Why don't we see more people cooling their mosfets?

[mad mikee]

Just saw the link about someone building a power reg cooling block for their NF7-S. Water cooling power regs

Even I could make up some variant of this. Anyone want to design something simple and easy to make? My first variation is to use 1/2" soft copper tube and kink it slightly to weave between the coils/caps.(more contact w/ copper plate and create some turbulence.) Will be adding it to the GPU - NB loop in my system.

Must be easy to make and non-restrictive of course

[jaydee]

Originally posted by mad mikee
Just saw the link about someone building a power reg cooling block for their NF7-S. Water cooling power regs

Even I could make up some variant of this. Anyone want to design something simple and easy to make? My first variation is to use 1/2" soft copper tube and kink it slightly to weave between the coils/caps.(more contact w/ copper plate and create some turbulence.) Will be adding it to the GPU - NB loop in my system.

Must be easy to make and non-restrictive of course

Depends on the board really. It is a pain in the ass in any event. The 2 newest boards I have, Asus A7V8X-X and the Abit KD7, both have mosfets 1/3rd the size of normal. And they are packed inbetween caps. Also I wouldn't use copper pipe inbetween the copper wire wound coils. Probably short your board out. Not sure though.

This is easier to do wih boards that have large mosfets though. Some boards even have heat sinks already attached.

[Tempus]

great, so now its cooler.

Did it help the OC?

Make it more stable under load?

Anything??

[mad mikee]

Reduces overall heat content in the box? (definitely!)
Anyone want to speculate how many watts of heat here (80-90c * 4-6 chips?)

Perhaps improves power quality? (Guessing yes, will know more after I try).

With all the precision stuff going on here, this shouldn't be that hard for many boards. However there are many it is impractical w/ such as mentioned above (small chips) and where they are split up.

Fortunately mine is one of the ones I think will work w/ this . Time will tell....

[Since87]

Originally posted by mad mikee
Reduces overall heat content in the box? (definitely!)
Anyone want to speculate how many watts of heat here (80-90c * 4-6 chips?)

I noticed somewhere that AMD 'assumes' 85% efficiency for the Vcore regulator. (Of course that's assuming stock operation.)

Assuming 85% efficiency holds for an overclocked mobo...

If the CPU is dissipating 90 Watts, then ~16 Watts or so might be dissipated in the Vcore regulator. Half of that 16 Watts may be dissipated in the MOSFET's and the rest in other components.

[jaydee]

Originally posted by mad mikee

Fortunately mine is one of the ones I think will work w/ this . Time will tell....

If you have larger mosfets and they don't have other parts inbetween them then you biggest hold up is mounting. You will have to epoxy them on and therefor it maybe a one time deal. If it doesn't work right the first time you stuck with it.

I measuered the temps the other day on my Abit with a 1700+@1800mhz 1.80Vcore under load the mosfets where 45C, the round things with wires coiled where 55C, and the caps in the area where 40-43C. Thats a lot of heat and it isn't even pushing the board yet. I might try and air cool the mosfets to try and stabilize the Voltage lines. There is an other article at OC somewhere that they showed some stabilization and some minor OC gains.

[mad mikee]

(runs off to search for inductor heatsinks ....- :shrug: nothin)

So anyone know about how to put heatsinks on inductors w/o mucking them up (em fields, etc) or is it time to build my own shroud and cool like OTIS, etc)
(hmm runs off to google that... finds MORE WC on the backside of the same board Same guy, different side of MB. )

So not a lot out there (so far in searching at least). He did say his power stability was better (missing the drop under load he got b4 (sounds very familiar)). I think the backside seems to give more bang for the buck but I can't do it until mikeputer case 3.0 (early next year when I build / design my own) But you can be sure that this will factor into that design.

In the meantime any other ideas water or otherwise will be appreciated.

[rocketmanx]

I cool the mosfets, clock generator (when I know where it is), and the north and southbridge.

