The Random Thread

Water and electronics don't usually go well together...
Yes. So I'm waiting for safe and efficient insulating coolant (fluid) to skip water cooling.

Another possible way could be put CPU sockets on the back side of motherboard and chassis to have heat spreader that is fixed with continuous welding with watertight manner and water cooling head to be placed at the opposite side of CPUs.
Code:
+--------------------------------------------------+
=                                                  =
|                                                  |
|  ------+------+-----------------------------     | Chassis
|        | CPU  |     Backside of motherboard      |
+========+======+==================================+
|     |            |      Heat spreader            |  fluid tube
|     |Cooling head|=========================================== to radiators
|     +------------+                               |
+--------------------------------------------------+
 
A lot of liquid-cooling devices on the market are referred to as Liquid Cooling and Water Cooling (including abbreviations LC and WC) interchangeably. Most of the time, the actual coolant is not really water anyway, there are different formulations on the market. T-Aoki , just take a look at what's available, there are designs that will blow your mind.
 
Do radiator fans sanely working?
Does the coolant sufficient?
Does water pump working?
Does thermal grease (or thermal pad) still there?
Sometimes grease spill out on long term usage, especially configured vertically.

Anyway, I dislike water cooling on PCs.
I would prefer liquid cooling once the coolant become insulating material that are fine even when the whole electric circuits are dipped into (intended or by accident). Or water cloling if all electric circuits and contacts (including CPU sockets, NVMe sockets, PCIe slots, ...) become water-proof.
Going to bring it to the repair guy. I think it's poor cooling but I can't find the cause. Is it possible that the board's thermal limiting system is faster than the CLI core temperatures? It's all around 70 degrees, so no visible overheating. Not a lot of dust inside either but maybe hidden.
This was my 1st watercooled system. It's disappointing. I expected minimal air flow but this thing still needs 3 massive case fans for the mainboard and PSU heat. No CPU fan is the only difference.
 
I think it's poor cooling but I can't find the cause.
The symptoms you describe are definitely poor cooling.

With no dust and three fans the cooling should be sufficient enough at low and medium loads the fans don't run at max rpm. This can have several causes, as T-Aoki already pointed out:
Sufficient coolant? You see, the coolant in "water cooling" is not actually water - or better not be. Pure water (worst case tap water, or even worse swamp water 😂) causes problems like rust, and over time algae grow which reduce the flow till a total block. That's why coolant consists of demineralized water with supplements like alcohol to supress algae. That's why the correct technical term is liquid cooling.
Plus the system needs to be filled up completely. If you hear some gurgle noises then this come from air bubbles inside. A correctly full filled up liquid cooling system delivers only the noise of the fans and pump. Over time coolant gets lost. Not much really. There must be no moisty spots. If so there is a leak: shutdown the machine immediately and replace the junk. But nothing is totally 100% completely closed. Those amounts are very tiny and are getting lost directly into the air, but over time it sums up.
Depending on the brand, the age and the coolant this shall take years. But it also depends on if the system you got was delivered already filled up correctly, like it should be the case with a ready to use already filled up delivered system, or an empty delivered system which needs to be filled first while assembling. When in the latter case the system was not filled up correctly to its full (you need to hold it right and turn it slowly to get the air out; see according handbook), then you get air in the system already in the first place.
A bit air doesn't do much besides gurgle noises, and of course it lowers cooling efficiency. But too much air will make the pump cannot pump enough water, so the cooler stops working. That's why you get coolant delivered also with already filled up ready to use systems, so to refill when there is a gurgle noise.
As T-Aoki also already pointed out, the cooler needs to be attached the CPU correctly. If this is not assembled correctly (not the correct cooler fitting the CPU, loose mounting, too much thermal paste used, or air gaps left) then the heat cannot flow sufficiently into the coolant and being transported to the radiator.

