2 XD5 pumps + 2 reservoirs in series

[TGAlex]

So after several months working on my latest build, placing several orders for new parts and dealing with shortages in between, as well as having to work around a number of problems that arose throughout the build, I have finally put it all together and it's running like a charm. Never thought I would be able to fully occupy an Obsidian Series 1000D but turns out I have. More on that build soon, for now I need some help with running two XD5 pumps, with their reservoirs.

Two things before I begin, one, I still haven't installed iCUE so the pumps are running in full speed (PWM disconnected), that's next on my list, had a ton of stuff to do, and second, I have already searched online and have found a few posts on different forums, and more or less they are all in line with that it's best to run a single reservoir, although two would work too. Now, I'm not interested in changing my reservoir configuration, unless there is a really good reason to do so, but I'd like to maximize the pumps' efficiency.

So, the pumps/reservoirs are connected with an adjustable Aqualink pipe that sits horizontally, length set to about an inch, plus a short male to female extender fitting to get my precise length, and connects to a 90 degree Corsair fitting on each reservoir, that's to give you an idea of how close the two pumps are, very close. Running 5L/minute on a 4 rad configuration (2x480 and 2x360), plus Maximus Extreme Glacial's massive block for the cpu and XG7 RGB 30 series TUF block, and a good number of 90 degree fittings throughout the loop, which I think is pretty decent flow.

Now, both reservoirs are halfway full, when the pc is off, and when I turn it on, the first reservoir loses a bit of coolant (level goes down by about half an inch) which goes into the second reservoir and gains that little extra. According to what I've read online, although there is nothing wrong about this, I mean the loop will still work fine (and it does, as said previously), it will not allow the first pump to run at full efficiency, as it also needs to compress the air that sits over the second reservoir. To fix that, I'm supposed to fill the second reservoir all the way to the top, tightly seal it, and that way the first pump can build liquid pressure into the second reservoir that from now on will act like a water block, effectively moving full liquid pressure into the second pump and then into the loop. Do I understand it correctly? And obviously no need to top up the first reservoir, that can be filled just like you would normally do in a single pump setup.

What I would like to know is, exactly what kind of efficiency am I losing right now? Is it like 5%, 10%? How much gain are we looking at, if I top up the second reservoir?

Also, you know, turning the second reservoir into a water block kind of worries me, so I was wondering, if I ended up doing that, instead of using the reservoir's plug, would it be more reliable if I fitted a short male to male extender on it, then XF ball valve and then the plug? That way I would have both the valve and a plug at the end, just for peace of mind. Do you think that makes sense? And that being said, I assume the valve is designed to handle the pressure?

One last thing, I've noticed the first reservoir is full of little bubbles that don't seem to go away at all (been a couple of days) while there are no bubbles on the second, literally none. How would you explain that?

Thank you guys, long post I know, sorryyy...

[LeDoyen]

using two pump/res combos is annoying because of what you're describing. ideally you want the two pumps in series, no reservoir in between. What you have here is just like having a big bubble of air in a radiator, kinda.

For efficiency losses i wouldn't know but, once the air is compressed you basically have water pressure at the inlet of the second pump which is what you want. the problem may come from the fact that you slightly reduce the volume of the loop when you do that.

Air is bad when it creates an air lock, a restriction to the flow of water, which is what happens when you have air in a difficult bend, or in radiators end tanks. those can kill the flow pretty drastically.

5L/min is nice, but realistically anything above ~2 will work perfectly, more than that is more headroom to be able to reduce the pump speed and have less noise while maintaining good flow for the waterblocks to do their job. I'm not saying that you should remove one pump, but one would be largely sufficient and solve your problem 😛

[TGAlex]

Thanks for your reply. So topping up the second reservoir isn't going to make any significant difference with respect to pressure. Perhaps I could just add a finger just so the first reservoir doesn't go too low, but still keep the second reservoir about an inch below the top.

Yes I figured that one pump might've been enough but I think barely. It's quite a big loop with a ridiculous number of restrictions, I'll post a few pics soon. If one pump could really handle the flow, chances are it would need to operate north of 4k rpm at all times, and probably max out under heavy load and as you just said, I was looking for a quiet operation. Still haven't installed iCUE, but I'm hoping the two pumps will provide good flow under 3k, ideally 2.5. Also, the radiators are massive and the 120mm fans are in push-pull configuration (2x8 so that's 16 fans in front and 2x6 that's 12 on top) which means neither the pumps nor the fans may ever need to operate beyond 60%, no matter the load. Of course that's yet to be tested, but that's why I went for such a cooling capacity. One other reason I went for a dual rather than single XD5 was to have a back up.

