Serious Flaw in watercooling?

[ibmkg]

I recently homebrewed a water-cooling system. I will pass on the details later as:

The water block is meant to cool processor only. The problem arises when the nearby caps and power regulator (8 phase) get hot. Conventional method (FAN) not only cool the processor but the nearby components as well.

What should be done about them? They get so hot that one cannot even tough them with bare fingers!

[bigben2k]

That would be an ASUS board, right?

It would be unusual for caps to overheat; are you sure you don't mean that the inductors (coils) get hot? That's normal, and usually harmless.

If any company builds a regulator circuit properly, you can bet Asus is one of them (they have close ties to Intel). If it needed more cooling, it would come with it.

If you're overclocking & overvolting your processor, then see a dramatic rise in temperature, that's a whole different situation: you might want to look into sticking some heatsinks any which way you can, without shorting out anything. Many have even water cooled their power regulator circuits.

[ibmkg]

Wow. You are good. It is an ASUS mobo wit intel chipset.

I do not know about inductors temprature since they come behind my waterblock. The near by caps 'do' get hot.

Asus already has mounted a small heatsink over these regulators and it gets fairly hot without overclocking. They used AC 3 paste, i removed it and used Arctic Silver 5. I cannot mount a bigger Heatsink as there is no room. I also cannot mount a fan over it for same reason.

Currently my rig is running idle 42C, casing temp 35C with fan running at 5V. Both caps and the heatsink over regulators are fairly hot.

Besides that RAMs also get warm (the heat spreaders actually). Processor fan used to keep them cool as well.

[JamesAvery22]

The VRMs, caps, inductors all have acceptable temperature tolerances. Do more than just the "finger test" and find out how hot the board and components actually are. If your PC is stable and the temps arent bad then why worry?

The CPU waterblock not providing air flow over the surrounding components is hardly a flaw in water cooling. Many boards need next to 0 air flow around the VRMs. If your motherboard requires airflow on the VRM then it sounds like a personal problem Replacing a heatsink/fan combo with a waterblock means you'll have reduced air turbulence around the surrounding components. Less fans = less air flow. Thats sort of common sense, don't know what else to tell you.

Thermalright makes after market VRM heatsinks that allow you to mount fans on them. Heck you could probably mount some 40mm fans on the heatsinks you have now. As I said above though, if the temps aren't too high and you're system is stable, why bother?

[ibmkg]

The only reason I have water cooled my PC is to OC it. This resulted in an aim to make WC under USD 40 with acceptable performance.

The air flow also kept the PCB cool. Hell I could leave the stock HSF if it weren’t for OC.

I was told this by a mechanical engineer that water cooling approach is not very practical. And one of the reasons he mentioned was this issue. I turned deaf and continued. Now I see his words as golden.

What’s done is done. For long term I am thinking either redesigning the WB or maybe leave it as it is. Currently I am using Al block and am going to switch to Cu (with similar design). The reason is to bust the myth of Al dissipation vs Cu heat conductivity while I am at it.

I can make (homebrew) heat sink for Voltage regulators but that was not my point. I thought the purpose of FAN was to cool processor only, which was never the case. A very important point I would say. The current HS on VRM says ‘Zero Noise ASUS DESIGN’.

And to conclude, water cooling can cool a single unit to be cooled. It is useful for serious cooling (like for OC purposes). Air cooling can be active and passive. Furthermore, air cooling is efficient as it as cost effective, lower power consumption easy to implement with low maintenance and if there is ‘oops!’ there is no bye bye.

[bigben2k]

Originally Posted by ibmkg

Wow. You are good. It is an ASUS mobo wit intel chipset.

I do not know about inductors temprature since they come behind my waterblock. The near by caps 'do' get hot.

Asus already has mounted a small heatsink over these regulators and it gets fairly hot without overclocking. They used AC 3 paste, i removed it and used Arctic Silver 5. I cannot mount a bigger Heatsink as there is no room. I also cannot mount a fan over it for same reason.

