Overclock.net banner

201 - 220 of 382 Posts

·
Registered
Joined
·
93 Posts
Do you need to remove the conformal coating when soldering? And also doesn't the solder material affect the resistance? And why is it safe to stack a 5mohm, wouldnt the 2x power draw blow the 20amp fuses?
No, Look at 7:00 of this video by debauer, I used his method to solder. Was really easy. If you are not confident, you can use kapton tape to mask the PCB.

 

·
Registered
Joined
·
249 Posts
a question, shunting the card (3090 FE) will bypass also the software controlling slider in MSI AB?
i was planning to shunt mine FE stacking 5mOhm 3mOhm, i read this cause a total board power of about 800w but:
i'm on custom loop and have another power connector on my z390 Dark PCIe lines.
Wondering if lowering the PL slider on MSI AB i can control the max PL output
 

·
High Clocker
Joined
·
3,436 Posts
The power slider still works. Its just the card thinks its using half the power since the resistance of the shunts are changed.
So 50% slider would be what use to be 100% with a 5 mohm stacked.
 
  • Rep+
Reactions: mattxx88

·
Registered
Joined
·
58 Posts
No, Look at 7:00 of this video by debauer, I used his method to solder. Was really easy. If you are not confident, you can use kapton tape to mask the PCB.

So did you put solder on the shunts directly without removing the conformal coating? I wonder if that would affect the resistance
 

·
Facepalm
Joined
·
8,623 Posts
So did you put solder on the shunts directly without removing the conformal coating? I wonder if that would affect the resistance
This is a very good question. Are there cards that do not have conformal coating on the edges of the shunts? And how can someone tell if it's actually conformal coating or just the way shunts are?
Because conformal coating should prevent a stacked shunt from even working if soldered, right? Because the FE cards have the edges lower than the middle while the AIB's are flush...
 

·
Registered
Joined
·
93 Posts
This is a very good question. Are there cards that do not have conformal coating on the edges of the shunts? And how can someone tell if it's actually conformal coating or just the way shunts are?
Because conformal coating should prevent a stacked shunt from even working if soldered, right? Because the FE cards have the edges lower than the middle while the AIB's are flush...
the heat from the solder should destroy the coating.
 

·
Registered
Joined
·
93 Posts
So did you put solder on the shunts directly without removing the conformal coating? I wonder if that would affect the resistance
So did you put solder on the shunts directly without removing the conformal coating? I wonder if that would affect the resistance
well in my case, I did the mod with silver paint first, so I scratched off the coating before soldering. But if you solder like debauer did, the heat from the solder should destroy the conformal coating.
 

·
Registered
Joined
·
249 Posts
The power slider still works. Its just the card thinks its using half the power since the resistance of the shunts are changed.
So 50% slider would be what use to be 100% with a 5 mohm stacked.
i just tried your tool for resistance calculator, gg really a nice work!

someone knows if in FE cards you shunt all with 005 except 008 on the pcie, will cap the 12pin power draw also?
 

·
Facepalm
Joined
·
8,623 Posts
i just tried your tool for resistance calculator, gg really a nice work!

someone knows if in FE cards you shunt all with 005 except 008 on the pcie, will cap the 12pin power draw also?
I believe that PCIE starts signaling "PWR limit" when it reaches 68W, which will cause a small drop in clocks and gpu voltage, but it can still go a bit higher before it starts throttling hard.
The 12 pin is capped by the two 8 pin power limits, which I believe are 174W for each pin. GPU Core is also capped at 300W. Any of these limits will trigger a power limit flag.

There's absolutely nothing wrong with shunting PCIE with a slightly higher resistance shunt. That's a hell of a lot better than having any of the 8 pins limiting you.
 
  • Rep+
Reactions: mattxx88

·
Registered
Joined
·
249 Posts
I believe that PCIE starts signaling "PWR limit" when it reaches 68W, which will cause a small drop in clocks and gpu voltage, but it can still go a bit higher before it starts throttling hard.
The 12 pin is capped by the two 8 pin power limits, which I believe are 174W for each pin. GPU Core is also capped at 300W. Any of these limits will trigger a power limit flag.

