WCing VRMs on motherboard backside

[pHaestus]

I saw part1 of this article a while back but I think I am even more impressed with part2:

http://www.overclockers.com/tips1115/

Good work.

[jaydee]

Nice work indeed.

I have been trying to figure out how to do something similar on my Abit KD7 for a while now. On that board the mosfets are tiny. They are at least half the size as the ones on the Epox 8K7A and they appear to be surface mount.

[Brians256]

Very nicely done, indeed.

[Cptn. Foo Foo]

Thanks guys for the comments. I am glad you guys liked the article :)

I am starting to get lots of emails from readers about the new article. I will also keep an eye on this thread if anyone wants to ask some questions here.


Cheers :)

[bigben2k]

Nice indeed!


One Q: did you consider mounting a waterblock on the back of the MOSFET directly, or was this not a possibility, because of how they're mounted?

[Cptn. Foo Foo]

I would have considered that option but I had already epoxied my top-side VR block on them. So unless I can get that top block off somehow I cannot get at them directly anymore.

A few people suggested I unsolder them and remount them verticlly with a block down the middle. I honestly do not know if that is even possible. The MOSFETS might not be arranged correctly for that to work.


LOL. I just thought of something. What if someone could make a block skinny enough (or flat enough) to fit between the motherboard tray and the back of motherboard. I'm sure you guys out there with nice milling equipment could come up with something ;) But if it was done that way then you wouldn't need to cut up the motherboard tray.

[g.l.amour]

hooah to you m8; although for me that is one bridge too far.

i do find alot of resemblance with my oc'ing probs in your conclusions:

1) my vcore fluctuates alot, when i set 2.05 it can range from a little above 2.05 to 2.00, that is alot if u ask me

2) anything i set above 2.05 makes no difference in oc capability

i guess that temp of the voltage regulators is responsible.

edit: would this problem not pose itself when oc'ing a p4? the vcore regulators have a lot less work if max voltage is only 1.85; or am i seeing it wrong here because in the end the total W for an oc'ed p4 or amd is the same?

[Since87]

Originally posted by Cptn. Foo Foo
A few people suggested I unsolder them and remount them verticlly with a block down the middle. I honestly do not know if that is even possible. The MOSFETS might not be arranged correctly for that to work.


Typically, the metal tab on the back of the MOSFET's is the drain of the part and will be connected to 12V. Substantial current needs to flow through the connection between the tab and the PCB. This current could be conducted through the waterblock, but the potential for 'galvanic' corrosion associated with having one part of a watercooling loop at 12V, is awfully high.

(Synchronous rectification fets excluded. The synchronous rectification fets are frequently smaller than the main switching fets. Connecting the drains of the synchronous rectification fets together will, quite likely, blow up the regulator.)

I'd suggest just soldering a strip of copper to the bit of the metal tab that stick out beyond the plastic packaging of the MOSFET. A copper strip soldered there, and sticking up into moving air, may cool just as well as a waterblock, regardless of which side of the FET the waterblock is mounted to.

Edit:

Instead of saying, "regardless of which side of the FET the waterblock is mounted to", it would have been better to say, "regardless of whether the waterblock is absorbing heat conducted through the epoxy encapsulation of the fet itself, or heat conducted through the epoxy of the PCB."

[Cptn. Foo Foo]

would this problem not pose itself when oc'ing a p4? the vcore regulators have a lot less work if max voltage is only 1.85; or am i seeing it wrong here because in the end the total W for an oc'ed p4 or amd is the same? That's something the Intel boys would know more than me...I have stuck to AMD ever since the P1 days.


But your question about wattage has made me wonder just how much heat is going into the coolant in my system. These are the sources of heat:

- Laguna PowerJet 2000 pump (which puts out 60-65w)
- Tbred running 2600-2700MHz @ 2.0v
- nForce2 ultra 400 chipset running 210-223MHz @ 1.7v
- GF4 Ti GPU running 310MHz @ ??v
- and the heat from the MOSFETs

I am curious to know if anyone here can estimate the total heat my two heater cores have to deal with :)


And on the subject of water cooling the bottom of the MOSFETs directly....it sounds more and more complicated. I am glad I didn't try it off the start because by the sounds of it I would have had a serious corrosion problem. My GPU block is the Swiftech MCW50 (which has an aluminum top...anodized but still aluminum).

[Cptn. Foo Foo]

I just found this cool website...it has a CPU wattage calculator :)
http://www.benchtest.com/calc.html#stress

It calculates that my 1700+ TBred running 2700MHz at 2.05v puts out around 170w of heat!!!

And the other day I tried for a better O/C and I got my system to boot Memtest at 2866MHz (I was going for a 100% O/C at 2933MHz). At 2866MHz the chip was putting out almost 200w of heat!

Are these wattage numbers for the CPU correct ??

[pHaestus]

That number is nowhere near accurate. I would be surprised if it were over 100W

[Dunno]

Mosfets are designed to dissipate heat through the side that is attached to the MB. Putting a HS on the other/top side has very little effect because of this. This is why Mosfet coolers have made very little difference until the Captain came up with this idea of his. Also putting heat into your MB is not good as it heats up other components on the MB which would be a lot better of not being warmed.

This is changing however!:
Link (http://http://www.irf.com/technical-info/whitepaper/thermalpcim02.pdf)
Link 2 (http://powerelectronics.com/ar/power_new_power_mosfet/)

Good one Captain! :)

[KnightElite]

Well Cptn., I'm quite impressed. That's some good pioneering work that you have done. Interesting results as well. I can just see some of the guys at XS getting a third prometeia to cool the mosfets now :).