Currently my rig is running idle 42C, casing temp 35C with fan running at 5V. Both caps and the heatsink over regulators are fairly hot.

Besides that RAMs also get warm (the heat spreaders actually). Processor fan used to keep them cool as well.

Asus is the only one, to my knowledge, that uses an 8 phase regulator . You have an Asus board from the P5 series.

Most of these components are rated to run at 75 deg C or above. Of course running these components on the high side of the max temp isn't desirable, because it brings down the expected lifetime.

AS5 would probably work well, but improvements would be marginal. I'd focus on putting a fan.

I use a Cooler Master Centurion case, which has a side inlet, venting air right over the core. You might consider something similar. Depending on the particular setup, you might simply be able to put a 40 mm fan on the other end of the heatpipe (if there is one). Small 40 mm fans push only ~3 cfm, but thicker 40 mm fans can push a lot more air. I've got a 40 mm fan that's almost 40 mm thick.

I wouldn't worry about the RAM. DDR2 puts out more heat (than DDR), but still negligeable. Just make sure that you have at least basic case ventilation.

[JamesAvery22]

water cooling under $40? Would love to see what parts you have.

The practicality and performance of any cooling system depends on its design and application. There is no single water cooling model. If your system isn't practical to you then you need to redesign your setup. Enough generalized statements, yatzy.

p.s., search more an Al. Its not a big deal if you have the right coolant.

[bigben2k]

Since it came up, I thought I'd share some of what I've found out, in my searches on the power regulators.

Beyond the basic of providing a steady voltage (within a certain percentage), a VRM (voltage regulation module) has to maintain that voltage regardless of the changes in current draw (i.e. power draw). Everyone can easily imagine a power supply's voltage dropping when you put a heavy load on it, it's the same concept.

Except that this is still a switch mode power supply. Meaning that it provides power in very small time slices. The main PSU works the same way, but at a lower frequency.

The reason for the time slices, is to split the power among several regulators, which would have to be very large, on a single stage. Since they're smaller, they can't handle the full power load, but can accomodate 1/2, 1/3 or in this case 1/8 of the full load.

The push in the industry is to make things smaller, and more efficient. The latest development is to use the coil's resistance, as a current sense (aka "power meter") to track and compensate for load variations, to maintain a steady voltage. Coils are, unfortunately, not very accurate, usually manufactured in a manual process, with variances in the order of 10%. New coils use a ferrite that completely enclose the coil, which is now just a fraction of a full turn, which can be manufactured in a completely automated process, and yield variances of 3% or less. Higher frequencies use a coil with a lower inductance value, so a high frequency is desirable. I believe it also makes for a smaller capacitor.

My last bit of searching (aka reading electronic trade magazines) put me across Vicor (vicorpower.com). They're pushing a VRM that powers a CPU straight off of a 48 Vdc power supply (48 Vdc power supplies are a common telecommunication equipment power supply). Assuming a 96% efficiency, that dumps 4 Watts of power next to the CPU.

Is the PC world ready for a 48 Vdc power supply, with dc-dc converters to bring it down to the other voltage requirements? Today's DC-DC converters can hit 96% efficiency, dumping minimal heat. The 48 Vdc supply can still get at least 85% efficiency, same as today's standard PSU. Voltage regulations can easily meet the 5% ATX standards.

Here's another little known bit of info: in large data centers, it's not uncommon to have power supplies that run off of a 220Vac source: the required electrical infrastructure is far less expensive, and actually makes up the cost of the heavier power supply. If you look at voltages and max amps, it makes a lot of sense:
110Vac line, max amps: 15 or 20 amps.
220Vac line, max amps: 40 or 50 amps.

Safety regulations state that one should keep the line loaded at no more than 80% of max for a line/breaker/outlet, with the loaded equipment's rating (I believe that this accounts for startup power, among other things).

That leaves 12 amps on a 15 amp circuit, 16 amps on a 20 amp circuit, 32 amps on a 40 amp circuit, and 40 amps, on a 50 amp circuit. Circuit breakers are rated for amps only.

