 

I wonder if I have a dodgy temp sensor.. Or is this just the effects of better heat transfer to the water at higher speeds? The sensor is an XSPC G1/4 plug sensor hooked up to the Commander Pro. My fan speeds are being controlled based on it's readings. http://www.xs-pc.com/temperature-sensors/g14-plug-with-10k-sensor-black-chrome

Link to comment
Share on other sites

That's is a bit odd. I have the same XPSC 10K fitting in my system. I will try and replicate it later, although I suspect this might be a sensor location issue in combination with the surge from speed change. In the loop, is it in the lower center part of the fixed tubing, right after the pump on outflow? I can't tell flow direction from the picture.

 

**I can't recreate this. I have two XSPC 10K probes stuck into the ends of a false T valve entering and exiting the radiator. I normally run my much smaller loop at 2000, but shifting to 4000 does not cause any change in temps whatsoever. I would like to figure out where in the loop your sensor is before theorizing further.

Link to comment
Share on other sites

That's is a bit odd. I have the same XPSC 10K fitting in my system. I will try and replicate it later, although I suspect this might be a sensor location issue in combination with the surge from speed change. In the loop, is it in the lower center part of the fixed tubing, right after the pump on outflow? I can't tell flow direction from the picture.

 

**I can't recreate this. I have two XSPC 10K probes stuck into the ends of a false T valve entering and exiting the radiator. I normally run my much smaller loop at 2000, but shifting to 4000 does not cause any change in temps whatsoever. I would like to figure out where in the loop your sensor is before theorizing further.

 

My loop order goes res/pump>slim rad>thick rad>temp sensor>GPU>CPU>res/pump

 

So the sensor is right after the radiators and before the GPU and CPU, where the EK T-block is at the bottom. Perhaps I should run the test again and use a laser thermometer to check the water temperature? do you think this would be accurate?

 

Also are your XSPC sensors both hooked up to a commander pro?

Link to comment
Share on other sites

Both my XSPC sensors are on the Water In/Out sensors on the Code. When I set it up, I was covering my bases for how I would control things. Off the C-Pro (and Code T-Sensor) are two more probes that run to the exhaust side of the HW Labs 280 for the GPU. The exhaust air temp is effectively Water In temp and my GPU temp as well. Lots of redundancies, but that was intentional and I can switch between C-Pro or BIOS operation control by moving the connectors. Also, I like data.

 

The only thing that really comes to mind for your situation is the pressure drop when shifting from max to half speed causes pause in flow with the post radiator T-sensor detecting lower temps and the subsequent CPU and GPU blocks show a momentary spike as the coolant slows (or stalls) passing through, something a bit like fluid backpressure.

 

How long does the condition last? As long as the pump speeds are low? Or only a few seconds before returning to normal? It is just 2-3C right? (about what you can take out in a single pass). Another way to test would be to step the pump down in smaller intervals. So drop from 4900 to 4400 to 3900 to 3400 to whatever until you get to 2400. Give it a few seconds at each speed, enough time to let 1 unit of water do a full loop. Do you still get the surge?

Link to comment
Share on other sites

Both my XSPC sensors are on the Water In/Out sensors on the Code. When I set it up, I was covering my bases for how I would control things. Off the C-Pro (and Code T-Sensor) are two more probes that run to the exhaust side of the HW Labs 280 for the GPU. The exhaust air temp is effectively Water In temp and my GPU temp as well. Lots of redundancies, but that was intentional and I can switch between C-Pro or BIOS operation control by moving the connectors. Also, I like data.

 

The only thing that really comes to mind for your situation is the pressure drop when shifting from max to half speed causes pause in flow with the post radiator T-sensor detecting lower temps and the subsequent CPU and GPU blocks show a momentary spike as the coolant slows (or stalls) passing through, something a bit like fluid backpressure.

