There is no Spoon ..... there is no shunt mod.
I am 35 years in electronics repairs, I did watch on YouTube one experiment of how by manipulating the shunt value, the power limiter circuitry changing it behaviour.
The so called benefit was 2FPS.
All Kids screamed that this is a modification worth having.
I think, and believe it, that this is a plain stupidity.
But the man which did the experiment, he worth the credit as tester which he tried something new and his content it is not boring.

 

There is no Spoon ..... there is no shunt mod.
I am 35 years in electronics repairs, I did watch on YouTube one experiment of how by manipulating the shunt value, the power limiter circuitry changing it behaviour.
The so called benefit was 2FPS.
All Kids screamed that this is a modification worth having.
I think, and believe it, that this is a plain stupidity.
But the man which did the experiment, he worth the credit as tester which he tried something new and his content it is not boring.

How is it not worth the effort when everyone on this entire forum, and everyone everywhere, is complaining how clearly power limited these cards are? Like crank the slide to +20 MHz and instantly it's power limited.

 

Massive improvement for me going from 350W to 520W in benchmarks on my 3090 XC3.
In games depending on the game thats 5-12fps more which really matters when your after 4K max settings.

Ampere cards a really power gimped. Nvidia knew this when they decided to go with samsung 8nm and the cards should of really release with base power of 400W for 3080 and 480W for 3090 since you have perfect performance scaling up to that point.

But back to the question.
I guess the most proper way would be desoldering all the shunts with hot air and replacing them all. That way you know you have perfect balance between all of them and the bios will work as intended other then the card drawing more power because its reading is faked.
Then next best way would be soldering a extra stack ontop. Which is less perfect since you have the extra restance from each solder joint.
Followed by the conductive paint methods (which is even less perfect since different thicknesses of paint and pressure will effect resistance even further)

On mine I went with conductive paint method so it could be undone if the card failed or a better bios came out (Which it did).
Cleaning up the conductive paint was time consuming but you couldnt even tell it had been done after the clean up and iv been running the 1kw bios but it looks like ill be reshunting again when the new bios come out with resizable bar support.

 

Power gimeped? 520W is insane. 7nm would have done nothing. Its all down the crazy large CUDA cores and power-hungry memory. 3080 uses 110w just memory alone. Nvidia would have save a lot of power and gotten a lot of perf if they had gone HBM2.

 

People claim overclocking, or modifying components, isn't worth it but honestly you are spending very little to nothing for a gain in performance. So for my it is almost always worth it. This all assumes you know what you are doing and don't break anything.

 

I also wondered if the hot glue trick works. That is, stick the new shunt on top of the old one and glue it into place with hot glue.

 

No fuses on the pictures you posted.
Dont hot glue it, thats idiotic.

 

No, the pictured card does not have fuses but my 3080 XC3 does. It has 20A fuses on the +12 ports and a fuse of an unknown value on the PCIE slot.

So you think the silver conductive pen is the best bet for a mod that can be cleaned up for warranty purposes? I am wondering how thick you apply the compound, whether there is a risk of it dripping down on the PCB and what the new power cap would be since the resistance is unknown.

Another idea was to use conductive paint to try to 'glue' the resistors on top. That way you at least know the exact power you're working with. Something like this:

https://www.amazon.com/Bare-Conductive-Electric-Paint-10ml/dp/B00KBXT6JW

I guess alternatively you could also use that paint and just run it over the shunt as well.

 

Use MG 842AR.
1 layer = 15 mohm.
2 layers = about 13 mohm.
5 layers = about 10 mohm.

It drys hard.
Other wise use it between the contact patch of the shunt you want to stack ontop.

Both I and Falkentne have written guides on here how to do it.
And I also released this tool for working out what your new power limit would be.

GitHub - bmgjet/ShutMod-Calculator: Work out what shunt values to use easily.

Work out what shunt values to use easily. Contribute to bmgjet/ShutMod-Calculator development by creating an account on GitHub.

github.com

 

On a 3090 FE my PCIE makes at about 95watts, card drawing 550 total. Not sure how other cards split PCIE vs pins but you could test your average ratios and calculate your expected usage with the calculator then limit power so PCIE doesn't get close to 120 watts.

 

For reference, here is the 3080xc3 ultra gaming (taken when I was putting on the AIO)

 

For reference, here is the 3080xc3 ultra gaming (taken when I was putting on the AIO)

View attachment 2474136
View attachment 2474137

I see you! M-hmm.

 

sigh

 

Is it really necessary to mod the shunt for the PCIE slot? Or will avoiding it affect the card's load balancing between that and the 8 pins?

 

On the 3090 FE it was recommended to shunt the PCIE slot as well. The issue may be that if the card requires a certain ratio of PCI to pins then even if you shunt everything else the PCI slot may limit it.