READ ! Did you too forget your mosfets ???

M

Main

Limp Gawd
Joined
Jul 26, 2009
Messages
135
I see watercooling how to's here - there and everywhere....

All these guides almost always have one thing in common - they "forget" to tell you that you remove a VERY important part of the motherboards cooling when going from air to watercooling.

What you do remove when going from air to water is the airflow caused by the CPU cooler's fan over the mosfets (Metal Oxide Semiconductor Field Effect Transistor - What is a MOSFET @ http://www.techpowerup.com/articles/overclocking/voltmods/21)

So you will be very wise compensating the removal of the CPU cooler with a fan - or better watercool the mosfets.

Why am I writing this ? Just got a mail from a guy that now has killed his second motherboard because of this - he says that his original heatpipe is still attached to the motherboard. Because he has no fan anymore this original cooling is now more to a hinder to get rid of the heat from the mosfets

If you see a heatsink on your motherboard it's there for a reason :)

Everybody (or most of us :D ) are equipped with ten excellent tools when it comes to find out which parts on the motherboard that need's to be cooled by air - your fingers.... too hot to touch - means in my book that it needs some sort of cooling.
 
yeah this is definately worth mentioning. Getting a DC signal down to and cleaned up for Vcore isn't as sophisticated a process one might think like. Ohm's law is in full effect with wave rectifiers. The thing about these parts is really that their characteristic curves are reasonably temperature independant, so the performance of the chips themselves doesn't so much depend on heat --at least in theory. When they get so damned hot solder is starting to melt, and PCB's are starting to warp, the performance of the chip being solid is all well and good but suddenly your starting to get bizzar things happening on your voltage. Things maybe CPU's arn't so fond of.

Actually now that I think about it the cleaner the signal from the PSU the less heat and more efficient the CPU voltage regulation should be... I'll run that one by my prof tomorrow.
 
Mosfets are cooled! I actually just redid this loop with motherboard cooling over the weekend:

cooling.jpg


The full coverage heatsink that came with this Rampage Formula board was fugly and temps were getting out of control (even with the included blower fan). The only downside was coming up with a way to route the tubing. I ended up splitting off and using smaller tubing to connect the mosfets and NB water blocks, as you can see.
 
You are exactly correct.

Having been married for 20 years you think I would be used to people ignoring me, I preach this constantly ! Even to the air cooled crew.

http://www.hardforum.com/showthread.php?t=1467205

Mosfet blocks are sweet, but remember, while the newer type solid caps are not nearly as affected by premature aging/failure as the older electrolytic, they still can have their life reduced or fail when subjected to the current/filtering demands of a high OC especially with a Quad. A fan, even a quiet slow one blowing around the CPU socket area is a very very good thing. Usually the PPL clock generator chip is in the same area as well.
 
Last edited:
Unfortunately I found this out yesterday. Air cooled i7 920, pushed it up to 4ghz, Vreg mosfets got too hot and blew.
 
One shall also be aware of that a heatpipe can stop work (stall) if the surroundings get to hot - a fan is also needed for heatpipecoolers...


"When the temperature is above the vapor condensation point of the working fluid, the vapor will not condense back to liquid phase. This causes the heat pipe to stall, because no fluid circulation occurs. "

In other words the heatpipecooler will when that occurs act as an insulator - and will speed up the overheating...

Quote from @ http://powerelectronics.com/mag/power_facing_thermal_solutions/

I always put a little fan on mine but frankly from the policy "it has to be more better that way". A really interesting fact and I "see it." Good stuff. If should follow that the the more delta T between the temp of condensation of the fluid and the pipe/surroundings the better, up to a point, it will work.
 
Last edited by a moderator:
Main said:
One shall also be aware of that a heatpipe can stop work (stall) if the surroundings get to hot - a fan is also needed for heatpipecoolers...