The big problem with mosfets is they are all over the board usually. they also don't lay in one direction either, they are sometimes angled and as someone else has already said, they are sometimes packed very tightly inbetween other compnents.
For the new style mosfets (smaller ones) and the old ones, I found bga ramsinks work perfectly. I use the old style vga ramsinks on longer components like the clock gen. My Gigabyte has sinks on everything I could find. I figure it can't hurt and with ramsinks it only takes a short amount of time to do with AS Epoxy.
With chrome or copper sinks it looks pretty awesome too

[killrtech99]

Originally posted by jaydee116
I measuered the temps the other day on my Abit with a 1700+@1800mhz 1.80Vcore under load the mosfets where 45C, the round things with wires coiled where 55C, and the caps in the area where 40-43C. Thats a lot of heat and it isn't even pushing the board yet. I might try and air cool the mosfets to try and stabilize the Voltage lines. There is an other article at OC somewhere that they showed some stabilization and some minor OC gains.

Yup. I can cross-coroborate Jaydee here. My IR thermometer measures nearly identically temps to his on mosfets and coil things (what the hell are those things called!). I'm probably going to Dremelize some old ramsinks and stick them on the mosfets and attempt to get a little fan going in the area to keep convection up.

[iroc409]

i'd be kinda careful around those coil things tho. this sounds like some sort of cooling on the power control devices is a very good idea.

i haven't played with the ones on the motherboard, but i've seen similar devices on other power supplies. i was repairing a lappy psu one time, and had plugged it in to test the output. well, the thing was out of its case, and i got my finger on the coil.

i got _quite_ a shock out of that... so i'm sure a lot of power is running through those, and possibly touching copper and the like to those things could be rather dangerous. i noticed he does not have them touching in the article, and points that out. probably for good reason. i don't think i'd mess with actually cooling those.. sounds like he got good results without.

on an idea for cooling the backside of the mainboard tho, here's an idea. see what you think of it.

on the back of the radeon 9700 cards, there's a small aluminum plate. it's actually used to cool something back there (i believe it's the card's own voltage regulator).

what if you made a much larger plate, and made contact patches (preferrably aluminum or copper to match) for the back of the mosfets, and had a large, relatively thin plate cut out on the back of the motherboard? it would be thin enought to fit without modifying the case, and perhaps you could even insulate if from the mainboard with a thin layer of neoprene or something.

this way you could make it fairly large... possibly even the size of the board, and it might be large enough to dissipate a good amount of heat. the only thing i'd be concerned with there is the insulation on the back would probably have the potential to cause other things to heat up as well. maybe making a large enough plate, or some sort of modifications, you could connect it to any other devices that get hot on the back as well.

dunno if it would work, just brainstorming. i figure the less you have to cut up the case, the better.

[BladeRunner]

There was a topic about powerfet watercooling a while back, ifyou haven't seen it........

http://forums.procooling.com/vbb/sho...threadid=7933&

[bigben2k]

Given the heatload (thanks to Since87) I'd try this out with smaller tubing than 1/2: 1/4 ought to be plenty.


As for cooling the coils (aka inductors, aka "doughnut shaped component"), there is no standard solution (that I'm aware of): the very nature of the coil doesn't give you a surface to mate a heatsink.

BladeRunner however extended a copper "arm", which was thermal epoxied to the top of the coil, in his PSU water cooling project.

If you go that route, you need to know this: the wire that is wrapped to form the coil, has a varnish coating on it, so that it's not just one "blob" of metal; the curent actually "turns" round and round. That varnish is easily scratched, exposing a live wire.

[BladeRunner]

As bigben2k says it's worth considering the short out factors. I didn't actually glue anything, being concerned about short out I used a mica shim....... I'm not a great lover of gluing anything if it can be avoided, although I had to give in and glue my mobo powerfet block & southbridge sink on as I couldn't see any other realistic options :shrug:

from the Psu build update on my site.......