 

How long does the condition last? As long as the pump speeds are low? Or only a few seconds before returning to normal? It is just 2-3C right? (about what you can take out in a single pass). Another way to test would be to step the pump down in smaller intervals. So drop from 4900 to 4400 to 3900 to 3400 to whatever until you get to 2400. Give it a few seconds at each speed, enough time to let 1 unit of water do a full loop. Do you still get the surge?

 

The effects appear to be lasting, lower pump speeds = lower coolant temps but higher component temps when under load, higher pump speeds = higher coolant temps but lower components temps under load. When idle however, higher pump speeds = higher water and component temps and lower pump speeds = lower water and component temps.

 

We are only talking 2-3C but it's pretty consistent. If I'm gaming and I turn the pump speed down, my GPU temps will rise a couple degrees and stay up. If I turn the pump speed back up again they will come down again within a couple minutes.

Link to comment
Share on other sites

If the temp change is steady and long lasting, it seems pump speed does matter in your loop. I don't see a lot of restrictions, but a reduction in the fluid movement seems to be the only real explanation. Perhaps the back up is at in the GPU block, with the coolant slowing down in the GPU, raising the temperature of the water in the block since it is there longer. The reduction in speed also means the coolant is in the radiator longer and has more time to release heat and thus comes back to the 10K sensor cooler. I don't know. I can see this at a 1C difference but at 2-3C it feels like a stretch without a dramatic difference in actual coolant flow (not pump speed). At my normal 2000 rpm on the D5, it is hard for me to detect fluid movement. At 4000 it is like someone is blowing bubbles with a straw inside the GPU block. Maybe see if you can find where you tipping point is for clear fluid movement and if that has any temperature effect.
Link to comment
Share on other sites

If the temp change is steady and long lasting, it seems pump speed does matter in your loop. I don't see a lot of restrictions, but a reduction in the fluid movement seems to be the only real explanation. Perhaps the back up is at in the GPU block, with the coolant slowing down in the GPU, raising the temperature of the water in the block since it is there longer. The reduction in speed also means the coolant is in the radiator longer and has more time to release heat and thus comes back to the 10K sensor cooler. I don't know. I can see this at a 1C difference but at 2-3C it feels like a stretch without a dramatic difference in actual coolant flow (not pump speed). At my normal 2000 rpm on the D5, it is hard for me to detect fluid movement. At 4000 it is like someone is blowing bubbles with a straw inside the GPU block. Maybe see if you can find where you tipping point is for clear fluid movement and if that has any temperature effect.

 

Yeah I know it's bizarre, that's why I'm questioning the reading of my sensor. I know pump speed shouldn't make that large of a difference so sudden, especially when we are talking about the 2500-4900RPM range. Also the idle temps are sometimes questionable.. That's why I'm wondering if a laser thermometer would be a good way to verify the actual water temp, if I were to point it at the tube right before the sensor?

 

The Phanteks GPU block I have is actually one of the less restrictive GPU blocks on the market (but slightly underperforming thermally as a result) and yeah I agree with you in that there shouldn't be a tonne of restriction in this loop, so I would have expected 2500RPM to be plenty. It's awfully strange. I'm kinda kicking myself for not including a flow indicator.

 

I'm gonna have to tear it apart again sometime soon because it's already looking a bit dirty inside the loop.. The inside wall of the reservoir has that dirty look like car glass after the rain. Now that all the bubbles have let go inside the system there are these sort of watermarks where they were in the res and tubing. I don'y know if it's an issue with the EK coolant or if it's insufficient radiator cleaning, but either way once I've finished talking to EK about it I'll be taking it apart to clean. I'll probably use that opportunity to add a flow indicator to the system, and correct any other issues I hope.. It sure would be nice to have it just work normally so I can enjoy it lol.

Link to comment
Share on other sites

The only thing I am a little worried about is if the laser thermometer will give you an accurate reading. I am not sure you can get the fluid temp. You can probably get the tube temperature, which over a longer duration is the same, but I am not sure it will help identify unexpected changes in liquid temp. Clear materials can also be a bit problematic to scan. I'll try to think of another way to assess this.
Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...