"When the temperature is above the vapor condensation point of the working fluid, the vapor will not condense back to liquid phase. This causes the heat pipe to stall, because no fluid circulation occurs. "

In other words the heatpipecooler will when that occurs act as an insulator - and will speed up the overheating...

Quote from @ http://powerelectronics.com/mag/powe...mal_solutions/

I always put a little fan on mine but frankly from the policy "it has to be more better that way". A really interesting fact and I "see it." Good stuff. If should follow that the the more delta T between the temp of condensation of the fluid and the pipe/surroundings the better, up to a point, it will work.

--------------------------------------------------------------------------------
Last edited by BillParrish; Today at 07:36 PM.. Reason: damn typos.... :D
Click to expand...

Thanks for editing my post.... :D guess you did it by mistake ;) ( it's no prob tho)
 
Main said:
I see watercooling how to's here - there and everywhere....

All these guides almost always have one thing in common - they "forget" to tell you that you remove a VERY important part of the motherboards cooling when going from air to watercooling.
Click to expand...

More than that. I've seen posts where somebody wrote that you don't need to cool the MOSFETs or the NB/SB.

What you do remove when going from air to water is the airflow caused by the CPU cooler's fan over the mosfets (Metal Oxide Semiconductor Field Effect Transistor - What is a MOSFET @ http://www.techpowerup.com/articles/overclocking/voltmods/21)

So you will be very wise compensating the removal of the CPU cooler with a fan - or better watercool the mosfets.
Click to expand...

OK. I'm open to ideas. (See below.)

Why am I writing this ? Just got a mail from a guy that now has killed his second motherboard because of this - he says that his original heatpipe is still attached to the motherboard. Because he has no fan anymore this original cooling is now more to a hinder to get rid of the heat from the mosfets.
Click to expand...

Nobody wants to lose a $300 motherboard, and possibly the $300 CPU and the $300 (or more) of RAM. :mad: So here is my question. I'm going to build my next system around the ASUS P6TD, which looks like this.

product.aspx


If the image doesn't show up, go here: http://usa.asus.com/product.aspx?P_ID=1H6537GW0fPVHNhO&templete=2 Then look on the right side of the page and click on the left thumbnail.

There seems to be one big, interconnected set of heatsinks and heatpipes for the NB, SB, and MOSFETs to the left of the motherboard, and then wrapped around the CPU on 3 sides. If you are watercooling, where do you place a small fan to blow air over all of this? Or, do you remove all or part of this setup and use water blocks? That of course adds complexity and expense to your loop.

Is it even possible to remove part of this interconnected set and leave the rest of it alone? :confused:

Thanks.
 
If I remeber correct - Zalman was one of the first to use heatpipes in their VGA coolers (the first ? ) it's some ten years ago - so I might be wrong :)

Anyway I do remember "the hype" that heatpipecoolers should be up to 100 times as effective as conventional cooling - "high tech used in space shuttles" aso.
I did doubt that the heatpipe coolers we use ar so effective as it was stated at that time, so since I am born curious - I did cut open the mentioned Zalman VGA cooler in both ends .
Took out the wick and flushed out the 3-4 drops of liquid.
Thereafter I did connect the heatpipe to a tiny 80 mm alu heatercore, a Alpha fan and a Eheim pump - and voila - a L O T better cooling :p

So much for 100 times better ;)
 
All things being equal the MOSFET's are the most resilient components on your motherboard. Most of them are rated to temperatures of up to 120C. Usually just the normal case fan you have at the back of your case produces plenty of airflow to keep them working just fine. Obviously it isn't bad to make sure that they aren't overheating, but this is a bit alarmist.
 
x509 said:
More than that. I've seen posts where somebody wrote that you don't need to cool the MOSFETs or the NB/SB.



OK. I'm open to ideas. (See below.)



Nobody wants to lose a $300 motherboard, and possibly the $300 CPU and the $300 (or more) of RAM. :mad: So here is my question. I'm going to build my next system around the ASUS P6TD, which looks like this.

product.aspx


If the image doesn't show up, go here: http://usa.asus.com/product.aspx?P_ID=1H6537GW0fPVHNhO&templete=2 Then look on the right side of the page and click on the left thumbnail.

