Motherboard mosfet replacement/upgrades!

[GordonFreemanInTheFlesh]

As a hardware enthusiast I have been having this fascinating question of replacing stock motherboard/gpu vrm mosfets with larger more beefier ones that require little to no cooling plus get 2x or 3x the wattage capacity!
I do have a hot air station/soldering station and will be doing some experiments on some old am3 boards I have laying around.
I have looked online everywhere and it's very hard to find smc varieties of mosfets that don't s*ck and those that can be found are very weak, so I will have to go with vertical ones where you can actually install proper heatsinks on like the one shown in the image:

https://imgur.com/a/xC1TIcn

(Note this is just for reference)

 

[natr0n]

I know many have swapped caps, but never seen mosfets swapped.

There is a tpu user who loves swapping caps on older gear and gets good overclocks that you wold never think.

 

[eidairaman1]

I know many have swapped caps, but never seen mosfets swapped.

There is a tpu user who loves swapping caps on older gear and gets good overclocks that you wold never think.

@buildzoid

 

[natr0n]

@buildzoid

There is another dude too I believe @delshay

 

[delshay]

Yes I've been swapping MOSFETS for some time now. Sofar I've found "turn off/on" "rise/fall times" to be better than going for MOSFETS that give outright higher output. Saying this I have a collection of MOSFETS in the TO-252 package I've never tested. Just because a MOSFET has higher output, does not mean you will get that, don't be fooled by it because it will always be limited by bondwire/trace layout,

These are my favorite MOSFETS in the TO-252 package IPD70N03S4L-04 Infineon Technologies | Mouser United Kingdom. Guaranteed improved undervolt/overclocking for CPU when fitted to high & low side DDR1/DDR2 platform.

If i could find better MOSFETS for my GFX cards, yes I would change them, but I've yet to find better. I've only seen one or two repair users on YOUTUBE change just one or two MOSFETS for better ones other than changeling the lot with better MOSFETS. That's what I would do as I feel changing just one or two is wasting my time. Beside this I don't want a repeat of the fault,

 

[ipo3nk]

sometimes the pcb trace itself limiting the current capacity.

 

[delshay]

sometimes the pcb trace itself limiting the current capacity.

I already pointed this out, but it's the bondwire the MOSFET itself that's the limiting factor, this is why I also look at the "diode continuous drain current" in the PDF.

 

[GordonFreemanInTheFlesh]

I already pointed this out, but it's the bondwire the MOSFET itself that's the limiting factor, this is why I also look at the "diode continuous drain current" in the PDF.

I took the idea of replacing/upgrading the mosfets to lower vrm temperatures at stock speeds as I'm not a big fan of overclocking .

 

[Hyderz]

im no expert but what does upgrading these mosfet ultimately achieve for the motherboard?

 

[delshay]

I took the idea of replacing/upgrading the mosfets to lower vrm temperatures at stock speeds as I'm not a big fan of overclocking .

I don't think it lowers VRM temperature. ...& I need to correct myself as it does not improve overclocking, in some case it does but it's very small. The biggest change is in the required voltage device needs.

Example: If a CPU needs 1.525v to do 4GHz stable, change in MOSFETS will drop that to 1.5 or 1.475v. This is what's happening here. The chances of them blowing up at high temperature is also reduce compared to the original MOSFETS.

 

[GordonFreemanInTheFlesh]

I don't think it lowers VRM temperature. ...& I need to correct myself as it does not improve overclocking, in some case it does but it's very small. The biggest change is in the required voltage device needs.

Example: If a CPU needs 1.525v to do 4GHz stable, change in MOSFETS will drop that to 1.5 or 1.475v. This is what's happening here. The chances of them blowing up at high temperature is also reduce compared to the original MOSFETS.

Hm that does make sense although I might have to add that the only reason I wanted to try vertical mosfets is because of the way the heatsink mounts(with a screw)

[And here was I...picturing myself adding massive aluminium heatsinks to satisfy my lust for excessive cooling]

im no expert but what does upgrading these mosfet ultimately achieve for the motherboard?

