Backplate cooling?

[billbartuska]

I found this photo on another forum. A few ppl said they had added fans to the bottom of their MB's and saw lower MB temps....one poster said the there was no effect on the elusive OC though. Anyone looked at watercooling? A GPU block here maybe? A non elecrical conductive base would be the issue to overcome.

[BillA]

this will 'manipulate' secondary path 'losses' wrt the principal sink
suggest experimenting with ducted cold air first (diff dias, mobo backside) to see the effect

[guandi]

I have known about motherboard backside temperatures for years as i have always cut out a section of the motherboard tray to allow fitting mounting plates for waterblocks and changing them quickly during testing.

Most of the heat in this area is radiated from the back of the CPU but there is some from the motherboard itself due to high component count in this area on some boards. As it would be highly inefficient to try and cool the cpu through the motherboard, air gap and plastic of the socket the only way would be to directly cool the back of the chip itself.

There are several ways to go about this each with pro's and cons.

Remove the socket from the board - This would allow free access to the rear of the chip on sockets with "holes" in (462 etc) but it would require a special wire "harness" to link it to the motherboard, this extra wire would however affect the cpu's performance as the wires would add resistance due to their length that would slow the flow of electrons (no matter how thin they are), and as anyone knows the reason they make cpu's circuits smaller and smaller is so that the electrons dont have so far to go, but this has the excellent effect of allowing more circuits in the same space, but i'm rambling, the reason you want to cool the back of a cpu is to squeeze extra power out of it, so the harness would outweigh any other potential gains.

Cool the Socket on the Board - This would eliminate the need for a harness and its problems, and allow for complete socket (939 etc) cooling. the only problems with this is finding a heat conductive material that is completely electrically non conductive and designing a clamping method for the cpu as the socket mechanism would just hamper cooling efforts. then you woul have to cool the socket "block" itself probably with radiating fins from each side cooled by the cpu cooler itself, for other solutions, off the top of my head maybe a small heatpipe loop (passive) attatched to each side or a water loop around the edge of the block (water, of course)

Hole in the Board - yep, have the Motherboard designed to have to have a hole behind the socket, that way you can directly attatch a cooling block to the back of the CPU with sockets with holes in. This of course would require a lot of work redesigning the PCB in that area, as motherboards get more sophisticated they have more layers of circuitry and components in this area, its just like a silicone sammich with everything on it.


So... it probably can be done, but its really just not worth it. Anyone obsessed enough to want to cool the *back* of their CPU would probably already have looked into phase change and pelt cooling, which can bring the entire chip, and even the back of the motherboard down to well below freezing. But in reality, water cooling is enough to get the back of most of my motherboards to a few degree's above room temp.

[bigben2k]

Originally Posted by guandi

...

Hole in the Board - yep, have the Motherboard designed to have to have a hole behind the socket, that way you can directly attatch a cooling block to the back of the CPU with sockets with holes in. This of course would require a lot of work redesigning the PCB in that area, as motherboards get more sophisticated they have more layers of circuitry and components in this area, its just like a silicone sammich with everything on it.


So... it probably can be done, but its really just not worth it...

Yep. It could be done with Socket A, but with today's pinouts, you can forget about it.

I explored the idea of a CPU back side cooler some time ago, here:
http://forums.procooling.com/vbb/sho...&threadid=5438
(never actually did it though).

[Dunno]

I have cut holes in a number of backplates behind older CPUs and mounted fans.
According to the temp probes temps drop a lot, but many argue that you are just cooling down the probe. I did see slight improvements on OCability/stability though.

I feel that for the probe to be cooler everything in the area has to be cooler?
I do concede that the traces going to the probe may be less resistive, thus given an 'egzadurated?' lower temp readings?

One must also remember that 90% of the heat dissipated by Mosfets is dissipated through the metal part of it thats soldered to the MB.
(New Mosfet designs that dissipate to both sides may be out already?)

Having a fan that blows on the area behind the CPU and Mosfets wont hurt but does tend to build up a lot of dust on a critical area of the MB. So high mantainence.

I have been thinking about a large 'waterbag' similar to Koolance's new RAM & HD coolers, that fits behind the MB.
Time to buy a box of wine, polish it, and experiment, perhaps?

PS: What about filling the socket with AS Ceramique to conduct heat better to the back of the MB. Is it non-elec-conductive enough?
Perhaps a small piece of copper, coated with something non-elec-conductive in the rectangular space of the older sockets?