There seems to be one big, interconnected set of heatsinks and heatpipes for the NB, SB, and MOSFETs to the left of the motherboard, and then wrapped around the CPU on 3 sides. If you are watercooling, where do you place a small fan to blow air over all of this? Or, do you remove all or part of this setup and use water blocks? That of course adds complexity and expense to your loop.

Is it even possible to remove part of this interconnected set and leave the rest of it alone? :confused:

Thanks.
Click to expand...

When I watercool nowadays I always do go fullcover + use a ir thermometer to expose the hotspots ( you dont need a ir therm.. your fingers are just fine exposing the hotspots).

Check ppcs or sidewinder for a fullcover solution for that mb.

This is watercooling --- It is even worse - much worse to dont have circulating air over the hotspots - and especially the cpu area if you do chilling or phase. The reason is of course that the airflow actually hinders condensation to a certain degree. it dryes it out (like when you have clothes hanging out on a mild summersday)....

It's all pure common sense really..
 
Erasmus354 said:
All things being equal the MOSFET's are the most resilient components on your motherboard. Most of them are rated to temperatures of up to 120C. Usually just the normal case fan you have at the back of your case produces plenty of airflow to keep them working just fine. Obviously it isn't bad to make sure that they aren't overheating, but this is a bit alarmist.
Click to expand...

We have now two dead motherboards on one day because of it ..... have you ever seen ir photos of a motherboard ?

These are old ir photos (the only ones I found now) - and the chipsets are redesigned to direct the heat outwards today - which they actually weren't at the time this pictures was taken (that's why we had fans on the back of the motherboards).
http://www.bigbruin.com/reviews/thermalimages/

So things are better today - no doubt about it. Yet motherboards still die...

From my side - this is a "better safe than sorry" advice :) - and not to scare or upset anyone :)
 
Last edited:
Upon further examination it seems my power supply failed under the overclock. It's currently unknown if my motherboard mosfets are still good or not, will post when I know for sure.
 
Main said:
We have now two dead motherboards on one day because of it ..... have you ever seen ir photos of a motherboard ?

These are old ir photos (the only ones I found now) - and the chipsets are redesigned to direct the heat outwards today - which they actually weren't at the time this pictures was taken (that's why we had fans on the back of the motherboards).
http://www.bigbruin.com/reviews/thermalimages/

So things are better today - no doubt about it. Yet motherboards still die...

From my side - this is a "better safe than sorry" advice :) - and not to scare or upset anyone :)
Click to expand...

Then you must be blind, because none of those thermal images (from 4-5 years ago mind you) show the MOSFETs getting anywhere close to a dangerous level. And just now you are talking about chipsets and not MOSFETs, so which is it? Was the chipset being adequately cooled when the person had the motherboards die? The chipset is going to be much less tolerant of high temperatures than a MOSFET is going to be. I am just pointing out the simple facts and you just keep changing your story around to keep it alarmist. If this were a real problem instead of just a small issue there would have been much more than this one person blowing out their board.

Perhaps it would do you some good to actually look at some datasheets on voltage regulators, here is one so you don't have to go looking. If you want to you could look up the ones on your particular motherboard too.

http://www.national.com/mpf/LM/LM317.html

Min operating temp : -40C
Max operating temp: +125C
 
Last edited:
When I watercooled my computer, I left mosfets off at first but with the way my case airflow is setup, it's getting too hot for my taste and causing lockups. I bought mosfet waterblocks along with the northbridge and it's fine since.

Your thread is very important to remind people about mosfet cooling even if it's the most resilient component. At least, make sure there is enough airflow around the cpu area.
 
i lost my old m3n78-vm because i didnt cool the mosfet, but some boards, like the crosshair II formula, comes with an optional fan that you can use if you are under water. OR you could just water cool them....
 