Lower board temperatures around the vrm area and also better cooling for the undividual misfets since vertical mosfets have a screw hole for attaching directly to an aluminium heatsink meaning you can go overkill with cooling!

 

[delshay]

Hm that does make sense although I might have to add that the only reason I wanted to try vertical mosfets is because of the way the heatsink mounts(with a screw)

[And here was I...picturing myself adding massive aluminium heatsinks to satisfy my lust for excessive cooling]


Lower board temperatures around the vrm area and also better cooling for the undividual misfets since vertical mosfets have a screw hole for attaching directly to an aluminium heatsink meaning you can go overkill with cooling!

I see, you want to bolt each & every MOSFET to an individual heatsink. Yes this will work, however each MOSFET contact point will increase. YOU also have the danger if one is knocked over high & low side will come into contact. Personally I think it's too dangerous & fragile for such a mod. Your better off soldering copper shimm & then solder a small heatsink to the MOSFET which I have already done in one of my older projects which worked very well. It turns a non overclockable motherboard into a very overclockable one at high voltage. With this mod I tried to blow-up the MOSFETS with max voltage, but I failed.
If your not into overclocking, it will run cooler because the heatsink is now soldered directly to the MOSFETS. You need a downdraft CPU cooler for this mod to work.

EDIT UPDATE: Sorry your idea can work. Bend the drain leg backwards to balance the MOSFET so that it sits flat/level with the motherboard. Don't forget you need to cut all MOSFET legs shorter so that it sits low profile but standing up.. ....The middle leg should look like a backwards L shape with the two outer legs inline with it so that it can stand upright on it's own on a flat surface.

2ND EDIT UPDATE: You can be adventurous & bend the two outer legs forward L shape while the center one is reversed L shape, not forgetting it must be able to stand on it's own upright on a flat surface. This mod should keep the MOSFETS stable if you knock it around, but there is still the risk if you knock it around too much you could lift PCB traces.

Don't forget keep High-side & Low-side heatsink separate. They must never come into contact.

 

[DaveLT]

You shouldn't try it anyway. They have too high inductances to actually switch at high frequencies that enable them to be low in power dissipation in the first place!
Best case it won't turn on. Worst case? It hard turns on and explodes!

TO220 mosfets are built for old applications and some cheaper current applications that dont need high frequency switching.
Even 252 mosfets are also too old for modern motherboards. also known DPAK they are very not suitable

 

[delshay]

You shouldn't try it anyway. They have too high inductances to actually switch at high frequencies that enable them to be low in power dissipation in the first place!
Best case it won't turn on. Worst case? It hard turns on and explodes!

TO220 mosfets are built for old applications and some cheaper current applications that dont need high frequency switching.
Even 252 mosfets are also too old for modern motherboards. also known DPAK they are very not suitable

Looking back what the OP said, the user wants to fit larger ones as a replacement. Well it will work, but it's still needs to be cooled, but the users wants to experiment with no cooler. I don't think it's possible to leave the FET in mid air with nothing attached. A larger FET will not fit in the original footprint, ...if the user wants to experiment on old AM3 motherboards, they can do so at their own risk.

I do experiments here, cranking up the voltage watching the original motherboard MOSFET blow-up one by one, then I change them to better quality one & repeat the test. It stops blowing up, "dammit". But I have a large collection of FETS to play with.

Here you would not believe some of the strange things I get up-to, but some experiments I do keep & put on my main computer.

 

[DaveLT]

Looking back what the OP said, the user wants to fit larger ones as a replacement. Well it will work, but it's still needs to be cooled, but the users wants to experiment with no cooler. I don't it's possible to leave the FET in mid air with nothing attached. A larger FET will not it in the original footprint, if the user wants to experiment on old AM3 motherboards, they can do so at their own risk.

