Skulltrail MOSFET
Oh man, wow, that's one way to cool them! I want to see that when you're done installing ;-)
Well, I've got good news and bad news. The good news is that I can confirm that the temperature reading was for the MOSFET "area".
Unfortunately, just putting waterblocks on the MOSFETs was not enough. Or at least thermal tape is not efficient enough for the blocks to do their best. There are smaller chips between groupings of MOSFETs which are taller than the MOSFETs themselves. This prevents putting the block directly on top of the targets. The thermal tape adds 1-2mm in height which is just enough clearance to put the block on and have it make contact with the targets.
The waterblocks are marginally warm, indicating that they are taking some of the heat. But, the surrounding chokes and capacitors are still "burn your finger" hot. My measured temperatures initially dropped by about 5C with the waterblocks. Though over time the reading does creap back up to the mid 90's again. At least, this was enough to complete several (as yet unstable) iterations of Intel Burn Test (at a disappointing 4.00 Ghz), whereas without the waterblocks it would always reboot before a single iteration completed. Moving above 4.00 Ghz requires more voltage which immediately causes failure as was the case before the MOSFET waterblocks.
So, I can now try to improve the efficiency of the waterblocks and/or add additional cooling to the surrounding chokes (5 per CPU).
Fortunately, I installed the MOSFET waterblocks in such a way that I do not have to remove the motherboard to change them out
This took quite a bit of finangling, but the blocks came with a fantastic set of hardware allowing me to get it done. Very impressed with the flexibility of these blocks/hardware/heatplates. As it turned out, I bought the wrong size heatplates. Yet, I was able to make everything fit by mixing and matching brackets.
I will likely custom cut some copper wafers (about 25mm x 16mm x 2mm) to lay over the MOSFETs, then use proper thermal grease between the MOSFETs, wafers and waterblock to get better heat transfer. I expect another 6C +/- 3C improvement by doing this.
Anyway, regarding the chips ya could probably slap some Enzotech BMR-C1 or C1L chips on them. Perhaps the same with the mosfets depending on your setup, the L is the low-profile version. These things are expensive but I use them on my watercooled GPU's memory and I'm quite happy with them.
Additionally, I will take Kouger's advice (a slight variation) and buy some passive sinks for my chokes. I found exactly the right size (10mm x 10mm) and exactly the right number (2 x 5 = 10 units) here:
http://www.sidewindercomputers.com/enmofocomu10.html
Just for the hell of it (and since I'm paying yet more shipping anyway), I'll grab smaller sized ones and see if I can sit them on top of the capacitors.
http://www.sidewindercomputers.com/enmofocomohe.html
I also switched out the two weaker case fans in favor of two stronger (higher CFM) ones arranged in a wind tunnel front to back across the northern half of the board. The stronger ones had been undervolted and attached to a double radiator. Now they're at full power as case fans and the original, weaker case fans have been moved to the radiator. No measurable difference, but more airflow can't be a bad thing
I'll post a build log somewhere once it's all said and done. My loop is not quite as sexy as it used to be since I've had to add connections for 4 more blocks! All in close proximity, no less, which means for some tall links in order to prevent kinking. The tubing now largely conceals all my expensive waterblocks underneath
Oh well, it is just a server after all
Mike