Still trying to figure why three days later the first reservoir resembles a transparent can of sparkling Evian while the second is pure mineral.

[LeDoyen]

your GPU will tell you if one pump is enough.

the waterblocks need a cerain amount of flow to work properly, below their threshold, the GPU temperature spikes very quickly by 10-15°C. Once you figured that rough RPM value with a simple test like playing with pump speed while in a game, you know how much headroom you have to play with.

with one pump, you're probably fine with a good margin, but if you need backup too, that can justify having 2 pumps. but if one dies, remember you'll have the dead pump acting as another restriction, and you still need to drain the loop to take the D5 out of it.

One thing too, flow does not matter as much as fan speed under load. Once the waterblock have enough flow to function properly, you don't NEED to increase pump speed when load increases. you can, to shave a degree or two more, but it won't do much. imagine with one pump your GPU block temps shoot up if you go below 2500 rpm, you can set the pump to like 3500 fixed and be done with it. just an example.

For the little bubbles, they eventually vanish, but it takes days.. weeks sometimes 😉

[TGAlex]

Alright, I'll post back when I've ran my tests. Thanks for your help.

[kano1977]

Did you ever figure it out. I just built a new rig with 2 aquatuning ultitube d5 reservoir combos. It's my first time connecting 2 reservoirs together and I am getting the same thing with having a ton of pressure in the second reservoir and having to tighten the cap so tight. I almost break it to prevent it from leaking. I have the one reservoir feeding into the inlet on the second reservoir and the outlet on the second reservoir going into my first radiator. 

[c-attack]

The solution is to get something in between.  That or you need to take out one of the reservoirs, which I am sure you'd rather not do.  You don't want to run pumps with reservoirs back to back.  

[LeDoyen]

.. or take out one of the pumps. it's extremely rare one needs two pumps.

If you have some mad setup with like 6 GPUs, 4 or 5 radiators and a massive CPU block, then it's better to go for a dual pump top like one of those for example ;

 

Aquacomputer makes one too 

ULTITOP DUAL D5 pump cover for two D5 pumps, G1/4 (aquacomputer.de)

Edited October 23, 2023 by LeDoyen

[TGAlex]

On 10/23/2023 at 12:50 AM, kano1977 said:

Did you ever figure it out. I just built a new rig with 2 aquatuning ultitube d5 reservoir combos. It's my first time connecting 2 reservoirs together and I am getting the same thing with having a ton of pressure in the second reservoir and having to tighten the cap so tight. I almost break it to prevent it from leaking. I have the one reservoir feeding into the inlet on the second reservoir and the outlet on the second reservoir going into my first radiator. 

Sorry for the late reply. No I didn't, just topped up the first reservoir too as it was getting on my nerves being halfway full, at least they are now about the same level.

One other thing I forgot to mention is that the inverse correlation between the two reservoirs keeps going indefinitely, I assume until the second reservoir eventually reaches the nearest to the top. Every time the loop is ran, coolant level will keep doing its thing, just takes quite a while until you notice a significant difference. On that PC I'm mostly in the weekends (use another one the rest of the week), say about 8 hours for the entire weekend. At that rate, takes about a month until you notice a 1/4 of an inch coolant level difference in either reservoir. So what I think happens into the second reservoir is that that little air pocket keeps being positive pressurized little by little, with coolant taking up its space, until it reaches a max pressure point (as per the pumps capacity), which I still haven't identified but is definitely under 1/2 an inch. The first reservoir should be on negative pressure as coolant goes down at the same rate. Last time coolants got there I unscrewed the caps a little (both reservoirs) with the PC off, which relieved pressure and vacuum respectively off the reservoirs and brought them to default levels.

You know what I think? If I have to do that like once a month, I guess it's no big deal, but honestly if air doesn't escape the cap, which it doesn't (leak tested the loop beforehand), water shouldn't either, so might as well just let it reach as high as it can. Definitely isn't going to reach the top, cause air isn't going anywhere, and two D5's can't build more pressure than my leak test so I shouldn't worry about the cap either. Still thinking that perhaps an XF valve in place of the cap, might offer some extra safety (still place the cap on the other end). But do I really need it? Probably not. And life goes on...