There's absolutely nothing wrong with shunting PCIE with a slightly higher resistance shunt. That's a hell of a lot better than having any of the 8 pins limiting you.
i'll start with your paint method then (finally i bought the pen MG chemicals for 50$ shipped to italy 😭), after that i'll go stacking resist and we'll see the difference 👌
 

·
Facepalm
Joined
·
8,623 Posts
Here's a 100% fullproof way of making the paint method work on a Founders Edition. It's more tricky doing this on a FE because the edges of the shunts are lower than the middle! This makes getting proper paint contact tricky. FE cards are not good for trying to use paint to stack shunts normally due to the edges being lower than the middle, which means the paint itself acts like a resistor, but we can use a shunt to "aid" in making sure the paint has good, cured contact with the edges (this is the most common failure with trying to paint a FE card!)

Grab some 842AR MG Chemicals conductive silver paint. You can use the jar (lasts longer) or the pen (this MAY be easier to apply without needing to do method #3 or #4, however people have reported that the pen dispenser can dry out!). You will also need a wooden sharp toothpick(s), tiny flat head screwdriver, as well as Torx T5 (backplate), T6 (GPU Xbracket), T8 (displayport bracket) and philips head for 4 backplate screws. I believe the philips head screws are m2*5mm or M2*6mm.
They can be replaced with pan head m2 screws but panhead will restrict that magnetic X thing from going on. Flathead screws are prone to stripping.

1) scrape the edges of the original shunt to remove the conformal coating fully. You want the edges to be completely silver and shiny. Do this for all six shunts.
This step is URGENT and must be done properly!

2) clean fully with alcohol. Then remember to FULLY shake the paint or pen before each application to mix it.
3) apply two dabs of 842AR to the silver edges of the shunts, making sure they are covered AND the 'dab' is higher than the middle of the shunt slightly.
The dab on the edges should cover the silver edges and MUST be slightly higher than the black middle of the shunt for the toothpick+temporary shunt method!

Note: if you are using the pen dispenser rather than the jar+toothpick method, you may be able to use the pen to paint over the shunt completely over the silver edges, then bridge it through the middle fully, skipping #3 and #4. Just make sure you don't apply too thin of a coat. Notice Sky3900 says "nice thick layer of paint." He used the pen, I used the jar+toothpick. RTX 3090 Founders Edition working shunt mod

He also followed this by baking the card it in the oven at 170 degrees FAHRENHEIT for 30 minutes to have the paint cure (this is 65 degrees CELSIUS and recommended by MG Chemicals for 30 minute curing). Air curing has to be done for about 12 hours. I have NOT tested this pen method. @Sky3900 used the pen method. Since the paint starts to thicken quickly and dry, the toothpick method is "probably" best done with the two dabs + temporary shunt. Remember--do not use too thin of a layer!

4) Now, wait about 1 minute after you applied the two dabs. This is so the paint will thicken slightly (this is very helpful) and will be less prone to running when you do the following. Now, lightly apply and stack a shunt on top of the dabs and push down very lightly. When you push very lightly, watch how the dab starts to spread and slowly push until the new shunt seems to be secure. Do not allow the dab to squeeze out so much that it gets on the PCB, but make sure you press down enough so the shunt barely makes contact with the middle black part of the original shunt. You can use any tool to press down very lightly!. If you did this right, the temporary shunt will act as a restrainer so the edge paint can cure.

Yes, this method requires that you have extra 2512 sized shunts to do this. But this is not to use the stacked shunt--only to help the paint "cure" to the edges of the original shunt. Note: you can use ANY 2512 sized shunt you want, 1 mOhm or 100 mOhm. DO NOT USE THE VIDEO CARD !!!

The purpose of this is to use the shunt as leverage to help the "paint" cure and fully contact the edges you scraped, and touch the conductive edges of the original shunt resistor.
I'm sure there is a term for this in the painting world but I'm not sure what it's called.

