SPCR

Hi,

Recently bought heat pipes from mcubed (Borg HPC). For various reasons (better fit to the case I'm building) I mounted them in the "wrong" direction (i.e., the end which are supposed to go on the cpu was put on the "cool" side, and vice versa ...).

In this case the cool side is an aluminum, finned wall, which should be sufficient to cool the cpu passively. However, the temp readings I get are too high, even without any stress on the cpu: after about 20 minutes in BIOS the temperature readings in BIOS indicates a CPU-temp of 64 (Celsius), but at the same time the temperature at the aluminum wall, at the opposite side of the point were the heat pipes are attached, the temperature is measured to about 20 degrees less (45 Ceclius).

I have not worked with heat pipes before, but a 20 degrees difference between the hot and the cold side seems way too much.

I see at least four possible explanations for this:
1. Heat pipes are broken.
2. Heat pipes don't work in the "wrong direction".
3. CPU readings from mainboard is faulty.
4. Heat pipes/heat sinks incorrectly mounted (too much or to little paste, surfaces not clean or plane, ...).

Any hints on this would be greatly appreciated,

/jon

Maybe something in Q: Heat pipes need gravity - do they work better vertically will be of help to you.

Hi, thanks,

The pipes are positioned so that the cold side is above the hot one (which, if it matters, should be "a good thing").

However, the pipes I bought are L-shaped, with one leg longer than the other, so when I was referring to direction I was more wondering if the internal structure of the heat pipes dictates a preferred direction?

Thanks again!

the temp readings I get are too high, even without any stress on the cpu: after about 20 minutes in BIOS the temperature readings in BIOS indicates a CPU-temp of 64 (Celsius)

FYI, sitting in BIOS is the equivalent of 100% CPU usage, no-one knows why. so there is stress on the CPU.

it's possible the internal wicking structure only works in one direction, although from the little I know of capillary action it should work both ways.

Ah, didn't know that (about stress on CPU in BIOS), interesting!

But still strange there should be a 20 C difference between the two sides of the heat pipes (maybe temp-readings on MB wrong?).

Guess I'll have to try both directions to see if there is a preferred direction.

Thanks a lot!

Got answer from Mcubed. No preferred direction.

jaganath wrote:FYI, sitting in BIOS is the equivalent of 100% CPU usage, no-one knows why. so there is stress on the CPU.

The BIOS heats the CPU because it is constantly executing instructions in a simple polling loop looking for keystrokes, CD loads, etc, as well as monitoring sensors. It does not use any of the (at least four) low-power modes provided by the x86 that are used by Windows and Linux when idling. Any sophisticated operating system goes into a "wait" or "sleep" or "deep sleep" mode when waiting for an interrupt in the absence of any OS or user CPU work to do. These modes turn off the clocks of most of the CPU function units, dropping their power consumption significantly (leakage current only). I have never encountered a BIOS that does this.

Thanks for that in-depth explanation cmthomson.

deleted

Just a few thoughts from a non-heatpipe-expert

45 deg C on the alu block, indicates heat is transferred away.

45 deg C, after how long time with CPU on? And how big is that alu-block? If it's a big one, it would take quite a long time to heat up !

Are you aiming for a passive solution? Maybe it's worth trying with a slow fan on the alu-block, not meant as a final solution, but to get an idea of the alu-blocks capasity...

Do you have any temp-data with same CPU, but more conventional CPU-heatsink/fan? That could maybe give a good indicatiuon of what to expect.

Also, I'm curious, can you provide a pic?

Good luck with the work

Thomas wrote: a few thoughts from a non-heatpipe-expert

Me neither. Did something really stupid, which I think explains the behavior I saw.

The heat pipes I bought (Mcubed) are supposed to be attached via heat spreaders on top of a block of copper at the CPU. I thought I could do it without the block, and instead mounting the heat spreaders directly on the cpu. Haven't tested yet (system is in an other town, and I only had a few days to work on it, with a baby and a three year old, it is harder than it used to be to get permission for these kind of activities...), but the heatspreaders have a small hole in it (M6) for pulling them together over the pipes (if this is hard to follow, check mcubed.com for details of the stuff). Anyway, this all means that there is a small part of the cpu which isn't covered by copper. Fairly certain this explains the differences in temperature I see.

As I said, stupid of me

Thomas wrote: 45 deg C, after how long time with CPU on? And how big is that alu-block? If it's a big one, it would take quite a long time to heat up !

It was only on for about 30 minutes, I'll do it properly after I have fixed cpu/heat pipe connection. I will put in a smaller pice of copper between the CPU and the heat spreaders, 1-2 cm thick.

Thomas wrote: Are you aiming for a passive solution? Maybe it's worth trying with a slow fan on the alu-block, not meant as a final solution, but to get an idea of the alu-blocks capasity...

I've made another system, totally passive, in that case the cooling wall was 25 x 40 cm, 3mm thick, with some fins on it. This was perfectly OK to cool a 62 W CPU (Semperon 3000+). Reached about 25 C above ambient after around 3 hours of full use of cpu. No pictures, sorry, (sold it). Also, note, no GPU in this system, only internal graphics, so cooling probably not enough if you want a gaming machine.

Thomas wrote: Do you have any temp-data with same CPU, but more conventional CPU-heatsink/fan? That could maybe give a good indicatiuon of what to expect.

Should have done this, will do next time...