As I said, ended up topping up the first reservoir too, cause at least it now looks good.

 

On 10/23/2023 at 4:56 PM, LeDoyen said:

.. or take out one of the pumps. it's extremely rare one needs two pumps.

If you have some mad setup with like 6 GPUs, 4 or 5 radiators and a massive CPU block, then it's better to go for a dual pump top like one of those for example ;

 

Aquacomputer makes one too 

ULTITOP DUAL D5 pump cover for two D5 pumps, G1/4 (aquacomputer.de)

That looks smart, guess I could've fitted that instead and saved me from the coolant level trouble, but oh well...

Going to post some pics in a few minutes. You'll notice the 90 degree restrictions all over the loop (just didn't wanna get into bending), the flow indicator and filter,  plus 4 radiators (2 x 480 and  2 x 360), an XG7 RGB 30 series TUF block for the ASUS 3090 TI OC, ASUS Maximus Extreme Glacial's massive block that cools nearly the entire MB including the 13900KS CPU which is notorious for running hot so I needed to maintain a minimum flow at reasonable rpm (I know fan speed is more important at keeping coolant temps in check, but minimum flow was also a concern). I doubt a single D5 would keep up except maybe if running maxed out, which I didn't want. That, including having an extra pump as back-up was what I was after.

Pics in a while...

Edited November 25, 2023 by TGAlex

[TGAlex]

The following is playing Assassins's Creed Mirage in 4K and all settings maxed out. Note that the CPU load shows 1%, just like the PSU shows 200 something watts but that's because I wasn't quick enough to capture the screenshot. That said, it's a GPU intensive game anyway.

So that's 40C coolant at a maximum 3700rpm pump and 900rpm fans. 3k could probably be enough for the pumps but honestly I can't hear them while playing a game so why not, and the same goes for the fans. Room temp ~ 24-25C and the case really does considerably lift the room temp all by itself, you can feel it just walking into the room. And as you all know, when you notice heat coming out of your case which increases as load increases such as when playing a graphics intensive game, that is a clear indication that your loop, or whatever it is that you're running for heat dissipation, is doing its job.

In Cyberpunk 2077 both GPU and CPU temps, and coolant of course (and room temp lol) are slightly higher.

[kpeklund1991]

I just completed my first build and almost have the exact same setup as you.  I didn't realize running two combos would be such a task.

The reservoir filling is very tedious and almost kind of troublesome and scary.

 

I may end up splitting the loops in the long run.

 

Can you give me any tips? When I try to bleed air the left most reservoir almost overflows or fills up completely and I really don't like the idea of it being maxed out to the cap with pressure.

Any tips or "here's what I did" greatly appreciated...

 

 

[TGAlex]

You know what, it kept bothering me as well, but honestly? Let's look at the facts here. This is hardly the weakest spot for a leak, when you have 20 90° fittings in the loop (90° = resistance and 20 of those  = great resistance). Obviously you have far less of those, but still, pressure against the filler cap is nowhere near that. Furthermore, the filler caps themselves use the exact same o-rings as the fittings and they screw in the exact same way. It's why I ended up not replacing them with an XF valve after all, which I thought would provide better safety at first. It wouldn't. There was no point. Last, if you leak test the loop beforehand and no air escapes at like 10 psi for 30 minutes, there is no way coolant will, and certainly not from the filler cap. However, as with all loops, whether you have 20 90° fittings or none, whether you're cooling 1 component or 5, you need to check your o-rings every couple of years. Just take the chance to do that right after a full drainage. Unscrew each fitting, check, screw back if ok, have any doubts, replace o-ring, simple as that.

So that being said, I did nothing, just totally ignored it. I know that's probably not what you wanted to hear, sorry. 🙁

Last, for what it's worth, with the PC turned off, unscrewing the right hand side reservoir's filler cap first, makes no real change to the coolant level. Then, unscrewing the left hand side cap causes the left hand side reservoir's coolant level to drop, while the right hand side goes up, until they are the same level.

[LeDoyen]

On 11/25/2023 at 5:06 PM, TGAlex said:

Going to post some pics in a few minutes. You'll notice the 90 degree restrictions all over the loop (just didn't wanna get into bending), the flow indicator and filter,  plus 4 radiators (2 x 480 and  2 x 360), an XG7 RGB 30 series TUF block for the ASUS 3090 TI OC, ASUS Maximus Extreme Glacial's massive block that cools nearly the entire MB including the 13900KS CPU which is notorious for running hot so I needed to maintain a minimum flow at reasonable rpm (I know fan speed is more important at keeping coolant temps in check, but minimum flow was also a concern). I doubt a single D5 would keep up except maybe if running maxed out, which I didn't want. That, including having an extra pump as back-up was what I was after.