5) Let it cure about 12 hours like this (best to not use the card. If you did not use 10 or 15 mOhm shunts for this, do NOT even think about using the card!! This will save you a LOT of work!).

6) Now, pry off the shunt you added (it comes off easily) by prying it off from the middle section with your fingers and it should just detach cleanly, You should now have cured paint on the edges with a black exposed shunt in the middle. Now you have cured edge paint and now you need to complete the circuit.

The paint dab should still fully or mostly be on the shunt on the motherboard, with only a small residue on the shunt you pried off. If there is a tiny amount of silver showing, then you can bridge that in #7.

7) Now fully paint the shunt, topping off and bridging the edges of the old paint, applying a layer on top of it and painting over the entire shunt fully. It's better to apply a little too much paint than too little. Since the original paint is on the depressed edges of the onboard shunt, it should now be much easier to just top off and paint the entire shunt now. Be slow and careful and make sure none gets on the PCB. Just paint over the original dabs, then connect them through the middle in a nice layer.

8) let it cure, but you can use the card after 1 hour (don't forget to re-paste and re-pad, as well as padding over hotspots as needed).

This will always work but is time consuming.

If the original shunts were fully flat and flush, so much of this work could be avoided, and you could even apply a very small amount of mg chemicals for adhesion, to stack 10 mOhm shunts, as then the shunts would electrically touch each other (don't forget to scrape the conformal coating), and then you can secure them after it cures, with liquid electrical tape, but this method is too unreliable with edges of shunts being lower than the middle (in which case, soldering works best, or the dab +cure method I mentioned).
 
  • Rep+
Reactions: mattxx88

·
Registered
Joined
·
58 Posts
Can I stack a very low resistance shunt for a super high power limit, but it should be safe because I'm still limited by voltage?
 

·
Registered
Joined
·
718 Posts
putting 5mOhm on the pcie shunt risks damaging the card? is that the reason we put 8mOhm or higher on it?
 

·
Registered
Joined
·
93 Posts
putting 5mOhm on the pcie shunt risks damaging the card? is that the reason we put 8mOhm or higher on it?
it really depends on the bios. I am running the strix bios on my reference card, with 10mohm shunt on the pcie resistor. Card only pulls about 43w max (64.5W real) from the pcie slot. Now I kind of wish I shunted with 5mohm, but at least I know I will not blow up my PCIE slot. My only concern is that I am pulling close to 300 watts from one of my 8 pin plugs... don't know if this is safe.

2465960
 

·
Facepalm
Joined
·
8,623 Posts
Can I stack a very low resistance shunt for a super high power limit, but it should be safe because I'm still limited by voltage?
You can, but I wouldn't do that unless you are on water cooling or better. You also have to know if there are any 10 amp fuses on the card or not.
And it's not voltage that limits you. It's how much power the card will pull.
Playing at 4k or using Supersampling AA draws more power than playing at 1080p. Other features being enabled may also add to the power.
It's possible for a card to potentially try to pull up to 700W just by maxing the resolution and detail settings.

putting 5mOhm on the pcie shunt risks damaging the card? is that the reason we put 8mOhm or higher on it?
Yes, that's true, but PCIE will not really damage the card directly. It's the PCIE slot that can heat up, but I think someone here @olrdtg maybe tested 110 mhz PCIE and saw only a small temp increase, but he has direct airflow over the slot. Plus some cards have low 10 amp fuses on PCIE and 10A * 12v=120W...
 

·
Facepalm
Joined
·
8,623 Posts
it really depends on the bios. I am running the strix bios on my reference card, with 10mohm shunt on the pcie resistor. Card only pulls about 43w max (64.5W real) from the pcie slot. Now I kind of wish I shunted with 5mohm, but at least I know I will not blow up my PCIE slot. My only concern is that I am pulling close to 300 watts from one of my 8 pin plugs... don't know if this is safe.

View attachment 2465960
Where are you pulling 300W from an 8 pin? ...
 

·
Facepalm
Joined
·
8,623 Posts
201 - 220 of 382 Posts
Top