For those of you wanting to play with the 20xx series and 30xx series, here are the efficiency curves of the MOSFET Clevo uses.

[Meaker]

[MS-16L13]

Before remodeling

After remodeling

 

 

 

[MS-16L13]

3dmark

[MS-16L13]

A brief analysis of the power supply of the CLEVO 30 series MXM graphics card.
A total of 11 phases of power supply, 7 phases to the core, 2 phases to the memory, 2 phases of auxiliary power.
After multiple data checks Mos tube parameters stable 50a maximum 60a, inductor isat 70a irms 45a.
This time the question arises, the discerning eye can see at a glance that the parameters of this inductor has a big problem. This package of inductors can have such high parameters, in fact, there is really no, and the false label is very serious. The real isat may only be about 30a, and the inductance drops very quickly, and the datasheet does not match.

 

[MS-16L13]

Original inductor test 1khz frequency 19a when the temperature drift began to appear What concept isat and irms all false mark more than 30%.
After replacing the new inductor without heat dissipation 1khz test 30a temperature 76 degrees inductance value no change, datasheet isat standard 37a 20% inductance value decay fully met the standard. Heat gun heating inductor over a hundred degrees 1khz test 25a still no significant change.

 

[Meaker]

6 hours ago, MS-16L13 said:

Original inductor test 1khz frequency 19a when the temperature drift began to appear What concept isat and irms all false mark more than 30%.
After replacing the new inductor without heat dissipation 1khz test 30a temperature 76 degrees inductance value no change, datasheet isat standard 37a 20% inductance value decay fully met the standard. Heat gun heating inductor over a hundred degrees 1khz test 25a still no significant change.

 

 

Did you do the soldering yourself?

 

What values did you choose for the shunt resistors and did you update the phase count resistor network?

 

Since I've updated my cooling I only get to about 57C at 1900Mhz core clocks during cyberpunk so I'd be interested in re-enforcing the VRM and changing the power limits.

[jaybee83]

19 hours ago, MS-16L13 said:

Before remodeling

After remodeling

 

 

 

 

impressive stuff, such a clean job! what kinda improvements, if any, are u seeing before and after?

[Meaker]

9 hours ago, jaybee83 said:

 

impressive stuff, such a clean job! what kinda improvements, if any, are u seeing before and after?

 

Looks like 10-15% over my simple shunt mod.

 

I'd guess from my 1900-2000 to 2150mhz type clocks.

[FTW_260]

@MS-16L13 Hello! do you have any information about which models of components were installed on the board?

[Meaker]

20 hours ago, FTW_260 said:

@MS-16L13 Hello! do you have any information about which models of components were installed on the board?

 

Same chips, just more of them.

[FTW_260]

33 minutes ago, Meaker said:

Same chips, just more of them.

I mean these:

 

I'd like to try shunt mod my RTX2080S like I did earlier on RTX2060 but I'm afraid 200W is the limit for the board(

[Meaker]

Probably similar values, I think it was more the frequency response he was looking at.

[FTW_260]

@Meakertell me please, do you know something about the power limit of the mxm connector? Is it safe to increase consumption to at least 250W?

[Meaker]

Most of this comes from the AUX connector IIRC, at 19V it's mostly just the input voltage so does not have a huge amount of filtering before it hits the card. So it's more about the capability of the MOSFETs on the board.

[Sniffy]

Meaker have you filled in your 3070s VRMs? I am wondering how you're moving 180W through your board with 4 MOSFETs for the GPU. Take away memory and aux power and you're looking at at least 130-140W which would push these components above 25A continuous which is their rating. The datasheet suggests there is current protection, is it allowing this limit to be exceeded if temps are fine? 

[Meaker]

Shunt modding has reduced the board limit and I liquid cool my system with a hybrid cooler. Also does the GA104 have separate core and uncore power planes? I think it's small enough to not need it so it's 6 + 2 phases so 6 for core and 2 for mem. So I'm riding pretty close yes but not quite over.