There are a lot of variables, sometimes a rear exhaust fan will keep enough air moving around to prevent serious issues. It just depends on the OC, the use of the cpu as far as the amount of load and the amount of time continously run at a high load, etc. etc.

Many many people are unaware of this issue and have never had a problem or dont keep a motherboard long enough for any reliablity issues to show up. But if you keep boards like I do, see sig, reworking all the cooling including the mosfets before installing a board intended for a high OC is a standard practice and cannot hurt and most certainly could help for the cost of a couple of hours and a few bucks for hardware and fans.

Thanks for editing my post.... guess you did it by mistake ( it's no prob tho)
Click to expand...

Yes, a common mistake I make, going for the "Quote" button I hit "edit/delete" in error and despite (most of the time, lol sometimes I do not notice right off) immedately exiting it still records my editing it even when nothing was changed. Sometimes people even get spellchecked for free.
 
Erasmus354 said:
Then you must be blind, because none of those thermal images (from 4-5 years ago mind you) show the MOSFETs getting anywhere close to a dangerous level. And just now you are talking about chipsets and not MOSFETs, so which is it? Was the chipset being adequately cooled when the person had the motherboards die? The chipset is going to be much less tolerant of high temperatures than a MOSFET is going to be. I am just pointing out the simple facts and you just keep changing your story around to keep it alarmist. If this were a real problem instead of just a small issue there would have been much more than this one person blowing out their board.

Perhaps it would do you some good to actually look at some datasheets on voltage regulators, here is one so you don't have to go looking. If you want to you could look up the ones on your particular motherboard too.

http://www.national.com/mpf/LM/LM317.html

Min operating temp : -40C
Max operating temp: +125C
Click to expand...

I knew them pictures were bad examples, I also know that there are ir pictures out there that showing temperatures exceeding the limit. But I just can't find them now.

I have myself killed at least three mobo's because of this, the two first were rma'ed - the last one I soldered on new mosfet's and it were ok again. This was with phasecooling and very high vcore.
After the third motherboard died and was repaired we started to look into it - and found that the mosfet's were extremely hot. After adding heatsinks and a fan - that board lasted until I threw it....
And I have never had problems with mosfets dying on me after that since I always cool them - one way or the other.

I reported this problem already in 1998 or 1999 when I used a phasecooling to cool down slot1 processors...

Stability is a issue too - a mosfet is not a single transistor but it consists actually of a lot of fet's - just like a car battery consists of several cells. You can therefore get internal hotspots in some of these fet's because of bad interconnects. If enough of them are affacted by this - you will see fluctating poweroutput, low power / spikes from the mosfet.

Now I am not a electrical engineer - so someone else might explain this somewhat better to you :)


What I have been "warning" about here is nothing new - it's well known - thats why I find it strange that mosfet cooling is overlooked in most of the watercooling guides .....


Why do you think they sell dedicated mosfet coolers ? with passive heatsinks, heatsinks with fans - and even waterblocks ?

Lastly - if one are a bit skilled with a solder iron it is not much difficult to replace a dead mosfet.
 
Last edited:
BillParrish said:
Yes, a common mistake I make, going for the "Quote" button I hit "edit/delete" in error and despite (most of the time, lol sometimes I do not notice right off) immedately exiting it still records my editing it even when nothing was changed. Sometimes people even get spellchecked for free.
Click to expand...

no prob - btdt ;)
 
Main said:
Now I am not a electrical engineer - so someone else might explain this somewhat better to you :)


What I have been "warning" about here is nothing new - it's well known - thats why I find it strange that mosfet cooling is overlooked in most of the watercooling guides .....


Why do you think they sell dedicated mosfet coolers ? with passive heatsinks, heatsinks with fans - and even waterblocks ?

Lastly - if one are a bit skilled with a solder iron it is not much difficult to replace a dead mosfet.
Click to expand...

