I don't think the rise/fall (turn on/off) times are comparable. The test conditions are different...

As long as the voltage and amps specification is double what is to be expected on the mainboard, and power dissipation is in the same region, it should be fine. I would use the STD70N02L.

Thank you very much for your advance - I found another one which is in stock https://www.ame.cz/AOD452A-D452A-a254613.pdf , which is faster but has a lower maximum current - 55A instead 60A (Nikos). There will be no more than P4 Prescott 3.0 GHz on the board, nothing more demanding. Would this be enough, or would it be a risk?

Ydee wrote on 2025-11-20, 15:58:

Thank you very much for your advance - I found another one which is in stock https://www.ame.cz/AOD452A-D452A-a254613.pdf , which is faster but has a lower maximum current - 55A instead 60A (Nikos). There will be no more than P4 Prescott 3.0 GHz on the board, nothing more demanding. Would this be enough, or would it be a risk?

You could approximate it like this: The CPU has a TDP of around 100W and a voltage of 1.40V, so 70A. Your board has a three-phase power design, so ~70A / 3 = 23A. This is what each FET must be able to deliver. In practice, it's more than that, because this simple calculation does not account for ripple, switching and heating losses, so you need to have a safety margin. 55A should be well above that.

But that current rating alone is not enough, there is also a temperature rating for the internal silicon die. 55A could generate little heat or lots of heat inside, depending on the internal resistance, which depends on the operating conditions - it's a lot of complicated parameters and math. You also need to take into account the maximum power that your motherboard can dissipate - whether the FET dissipates 2W or 4W can make a big difference. To avoid overcomplicating this, look for a FET similar to or better than the old one, based on the following parameters:

-comparable current rating (gives you a general idea of the robustness)
-Rds(on) at Vgs=10V that is ideally lower or at most 20% above the old FET (this is very important, even tiny increases produce lots of added heat)
-comparable gate threshold voltage
-comparable Vds(max) rating of around 25V (higher=more robust, but also less efficient)

Some parameters depend on whether the FET that you replace is the high-side FET (the one connected to 12V) or low-side FET (the one connected to ground). The high FET needs a low Rds(on), a low gate charge Qgd and a short rise time. The low FET needs a low Rds(on) above all else, a short rise time is nice but second to that.

The one you chose should be a good replacement, just be aware that when you buy MOSFETs from ebay and especially Aliexpress, there's a good chance of them being fake. You need to test them before, a fake one could destroy your board.

The 130ns rise time in the ST FET is indeed puzzling, because the gate charge parameters would suggest a much lower value. Can someone explain this?

Thank you for the detailed explanation, I appreciate it. On the board are on phase P70N02LDG as high side and P60N03LDG as low side FETs.
I'm afraid even low is short-circuited and so I'll be changing both and praying that something else isn't destroyed.

P.S. Those ST FET's are in phases (on the low side, on the high side there are 90N03L) on the board from the Dell Precision T3400, so I thought there would be a similarly designed VRM for the CPU.

Another question, are you sure that the FET you short circuited is really damaged? What happened with the board at the moment you short-circuited it? How did you determine that it's damaged?

About the ST FET, it would make sense as a low-side FET, but a slow rise time would be bad for the high side. The voltage at the high side FET transitions from 12V to 0V during switching, and the slower this transition is, the more power is wasted inside the FET, because P=V*I. The low side only needs to switch a tiny voltage just above 0V, so even if it switches slowly, it doesn't waste much in absolute terms.

If you use it as a high-side FET, it might work, but waste more power, possibly so much that it fails or the board overheats. You would have to calculate the switching loss to estimate the impact. I wouldn't know how, but I'm sure there's a formula for that.

Edit: Here it is, https://www.ti.com/lit/an/slyt664/slyt664.pdf on page 23 and 24. It's not very simple unfortunately. But seeing that the formula relies on gate charge and current, and these parameters are similar between your original and the ST FET, it shouldn't perform much worse. In fact, their rise times should also be similar based on that, I still have no idea how it measures 130ns vs. 7ns.

The PSU has shut down, and after the boot is attempted again, the diagnostic card flashes momentarily on display, and the fans move and turn off, including the PSU. When measuring the resistance of a phase with a short circuit, there are several ohms on all S-G-D; FET on the other phases does not. I suppose I shorted G to D (9V to 1.10V when I measured another phase.)

Finally done, replacing both phase FET's worked. Of the currently available, I chose https://www.ame.cz/NTD60N02RG-T60N02RG-a14331334.pdf for the high side and https://www.ame.cz/NTD4806NG-4806NG-a252405.pdf for the low side. According to the datasheet, while they are not as fast as the original NIKOS, there should be no problem in the other parameters, and they also have Schottky diodes instead of ordinary ones that had NIKOS. They seem cooler than the existing ones on the other two phases, but I still intend to plant them all with passive heatsinks.

So I'm hoping the VRM will have a good fight with it and the board will be stable even under full load.
Thank you for your help and advice.

real life history: asrock p4vt8+ fries same mosfet (top left near lpt) if you use for long period cpus >p4 prescott 3 ghz

if it happens to you, maybe that motherboard are not recommended for >3ghz, because it has fewer power phases. i mean, it will happen again

Thank you for the experience, but I damaged the mosfet myself with my own clumsiness, it didn't burn with overload.
The board officially supports dual core CPUs, so single-core Prescott operation shouldn't damage it. We'll see in the future.

Glad to hear that it's working and nothing else was damaged. Really nice job on picking the parts and the soldering. I bet it felt great when you realized that you not only repaired it, but even made it run better than before.

Thank you, but it was only possible because of the help and advice of you, the members here.