Calculated power (V*I) :
110Vac @15 amps yields max 1320 (safe) Watts
110Vac @20 amps yields max 1760 (safe) Watts
220Vac @40 amps yields max 7040 (safe) Watts
220Vac @50 amps yields max 8800 (safe) Watts

[ibmkg]

bigben2k:

Nice.


---------------------------------

Updated Temps:

Room Temp = 32C

[120CFM Fan @ 12V, Currently running @ 5V]

Water Temp in Loop = 34C
CPU = 42-45 idle [50-55max on load]
Regulator Temp = 40-41C
Cap (near LGA775 Socket] = 40C
HDD = 40C
RAM = 35C

System Temp = 35C




Questions:

a) Should I worry about Cap temp?

b) Is regulator temp ok? This temp is actually of the heatsink located on the regulators.

c) Should I consider using vapour cooling? I can implement that in the existing loop (a little modification to the reservior, I will not remove the rad].

[bigben2k]

-Room Temp = 32C

That's a hot room!!!

-Water Temp in Loop = 34C

2 deg above room temp means you've got an excellent radiator/fan combo, but if your room temp is different...

-Cap (near LGA775 Socket] = 40C, Regulator Temp = 40-41C

That's not bad, actually.

- CPU = 42-45 idle [50-55max on load]

if your room temp is really 32C, you're doing very well. The odd thing is... you're saying: 32C. I've only seen that from one other member, based in some Mediteranean country, Saudi Arabia if I remember correctly, where "summer" is 54 deg C (outside). there might have been another South American member... You're in the US, are you in Florida or Texas?!? Are you running this computer in an outside garage or something?

[RADCOM]

can't you just get a couple of those antec spot coolers they are bendable ( with blue leds to boot lol )and they would convey the air away from the VRMs towards the exhaust fan at the back of your PC.

[ibmkg]

bigben2k:

Actually, I am running this PC in my home workshop (garage so to say). I play around with power tools just for hobby and try NOT to break things. The design of garage is such that its sound proof. However, the very design results in room to be hotter.

If I turn on the AC, then I cannot solder

Furthermore, I am using Al block. Converting it to Cu would yield reduction in temperature [hopefully]. If the difference were marginable, I would redesign the reservoir so that it becomes vapor cooled as well.

Finally, I am also working on Microcontroller 8051 series that would control this rig. It would sense and report temperature, flow of water and fan speed (control it as well with temperature as reference). Using a Microcontroller has an advantage as well, since its standalone it can force shutdown PC if anything bad is to happen e.g. if temperature is too high.
MC would report everything to PC using COM port. Its simple but would make the rig complicated.


RADCOM:

SpotCool
I had the very same design in mind. Infact I was searching for materials to construct it. I see its thermally controlled as well. Price USD20 max.
Aces!

[bigben2k]

Interesting.

I think in your case (given that the garage is attached to another building) I'd try to use a remote kb/mouse/monitor, and keep the PC cooler, but you're doing fine, really.

[ElMoIsEviL]

Originally Posted by JamesAvery22

The VRMs, caps, inductors all have acceptable temperature tolerances. Do more than just the "finger test" and find out how hot the board and components actually are. If your PC is stable and the temps arent bad then why worry?

The CPU waterblock not providing air flow over the surrounding components is hardly a flaw in water cooling. Many boards need next to 0 air flow around the VRMs. If your motherboard requires airflow on the VRM then it sounds like a personal problem Replacing a heatsink/fan combo with a waterblock means you'll have reduced air turbulence around the surrounding components. Less fans = less air flow. Thats sort of common sense, don't know what else to tell you.

Thermalright makes after market VRM heatsinks that allow you to mount fans on them. Heck you could probably mount some 40mm fans on the heatsinks you have now. As I said above though, if the temps aren't too high and you're system is stable, why bother?

Ummm actually newer motherboards need air flow near the VRM's as they employ far more of them cramped up closely together.

Asus have developed a heat pipe design to cool their's. When using watercooling Asus have bundled (with more of their boards) a fan that mounts ontop of the heatpipe cooler.