No worries, I am a Computer Engineer so I don't need anyone to explain it to me. You are correct that MOSFETs performance can decrease as temperatures start to reach extremes. However they are still rated to perform within specifications up to their maximum operating temperature, which is usually 120C or higher.

My point still stands that MOSFETs are a relatively minor issue, even in most water cooling applications. They don't need a dedicated fan anymore than your BIOS chip does. They do need a decent heatsink and moderate case airflow. Unless you are running at extremely high voltages, like with your phase change setup, or have a poorly ventilated case because of an external water cooling setup you will be fine.

Trying to say that because products exist on the market shows that there is a legitimate and widespread issue is a red herring. There are products for everything, whether you need it or not.
 
Last edited:
Erasmus354 said:
No worries, I am a Computer Engineer so I don't need anyone to explain it to me. You are correct that MOSFETs performance can decrease as temperatures start to reach extremes. However they are still rated to perform within specifications up to their maximum operating temperature, which is usually 120C or higher.

My point still stands that MOSFETs are a relatively minor issue, even in most water cooling applications. They don't need a dedicated fan anymore than your BIOS chip does. They do need a decent heatsink and moderate case airflow. Unless you are running at extremely high voltages, like with your phase change setup, or have a poorly ventilated case because of an external water cooling setup you will be fine.
Click to expand...

I always run my computers as overclocked as possible, that most often includes vcore way out of specs for the cpu.
I recently saw ir photos of a 780i motherboard which his temperatures excedeed well over 100 degrees C, it's freaking irritating that I cant find those pictures just now :(
Anyway that motherboard did run on "middleclass" water and with passive heatsinks for the mosfets. He experienced better overclock after he did cool down the mosfets with water.

Myself also have a Abit NF7 motherboard with modified Thermaltake Big Water cooling in my workshop ( I don't have a hearth to get rid of it :) ) - the mosfets will almost instantly get so hot if I remove the fan that I will burn the skin off if I touch the mosfets. So I wouldn't dare to run that machine without active cooling for the mosfets :)

Electronic components will live longer if they are not permanently exposed for very high temperatures - that is also a well known fact.

Trying to at least hold the same temperature on the mosfets which they have with voltage settings at stock is in my book a minimum

Erasmus354 said:
There are products for everything, whether you need it or not.
Click to expand...

That is true :)
 
A finger test is a poor test considering you will get a burn nearly instantly at anything above 70C. So there is a margin where the heatsink will burn you but the MOSFETs will still be operating within specifications. (Keep in mind that if the heatsinks are 70C then the MOSFETs are hotter) As for the longevity argument, your components will be obsolete long before damage from prolonged high operating temperatures becomes an issue.

Don't get me wrong, keeping your MOSFETs cooler *will* help deliver cleaner power and may help you get a higher stable overclock. I am merely arguing against the notion that not actively cooling your MOSFETs will cause your motherboard to break.
 
Erasmus354 said:
A finger test is a poor test considering you will get a burn nearly instantly at anything above 70C. So there is a margin where the heatsink will burn you but the MOSFETs will still be operating within specifications. (Keep in mind that if the heatsinks are 70C then the MOSFETs are hotter) As for the longevity argument, your components will be obsolete long before damage from prolonged high operating temperatures becomes an issue.
Click to expand...

A finger test is not a accurate "instrument" of course - but it sure as hell gives a very good indication, don't expect that John Doe have a thermometer lying around :)

Erasmus354 said:
Don't get me wrong, keeping your MOSFETs cooler *will* help deliver cleaner power and may help you get a higher stable overclock. I am merely arguing against the notion that not actively cooling your MOSFETs will cause your motherboard to break.
Click to expand...

We do somewhat agree that one is best off having the mosfets cool - don't we ? ;)
 
Here's a good place to ask this. I have a my CPU, NB, SB and Vreg Mosfets on water for my EVGA X58 E758 board, and the mosfets are still getting to 66-70C. I am using Arctic Cooling ceramique as TIM, but I still think this is too high, considering my CPU is hitting 60 at the absolute max. And opinions?
 