I do experiments here, cranking up the voltage watching the original motherboard MOSFET blow-up one by one, then I change them to better quality one & repeat the test. It stops blowing up, "dammit". But I have a large collection of FETS to play with.

Here you would not believe some of the strange things I get up-to, but some experiments I do keep & put on my main computer.

Speaking from my EE experience, leaded mosfets have very high lead inductances and other issues that led mosfets to have smaller leadless packages. In fact, components in general.

 

[Ferrum Master]

The greatest problem is to match them... You ain't a factory where you can match several rolls of them... the capacity and gain differs, not sure if the circuitry have means to fight parasitics and just rely on perfect match.

There are many variables... but none makes sense to really waste time on it.

 

[DaveLT]

The greatest problem is to match them... You ain't a factory where you can match several rolls of them... the capacity and gain differs, not sure if the circuitry have means to fight parasitics and just rely on perfect match.

There are many variables... but none makes sense to really waste time on it.

To drive them to frequencies possible with leadless mosfets, you need a WAY stronger driver that entails way higher power consumption on the drive or lose the efficiency benefits of higher frequencies.

As for matching, they are driven to either fully on (saturated) or off, there is no need for any matching as their Vgs off is fairly equal. It's only necessary to match if you are running mosfets in linear mode.

What is a problem however, is parasitics caused by leaded mosfets and slower mosfets because they aren't able to turn on or off in time and its that region that blows up mosfets with power switching because they suddenly dissipate a hell lot more power than they would normally

 

[delshay]

I can carry out the experiment here what the OP is trying to do. I have some big & I mean very big MOSFETS here. The problem before even starting, is getting such large FETS in a smaller space. I have to do a search & try to remember where I stored them. I also only have one motherboard to spare to carry out this experiment.

I like to point out when I was blowing up original FETS on motherboards it took around 1.8v+ to blow them up. The CPU is rated 1.55v Max but it actually survived up-to 1.9v before the FETS departed.
Flames can be seen as a hole is burned into the FET. ...After about the third or forth FET is blown, the motherboard shuts down. I carried out this experiment twice

As for what the OP wants to do, It will make the motherboard more fragile. Not sure if there is anything to gain from fitting bigger FETS without a heatsink.

 

[Ferrum Master]

What is a problem however, is parasitics caused by leaded mosfets and slower mosfets because they aren't able to turn on or off in time and its that region that blows up mosfets with power switching because they suddenly dissipate a hell lot more power than they would normally

Exactly.

While these days the difference is really small, there were some weird cases when you just replaced one mosfet in a VRM train for GPU(one phase blown). Despite being the same number, that critter acted weird, more hot or shutdowns. It was fixed changing all critters for the same batch ones... or you have a donor and put a known good one.

I have modded motherboards for OC purposes... LN2 mostly. There were boards were even place for the second mosfet was vacant. It should work well on older analog or hybrid style VRMs, haven't looked at modern digitals as those VRMs are beefy enough by default. Better results are just using Mosfets with lower RDS ON, not current or voltage... It still really works redoing some scooter ESCs where in the orginal place usually sit some fake/counterfeit china mosfets either way.

 

[silentbogo]

"Upgrading" mosfets isn't new, but you have to be very-very careful when it comes to spec.
I've done it quite a few times when fixing dead laptops and motherboard, and my primary consideration is always a lower Rdson, rather than higher current throughput. Less heat = better performance and stability. Higher current rating ≠ Better performance.
Plus there are many more considerations, especially when it comes to AM3 boards. First of all - quality of PCB. Most 760/770/960/970 boards have shitty power traces(including some "enthusiast"-level boards), only meant for handling sub-90W parts. Upgrading mosfets will do no good, because textolite will still burn and char once enough current passes through it.
The whole AM3 ecosystem was such a mess, I avoided it at arm's length. The only reason people bought it - is price.

Also, as @Ferrum Master noted, counterfeit parts are quite common. I get my new parts from very few good suppliers, and even then I may get some defective chips. That's why I have a sizeable stockpile of dead donor boards at work.