Pics in a while...

That one went over my head 🙂 i didn't see the reply.

THe hot temps on CPUs are mostly due to excessive voltage, not lack of flow. GPUs are a bit more sensitive in the flow area but the threshold at which temps will increase is quite close to the minimum speed of the pump usually.

Even in a 1000D decked with radiators, a single D5 or DDC is way more than enough. Even workstations with 5 or 6 GPUs don't need more than that.

If a CPU overheats on custom loop, it requires a bit of tweaking to keep the Vcore in check, more than water flow.

[TGAlex]

16 hours ago, LeDoyen said:

That one went over my head 🙂 i didn't see the reply.

THe hot temps on CPUs are mostly due to excessive voltage, not lack of flow. GPUs are a bit more sensitive in the flow area but the threshold at which temps will increase is quite close to the minimum speed of the pump usually.

Even in a 1000D decked with radiators, a single D5 or DDC is way more than enough. Even workstations with 5 or 6 GPUs don't need more than that.

If a CPU overheats on custom loop, it requires a bit of tweaking to keep the Vcore in check, more than water flow.

Maybe you're right, I dunno, you seem pretty certain. Just went with two for peace of mind. Though, running two, doesn't that mean they can provide the same flow at lower rpms and therefore wear out slower?

Yeah I was once tempted to tweak the voltage after watching a JTC video, I'm sure you know which one I'm referring to, but I totally chickened out and I feel absolutely no shame to admit it. 🤣

[c-attack]

On 4/25/2024 at 8:43 PM, TGAlex said:

Though, running two, doesn't that mean they can provide the same flow at lower rpms and therefore wear out slower?

Not exactly. Two pumps means 2x the fluid pressure, but the actual flow rate may not change. Pressure is still useful and some loops may need it. The more vertical travel you have in your loop, the more pressure you need. Other fluid restrictions like extra dense CPU blocks or extreme angular changes all through the loop can add to the flow resistance and make more pressure helpful. However, if you are not pressure restricted, then it won’t change flow rate. 
 

My system is very tall and there is a an external radiator below the desk making total vertical travel almost 1m, not including the up/down in the 480mm radiators. Very clear difference in flow rate with one vs two pumps, but even then it was fully functional and effective with one D5. I also use a pump only with no reservoir for the second unit.
 

That doesn’t mean you have to drop the other pump. You can run separate loops for cpu and gpu. You can get creative with your planning and work out a routing scheme to avoid the direct connection between them and help keep some balance in the reservoirs. However, as mentioned earlier the one thing you really need to do is make sure they are running the same speed. I’d probably stick with a fixed pump speed for both units, manually changing it when needed to avoid unnecessary frequent pressure changes. 

Edited April 27 by c-attack

[TGAlex]

I think I got it, thanks. Two pumps can handle higher head pressure but for actual flow rate to increase, you'd need a pump that can operate at higher rpms. Academically speaking of course, as it's not needed in a computer loop.

Yes running a 2nd pump with no reservoir is probably better, technically. But I have to admit, I also liked how they looked together. Got the idea from the Corsair site.

I don't think I'm going to make any changes in the near future to be honest, but splitting the loop sounds like a great idea. I'll get to keep the two reservoirs next to each other too that I want for aesthetic reasons. I may go for it when it's time to check the o-rings.

[kpeklund1991]

What kind of idle temps do you manage?

we almost have the exact same build, im running a 13900ks and a 4090, dual pump loop

my idle jumps all over from 30 to 40.. seems strange.

pumps running at full speed. just seems to go all over the place. 

[c-attack]

10 hours ago, kpeklund1991 said:

my idle jumps all over from 30 to 40.. seems strange

Coolant temp?  Or cpu/gpu temp?

[TGAlex]

14 hours ago, kpeklund1991 said:

What kind of idle temps do you manage?

we almost have the exact same build, im running a 13900ks and a 4090, dual pump loop

my idle jumps all over from 30 to 40.. seems strange.

pumps running at full speed. just seems to go all over the place. 

As @c-attack said, we need to know if you're referring to coolant or cpu/gpu.