"This copper bracket was attached to the waterblock via some screws with artic silver II paste, as a thermal interface. The other end of the bracket rests on the top of the coil with a thermal mica shim to prevent any electrical connection between the earthed water-block and coil, to ensure good contact a cable tie is also used around the coil and bracket. This simple, but admittedly agricultural solution has resulted in a max coil temp of around 43C to 45C. (55C when the high summer ambient was nearer to 40C)."

I also considered using the ceramic top and bottom from a small TEC to sandwich the hot PSU coil, but it was a little to buried by other components to do this easily.

I can't speak for all mobos, but on the NF7-S 2.0 I'm using the coils tend to be heated more by the fets heat, and the general heat soak of the whole area of the pcb. I found the caps around the fets actually got hotter when I tried this sink, and it did little to help the fets, (in a zero airflow system)



The sink solution failed to solve the heat issue to my satisfaction, but a simple water-block seems to work just fine not only cooling the fets, but preventing heat soak from them in effecting the area and all the other associated components.

[Blackeagle]

Go Newegg and search for Dynatron 1U copper heat sinks. Costs only $8 and cut it up for some nice easy to attach heat sinks.

With a side blower in your case of even modest size you should be all set.

If you want the back cooled as well, cut a opening in the motherboard tray and then install a fan on the back case wall as a blower there as well.

These mods will do all the cooling you should need in these areas and will both cost less and be easier to do.

Only reason I can see for water cooling of these components is if you are, like BladeRunner, in search of silence.

[killrtech99]

Originally posted by Blackeagle
If you want the back cooled as well, cut a opening in the motherboard and then install a fan on the back case wall as a blower there as well.

I think you meant cut an opening in the motherboard tray


I agree with you though. You'd have to be WAY into quiet PC to want to watercool the mosfets. But eh, if peeps wanna....

I thought you all might get a kick out of this. It's thermal imaging of an Epox 8rda+. My favorite quote from the site:

Mind you, being a resistor it is probably supposed to be hot and it is exceptionally tiny, therefore if you ever consider heat sinking this, even for a moment, you would probably be advised to attend the next meeting of Heatsinkers Anonymous.

Heatsinkers Anonymous!! No, we're all on Procooling trying to cool off!

http://homepage.ntlworld.com/bubo/index.html

[Blackeagle]

tray????



oooppps! Thanks, I fixed it!

In spite of my referancing some low cost ways to do this with air, and equally effectivly, I still love the finely detailed and lovely work done by BladeRunner.

Just look at that mosfet cooler! What a masterfull touch!

[iroc409]

ok, looks like nobody liked my other idea, and i was thinking about it last night, and came up with another solution that's still fairly simple but doesn't include watercooling.

now this will only really work on an aluminum case, if it works at all. again, it's just a thought i'm throwing out there.

what if you made small copper (or aluminum) blocks, cut the size of the back of the hot parts on the board (obvisously we're focusing on the fets here).

make the thickness of the blocks the thickness between the board and the mb tray (case). mount the blocks on the back of the board using frag tape or some thermal epoxy. then toss a little thermal paste on the other side of the blocks, and mount the motherboard. the thickness has to be pretty exact, so it contacts properly.

now what i'm thinking here is this would use the aluminum case to your advantage. the aluminum case parts would be used as a gigantic heatsink, providing what i would think to be a considerable amount of silent heat dissipation.

you could probably use ramsinks or something similar as well, as long as they were the right thickness, but i'm not sure as that would work any better than just solid metal blocks.

[iroc409]

Originally posted by BladeRunner

dude, you do some seriously kick-ass work, you ever build blocks and stuff for other people, or just for yourself?

[BladeRunner]

Hell I think you guys could be right and I need to visit Heatsinkers Anonymous!!

