Grouper Mosfet Cooling/Block Discussion!

[fareastgq]

Ok guys, sorry for the messed up title, hehe, that's the only way I could get it in there. Anyway, What I want to get from this is exactly what the title means,.. it's simply just a discussion thread, feel free to conject, hypothisize, blah blah whatever..I just want to know what ppl think fact and opinion. Hopefully this can turn into something interesting,... if not, oh well. Keep in mind the size of the mosfet area and how hot they can get. Some questions to get started :

-What do you think the best block will be for it at this time?
-What would be the best designed block for an area like that?
-How much heat do you think that area puts out?
-If I had a Storm (g4-5), tdx, etc../Mosfet block/Dual 6800 blocks. what would my temps be like if ambient was @ 35c.
- Feel free to comment on what kind of plumbing u think would be good for a loop like the above mentioned.

[Fairydust]

MOSFETS can take some heat, performance degration only occurs at significant heat, so you don't need a high performance design for it. Does the Mosfet grouping include Vdimm regulation or only Vcore?

I estimate a low 2 digit Watt heat output even in heavy OC scenarios for a complete power regulation circuit, so even the most basic designs can keep it relativly cool.

Any low restriction design should do imho.

[fareastgq]

I find myself wondering if it's even worth it to add that level of heat to your loop when a simple NB sink like swiftech's or a thermalright may be more than enuff cooling. either way, if it's not putting out alot of heat, or putting out massive heat, we might be looking at the same senario as "gpu water cooling is not needed" If I did do it, I might do it just for bling, but I wouldn't want to sacrafice my loops temps when I know a sink will do the same job just fine.

[MaxxxRacer]

then just use a standard heatsink :stick:

BTW I edited the title as it was unreadable and really annoying.

[Psyche911]

The best block for it would be a non-impingement block like the MCW6000/6002. It needs the larger cooling target as the Storm and TDX mainly focus on the CPU core, which is a much smaller area than the group of mosfets. If you look at the standard size northbridge heatsink on the board stock, it's base isn't even big enough to fully cover the mosfets. They stick out a bit on all sides.

I would not cool it in the same loop as the CPU or GPU. For both restriction and probably a fair amount of heat dump. Maybe if you're running an Iwaki and a very, very high heat exchanging radiator.

I may buy the board myself, but will stick with air.

If it were to go in a loop, I'd do a dedicated one with it and the northbridge probably.

[Bloody_Sorcerer]

i take it we're talking about grouper here and how a NB sink can be mounted on it or whatev? It's kinda hard as the thread title is still totally screwed up (whats a mosfter?)
I think (not sure...) that the little black parts of the FETs are still facing upwards and the part that gets hot is the backside so the most effective way to cool them would be a heatsink (or waterblock, waste of restriction IMO) on the back of the mobo.

[fareastgq]

Grouper Mosfet Cooling/Block Discussion! <----- can you change it to that Maxx? I changed it but it doesn't really work, I can still see the old title.

[MaxxxRacer]

I changed it.

bloody, that was my typo.. it was pretty late last night when i did it so i mess up when i type.

[Para]

What I'd do is a "device" with 3-4 bendable heatpipes, as to fit any and all mobos out there, connected to a high-flow tube-like waterblock.

Figure that a good Vcore stage has an efficiency of 80-90%, so you'd actually have to cool like 15% of the CPU output power, and that's almost nothing to a good-quality water setup.

[naTTen]

Heatsinks works for me but if you really want to watercool the mosfets i guess a block with a hole drilled right trough it would do the job.

[fareastgq]

Heatsinks works for me but if you really want to watercool the mosfets i guess a block with a hole drilled right trough it would do the job.

what do you mean "hole drilled right through" ? Also, a sink can be mounted on it fine. It has holes for multiple configurations.

[Psyche911]

I think he means just a block of copper with a tube through it for the water to pass through. He's suggesting that it's' not hot enough to need any kind of thought-out waterblock design for sufficient cooling.

[fareastgq]

He's suggesting that it's' not hot enough to need any kind of thought-out waterblock design for sufficient cooling.

That could be true, if you were running stock...everyone and their mom is going to push this board like hell though (well at least I am). Though the power regulation looks superb.. I'm anxious to see what kind of heat gets out with xtreme clocks on it. I remember the even on the great nf7s's my mosfets got quite hot running a 2.7 gig xp-m 24/7. Multiply that by 12 or how many ever are on there and that's a pretty good heat source. What would be kind of cool perhaps, would be a block that went front and back, That would be the ultimate cooling I think since like bloody sorcerer said the backside gets pretty toasty too. Now that I think about it though, those xp's could take like 2 volts though, these 9nm chips at 1.6-1.7 avg volts, feels like it's nothing.

[MaxxxRacer]

power regulation is excelent on the grouper. It was designed by big toe specificly to overcome all of the issues the DFI's have with power regulation. ironically, the new revision DFI nf4 uses the EXACT same power regulation setup for the cpu and mem as the grouper board.

[Holst]

Sometimes mosfets can overheat surrounding components (especially voltage regulator) watercooling them might be of benefit in some circumstances.

I did a quick scetch of a cheap mosfet block.

Basically its just a bit of copper with some microbore tubing soldered onto it.

This wont work brilliantly, but will be MUCH more effective than aircooling and quiet.

Quite how you would plum this in without killing your flowrate is awquard.

I think the best bet would be to make a manifold to split the flow into two parts one 1/2" and one 5mm (microbore size) using this flowrate through the block would be pretty low, but you wouldnt need much IMO. And this method wont affect overall system flowrate at all if done right.

[Bloody_Sorcerer]

there had been some FET watercooling ideas tossed around at procooling a while back.
I still say back-of-the-board cooling is the only cooling for FETs thats going to be worth a crap, so something like a short NB heatsink inside of a copper pipe (brazed up and stuff) would be easy and incredibly effective. even just a copper plate and a flattened copper tube (more surface area) would be sufficient. these components are almost always rated to upwards of 150 C; they don't require stellar cooling methods. the only heat problems they cause is heating up other nearby components.
I also recall seeing someone make one out of silicone; it actually worked quite well (he molded it into the top of his mobo; search for the thread about it over at procooling i guess :-P )