Sitting on the desktop for a while I have 28C coolant, 28C gpu and 31C cpu. Room temp, I guess about 25C because right after startup, iCUE reads 25C coolant and my UPS battery is between 25C and 26C as well.

This has nothing to do with the hardware though, I mean you could be running a 10 year old cpu or gpu and still have the same idle temps as your high-end system. It's all about the water blocks and radiators, but mostly water blocks and finding the pump and fan rpm sweet spot as per @c-attack and @LeDoyen very comprehensive write ups. Keep in mind, my loop cools nearly the entire MB as well, the block is freaking huge, it's almost twice as big as what you're looking at in the photos cause the rest of it is behind the gpu. In the third pic I've posted, look below the Corsair badge on the gpu, that little RGB thing is part of the MB block!

Btw I used to run the pump at 2800 rpm between 20C and 30C coolant (which are essentially idle or email/browsing temps) but turns out I'm getting the exact same temps at 2400, so I'm now running 2400. Still same temps at 2200 actually, but then I'm getting a faint low pitch humming noise that's a little annoying. When I'm writing emails or browse the internet without listening to music, I don't want the computer as quiet as an ant pissing on cotton, I want it as quiet as an ant not even thinking about pissing on cotton (Gene Hackman, "Heist"). That's 2400 for me.

Fans are set to 700 rpm between 20C and 30C coolant, but for some reason they don't go below 750, dunno why. However they're super quiet at 750 as well.

Here are my custom curves just to give you an idea, although keep in mind that what works for me, may not necessarily work the exact same for you.

Both pumps:

20C - 30C (idle/email/browsing) = 2400 rpm

40C = 3000 rpm

50C = 3600 rpm

60C = 4200 rpm

70C = 4800 rpm (full speed)

It's perfectly linear as you may have noticed, but I very well know that I'm never going to see anything north of 3300, ever. So what matters in this curve, is the part between idle and email/browsing temps (20-30C) and gaming or video editing temps (30-45C). In that part of the curve, the pump will operate between 2400 and 3300 rpm and that's all you want really. You don't need crazy rpm to keep coolant temp in check.

Fans:

20 - 30C = 700 rpm (but don't drop under 750 for some reason)

40C = 850 rpm

50C = 1000 rpm

60C = 1150 rpm

70C = 1300 rpm

Another perfectly linear curve and as before, what matters here is the part between 20-30C and 30-45C, where the fans will operate between 750 rpm (should've been 700 but whatever) and ~950. Now since fan speed is what matters most, rather than pump speed, I'm going to see if I need to get that 1000 and 1150 rpm earlier, like at 45C. Keeping idle temps in check may have nothing to do with the hardware, but temps under load does, and that being said, I've decided to stay with the 3090 Ti OC and skip the 4090 and wait for the next gen. Going to see what that requires in terms of cooling and make the appropriate adjustments. The 1000D with the kind of cooling it can accommodate, isn't just a case, it's a long-term investment.

[kpeklund1991]

Thanks for the replies!

So I am mainly referring to CPU temps.

I've heard the 13900ks is just really sporadic and the temps move around.

Icue shows my idle just sitting at desktop going anywhere from 35 to 40c, and that correlates with the little CPU temperature display on the motherboard.

That being said, hWinfo shows my CPU temperate at around 31c (all physical cores and virtual cores) with the occasional random core getting a boost and the temp spiking for a split second, and it shows the CPU package temp going all over like the icue temperate display.

It appears the icue readout is the "CPU package" temperature which essentially seems to just readout what the highest CPU core was at that given moment in time during the refresh interval.

Again, I've read that the 13900ks is just weird and randomly spikes a core for turbo boost or something of that nature. I don't know enough about it to speculate honestly.

My liquid temp reads a solid 30c almost all times, but the icue readings just go all over.

I guess I was just expecting much lower temps due to such a massive amount of radiator area and fan speed, etc.

Perhaps I should have not used the stock Corsair thermal paste on the cpu as well.

 

So I'm wondering wether to trust the icue readings or the hWinfo readings.

 

Thanks again for the replies!

[c-attack]

1 hour ago, kpeklund1991 said:

It appears the icue readout is the "CPU package" temperature which essentially seems to just readout what the highest CPU core was at that given moment in time during the refresh interval.