Guess I'm just a waterblock making junkie....... and when something is too hot I can get carried away



Seriously I'm obsessed, and sinks should work fine if you have some reasonable airflow. That's the big if because I've found in "fanlessness" things that often don't get hot or even warm, can overheat if there is very low or no airflow......



iroc409

Thanks .....not wishing to go off topic here, there's a little about my plans in my replies in this topic HERE

[iroc409]

i don't blame you at all, if i built stuff like that i'd go crazy with it myself.

my idea wasn't necessarily banking on the airflow behind the mb, but the physical size of the heatsink. the mb tray itself is fairly large, plus the fact that it's fairly well attached to the main part of the case. with that in mind, i was thinking it would do reasonably well for heat dissipation.

i've never tried it, it's just something i've been thinking about. i think i might try this, see if it works. maybe combine it with cooling on the topside as well.

[killrtech99]

Originally posted by Blackeagle
Just look at that mosfet cooler! What a masterfull touch!

Oh yes, no doubt. It's absolutely beautiful! Almost seems too purty to use! Every bit a work of art as well as an awesome piece of equipment. I really stand in awe (and jealousy) of many of the block makers on this forum. Talk about things that make you go

[killrtech99]

QUOTE]Originally posted by BladeRunner
Hell I think you guys could be right and I need to visit Heatsinkers Anonymous!!

Guess I'm just a waterblock making junkie....... and when something is too hot I can get carried away [/quote]

Nah, you don't need HA, you are simply raising this hobby to an artform. Onwards you!!!

You do raise an excellent point about going completely fanless, there certainly are a lot of things that need cooling in the absence of air flow. Even though I've got a fair amount watercooled (I guess the "big three") I still have tons of fans and my case is still really quite noisy. Just not as noisy as when I running a Delta fan. I'm currently in the process of setting up 3 (yes 3) rheostat bays to tune down all of the fans in my system to at least get it fairly quiet. I suppose if I could make waterblocks like you......

And apologies to Mad Mikee for some threadjacking. Just feeling the love man!

[V12|V12]

For passive cooling I just use a bunch of old SUN/HP server Heatsinks I chopped up to fit the Vregs, MOSFETS, and Caps if I ever notice they are hot. As for gluing, I use a mixture of AS-3 and some mild strength epoxy, depending on the ratio of epoxy to HS-compound. I did 1:2 and it was strong enough to hold firm, but is still removable if stressed hard. But nothing has come off, even with a purposeful bump to it. Water cooling all of those subcomponents, while innovative, seems like a waste, since a nice passive HS does WORLDs difference and since those components safe temp is WELLL WELL above what it would be w/o the HS, bleh, just add one glue, forget about them. What I have found to be the LOUDEST fan component in my comp is my PSU! I removed the noising and restrictive 80mm from it and now it's just a lone 92mm, which for some reason, while not running FULL RPM, it's still noisy. Much more so that the other 92s I have... I thought about maybe adding a thermistor so the fan will them run crazy loud ONLY when needed, but since it's the most vital part of my system, I can't be taking any chances with the PSU, cause if it has probs, it can destroy other components... Maybe I'll hook up the fan to a fanbus and manually control the temps based on the exhaust temps I get... Okay it's friday, too many posts, drinking time! WOOT

[kai]

I choped up an old 1700+ stock heatsink the big aluminum one you got with the old palimino's going to be epoxying them on soon , like when i get my new water block and new tubing but i havent decided which block yet , stuck between rbx and WW

[BladeRunner]

V12|V12

While I'd agree with you in most situations, I tried a passive sink as said, (using AS epoxy cut with as3 so I could get it off), and it didn't help much at all, actuallytended to heat up other components more.

It is hard to get across the fact that NO fans in the system is very different from even just one in the PSU. I know these components can be rated for high temps, but its still not good practice to run them overly high...... Do we take what AMD state as a maximum safe CPU operating temp as acceptable? I certainly don't, and I don't want anything in my PC at anywhere near 80C, it adds heat to the case ambient, but much more than that It just annoys me