66-70c is fine for MOSFETs. They're big beefy power transistors, not tiny sub-micrometer transistors, so they're much more tolerant to heat.
 
Dethred said:
Here's a good place to ask this. I have a my CPU, NB, SB and Vreg Mosfets on water for my EVGA X58 E758 board, and the mosfets are still getting to 66-70C. I am using Arctic Cooling ceramique as TIM, but I still think this is too high, considering my CPU is hitting 60 at the absolute max. And opinions?
Click to expand...

You are about 50C below the threshold for the mosfets (120C). Comparably your CPU is at 60C and has a maximum operating temperature of 100C, so you are only 40C below max there.


You are fine.


However, the temps are higher than I have normally seen for watercooled vregs. What block are you using, and have you checked the contact? You might have to use some thermal pads instead of paste because there is poor contact.
 
Yep you really have to watch board warp. Cranking down on the CPU water-block will "potato chip" (bend) the board in that area despite the motherboard hold down screws. Even a 1/10 of an inch is enough to make a lot of mosfets no longer contact or severely reduce the clamping pressure between the mosfet and its heatsink/block. I am glad to see more and more high end boards and water blocks with stiffeners under the mosfet sinks/blocks.
 
EdZ said:
66-70c is fine for MOSFETs. They're big beefy power transistors, not tiny sub-micrometer transistors, so they're much more tolerant to heat.
Click to expand...

well you are wrong... and right ! -- = a mosfet consists of thousands of tiny fets which sums up to a mosfet :)
 
Main said:
well you are wrong... and right ! -- = a mosfet consists of thousands of tiny fets which sums up to a mosfet :)
Click to expand...
Not quite: the structure of a MOSFET is made of many cells or lines of gate structures to maximise the switching area available. However, these are not individual FETs. A MOSFET is a single device, so the problems caused by heat in processors (quantum tunnelling breakdown between transistors) is essentially not an issue. The temperature limit for MOSFETs is the point at which the silicon itself will begin to deform.
 
Erasmus354 said:
You are about 50C below the threshold for the mosfets (120C). Comparably your CPU is at 60C and has a maximum operating temperature of 100C, so you are only 40C below max there.


You are fine.


However, the temps are higher than I have normally seen for watercooled vregs. What block are you using, and have you checked the contact? You might have to use some thermal pads instead of paste because there is poor contact.
Click to expand...

What temps are you used to seeing? I never have really payed attention to what others get with Vreg blocks since this is my first experiment with one. I am using the Bitspower board specific block for the E758. The board isn't warping visibly, especially since its clamped down by the stand-offs and screws. One issue I can think of is that I have a radiator touching one of the 45* barbs coming of the Mosfet block, and I guess may be pushing it off the vregs slightly. I guess I'll tear it down again and move to the thermal pads. 10+ more hours of work. :(
 
Dethred said:
What temps are you used to seeing? I never have really payed attention to what others get with Vreg blocks since this is my first experiment with one. I am using the Bitspower board specific block for the E758. The board isn't warping visibly, especially since its clamped down by the stand-offs and screws. One issue I can think of is that I have a radiator touching one of the 45* barbs coming of the Mosfet block, and I guess may be pushing it off the vregs slightly. I guess I'll tear it down again and move to the thermal pads. 10+ more hours of work. :(
Click to expand...

You don't have to, like has been said you are under what is safe. Generally I see watercooled vregs kept in the 40-50C range. You're just a little higher than normal.
 
Well - let us sum this thread up this way ? ; we all agree that active cooling certainly not hurt ? The old rule that electronic components work better the colder they are (until a certain point) still applies...

On the other hand - my preferences with tim on ie. mosfets goes in favour of shin etsu, ceramique, as5 - in other words thicker types tim.
There is virtually impossible to measure this - so this is only my feeling of what is correct regarding tim on such surfaces.

What do you guys think of this ?
 
You must log in or register to reply here.
Back
Top