Yeah, don't get too worked up over the CUE CPU temp reading or it's % load either.  If you want detailed CPU temp info, use HWinfo.  While we are talking about it, CUE's core temp numbering is incorrect for Raptor Lake and has been for a long time.  I don't even leave my cores up on the dashboard anymore.  You've already figured out it provides a different reading for package temp than HWinfo.  Use HWinfo when trying to accurately asses core behavior, both for the temp and the frequency/VID/Voltage values you need to make sense of it.  

 

1 hour ago, kpeklund1991 said:

Again, I've read that the 13900ks is just weird and randomly spikes a core for turbo boost or something of that nature

This is normal behavior these days for Intel.  One core spiking is some minor task being executed and the CPU is doing what it's supposed to do -- respond rapidly.  A slightly more complex issue is when all the cores respond to these minor tasks and this can sometimes happen with some manual overclocks.  The CPU behavior has become very sophisticated and the brute force manual OC is not a great option.  However, the temps are still fairly simple.  When a core sparks up, it will warm up.  No amount of pump speed, fan speed, or radiator real estate that prevent it from heating up as voltage is applied. The CPU IHS transfers the heat to the cold plate on the block, which in turn transfers to the water.  The water transports heat to the radiator for dissipation.  The cooling system is a waste heat removal service.   It's fun to set up a massive case with yards of radiator length, but when your CPU is sitting at idle and only punching 17-23W, it doesn't make a difference whether you are running a 3x3m square external cooling system or a little 120mm radiator.  The cooling system only has 20W to get rid off and all CPU temp results are voltage and conductivity alone.  If we pair your motherboard and CPU with any cooling method or brand, the idle temps will be effectively the same aside from minor differences in thermal conductivity on the cold plate.  It's when you load up your 600W GPU and 300W CPU that you can start showing off and blowing that heat all over the room while maintaining excellent hardware temps.  

 

1 hour ago, kpeklund1991 said:

Perhaps I should have not used the stock Corsair thermal paste on the cpu as well.

If you get the mount wrong, you will see CPU temps 15-20C higher than expected.  They will instantly hit that 95-100C mark when they should not.  If you get the mount just a little imperfect, you might see a large disparity between core temps when at full load.  However, this is no longer a reliable metric as these 8+16 CPUs all have a health 5-9C core differential between best and worst cores.  Too many cores, too close together.  There is some heat sharing and some of them will be better/worse for it.  So a slightly imperfect mount might run just a touch higher than expected.  Don't use idle temp to measure that.  Use a full stress test with a linear load.  I usually recommend CPU-Z "bench test" because it is under the 253W Intel power limit so you won't bang off the ceiling and downclock while providing an even load.  This is a good way to test multiple things, like pump speed vs CPU temp or your heat transfer.  

[TGAlex]

As a rule of thumb, the loop is properly set up with no significant restrictions in the blocks or elsewhere and you have properly installed the cold plates onto the right amount of thermal paste, when on idle coolant sits between 3 and 5C above ambient, CPU sits (on average) between 3 and 5C above coolant and GPU matches coolant or worst case is no more than 1C warmer.

Ideally, you actually want coolant a few degrees warmer than ambient because what it means is that it's doing what it's supposed to, dissipating heat. If you lower pump rpm below a certain efficiency point you will notice coolant temp actually drop, but at the same time, CPU and GPU go up. What that means is that coolant is lingering in the blocks, being unable to flow and dissipate heat and iCUE reads lower coolant temp because the sensor is located in the part of the loop away from the blocks and since coolant isn't fully circulating or not circulating at all, that part becomes cooler, getting closer to ambient.

Now, if pump rpm is good and you still experience what I just described, then there may be something that's causing restrictions, effectively reducing flow which you would be able to see with a flow indicator. I was having this issue on my retired GTX 980 Ti system and as it turned out, I had chunks of dye trapped into the top radiator partially blocking the exit, causing a temperature deviation of 10C between CPU and coolant (as opposed to the 3-5C that's considered normal), same result as reducing pump rpm below the efficiency point. GPU wasn't really affected because it was the Poseidon Platinum series that was both liquid and air cooled, otherwise I believe I would've noticed the same deviation. And I didn't have a flow indicator on that loop so I had to guess what the problem was. If it wasn't that, I would've started looking for thermal paste problems, which wouldn't have made sense though as it all happened very quickly. For what it's worth, I managed to dissolve the chunks with EK products and it all went back to normal.

From what you're saying @kpeklund1991, I see nothing wrong and @c-attack has once again very nicely explained CPU temp spiking.