What also can be a cause is if the pump works sufficiently enough - or at all. But since you did not report any thermal shutdown the latter one cannot be the case.
As a mechanical module the pump suffers wearing. Depending on the age and if you do not bought the cheapest junk the pump normally last many years.
But there is also the settings. In most today's motherboards BIOS settings you can do several adjustments for the fans and one for the pump of a liquid cooling system. It's for to find your personal compromise between silence and cooling power at different loads.
If the pump runs too slow or its rpm is not adjusted to higher values when the temperature rises, the fans can work their asses off but not helping the fact that there does not come enough heatflow from the cooler they can blow off the radiator, 'cause the pump delivers too few, it simply pumps not enough coolant.

If all this is checked as being OK, there could be some dirt somehow gotten into the cooling system which blocks the flow. Then this can be a warranty case but anyway the need to replace the cooler.
 
What's the usual cause when a watercooled PC can't reach full load anymore? It's a i9 13900K. Please let it not be CPU circuitry decay...
I'd suspect CPU voltage stuff depending on if it was ever over-volted previously (possible if auto-BIOS was high/errata). Cooling system shouldn't matter as long as temps are still ok.

Not sure how newer Intel cooling is done, but with i5-8400H if I ever saw idle not at max 4.1GHz I'd suspect temps to be 90C+ either from a bad fan or needing to replace paste.

What's full load mean? If I max CPU cores I go from 4.1GHz to ~3GHz due to temperatures being too-high to be able to maintain higher frequency, but I can still do full load if temps weren't an issue. If power delivery was an issue i'd imagine the PC would power-off and the PSU being the issue, or powering-off from the motherboard if it's something with the mobo's VRM to the CPU.
 
What's full load mean? If I max CPU cores I go from 4.1GHz to ~3GHz due to temperatures being too-high to be able to maintain higher frequency, but I can still do full load if temps weren't an issue. If power delivery was an issue i'd imagine the PC would power-off and the PSU being the issue, or powering-off from the motherboard if it's something with the mobo's VRM to the CPU.
There's no clear indication but the case fans used to go full-on during a buildworld and it took 12 minutes. Now 25 or so. I can't find any problem.
This is the infamous voltage-patched
SkyRaptor lake. I had 2 total freezes over time with unknown cause but clearly hardware failure...
 
There's no clear indication but the case fans used to go full-on during a buildworld and it took 12 minutes. Now 25 or so. I can't find any problem.
This is the infamous voltage-patched Skylake. I had 2 total freezes over time with unknown cause but clearly hardware failure...
How's your spare space remaining in your partition / dataset your /usr/obj is placed?
Even if CPUs are fine with proper cooling, the above matters when the remainder is below 20% of the paritition / pool size. This may be more significant on ZFS than UFS, but both would be affected.
Note that, if I understand correctly, the spare spece includes the space reserved with zfs set reservation=.
 
How's your spare space remaining in your partition / dataset your /usr/obj is placed?
Even if CPUs are fine with proper cooling, the above matters when the remainder is below 20% of the paritition / pool size. This may be more significant on ZFS than UFS, but both would be affected.
Note that, if I understand correctly, the spare spece includes the space reserved with zfs set reservation=.
Not really sure what you mean. It's on a UFS SSD without the traditional partitions but just 1 filesystem on /.. Its copying speed at many small files, like I often do is as expected.
The entire build happens inside a chroot dir. /usr/obj is a subdirectory of that. I don't think it's any different tnan anywhere else.

I also have had strange ZFS behavior of a dataset on a different computer. 1 time it's fast, next time it almost clogs. It appears to have a struggle with particular data formats and/or special characters in filenames but I'm no ZFS expert. If something happens again, I'll make a post. I have no idea what happens. I started using datasets only to fast delete millions of files without accessing them all.
 
How do you define and measure "full load"?
Not sure, to be honest. It's based on what I've seen from this machine. It includes storage and RAM speed. I think a real test requires a program that uses all cores separately with unlimited possible load like deep chess analysis.

My suspicion is that the cooling pump doesn't reach the expected flow, so a mechanical failure. Tomorrow I get it back. See if it performs like before.
 
Back
Top