Shunt modding 3080 ti. Thoughts.

S

Samuraijack23

Weaksauce
Joined
Oct 23, 2022
Messages
119
So I decided to do it.

Those are 3090s but 308" ti are the same.

I will use liquid metal. Casious measures out. So I just apply LM on em/resistors/ and job done. How long it takes for LM to dry out?

Lots of people are buying resistors to avoid damage but still...3090
 
Where is the danger zone?
9829E021-F202-4BF0-A22F-FB843F7E2C8C.jpeg
 
Bad. So it can contact with areas I marked(what are these squares?) and 8pin tubes and short circuit? Or it takes time to erode em?
 
What is your goal with this? A quick benchmark run and then you remove the LM the same day, or a permanent modification? LM is not the way to go for a permanent shunt mod. Do anything else... stacked resistors, swapped resistors, 5cm of 20ga copper wire wrapped in kapton tape, hell even just blobbing solder over the shunts would be better (the last one probably won't work, but it's less likely to permanently damage the card than LM)

There is two issues here, short circuits due to LM dropping on other components during application, and LM just straight-up liquefying solder after it's applied. Again, if you paint LM over the shunt resistors you'll probably get a functional shunt mod for a short amount of time, and then after a time, you'll find the LM has fucked up your card. Could be days, weeks, months, depends on the exact LM and solder but the risk is extremely real.

Since the pic Furious_Styles posted didn't seem to do the trick, here's a more specific guide to acceptable places for LM to be applied to a GPU (Green= OK / Red=NO)

LM-no-no.jpg
 
I don't need it for long.

But there is a vid of derbeu of 1080 where he did 6month run and had no problems.
 
Enigma has the right answer. Liquid metal really has no place anywhere near any graphics card you intend to use for purposes other than competitive overclocking at any point.

There is basically no place on the card you can apply liquid metal without risk of it getting into somewhere it shouldn't be, pretty much regardless of how careful you think you are. A better solution, if you intend to use this card for normal gaming, desktop stuff, etc, is to just not mess with this, but if you must, do it the right way, with either a potentiometer or extra shunts in parallel.
 
Samuraijack23 said:
It is all bout gaming
Click to expand...
Then go for longevity instead of benchmarks. For gaming there is no reason to do a shunt mod on this card. The most I would do is look for a custom vBIOS to flash that works within the stock power limitations.
 
I wager within a month or so we will see a post "hey I modded my card and strangely it died, I cant warranty it and cant afford a replacement...so how to make it work again?"
 
RazorWind said:
with either a potentiometer or extra shunts in parallel.
Click to expand...
A potentiometer or POT for short does not have fine enough adjustment for a shunt mod or the current capacity for that matter... Stacked resistors provide a fixed resistance and the P-Channel MOSFET will give you a variable resistance setting when using a POT in conjunction with the MOSFET. Here is a video that describes the device in detail from der8auer:

When using the P-Channel MOSFET it basically gives you an adjustable resistance at a much finer adjustment with a higher amp rating that the POT can not directly give you.... (A POT will burnout because it can not carry the amps that the P-Channel MOSFET can)

A stacked resistor mod or a resistor replacement will only give you the ability for a fixed resistance reading aka a fixed power limit increase...
___________________________________

The cons that exist with an LM Shunt mod:

1. LM will eat solder and the traces on the board.

2. With LM you have no fine control over the resistance of the mod. (you just get lucky)

___________________________________

The pros that exist with a shunt mod in general:

1. When you are gaming and you are hitting the power limit of your card it will increase the power limit.

2. No Bios with a higher power limit is needed. (This is helpful if you are running a card and you cannot find a bios with a higher power limit that is compatible)

____________________________________

Points to consider with all shunt mods:

1. A BIOS might get you where you want to be with no mods.

2. Higher heat output (You better be running a killer custom loop, Phase Change, DI, or LN2)

____________________________________

If this is for a benchmark run use The Elmore PLM, If this is your daily driver use stacked resistors and/or the elmore PLM device, If this is for a card that is headed to the scrap heap use LM. Either way I salute your decision because at the end of the day it is your card and your money... I am appreciative that someone else on these boards even has the balls to discuss these kind of mods here because it has been a long time since [H] has been this [H]ard. If you need help or hints/tips feel free to respond.

Wishing you luck,

Enigma
 
Last edited:
Does shunt modding actually tangibly improve performance in day-to-day use? It's certainly necessary to get good benchmark runs under LN2 (let's not forget the Untouchables!) but my impression was that voltage scaling was pretty poor at ambient temps on modern nodes.
 
You might as well throw it into a microwave. There is absolutely no reason to shunt mod a GPU that you use as a daily driver. Hell I think all the world record chasers are dumb to bring with.
 
bwang said:
Does shunt modding actually tangibly improve performance in day-to-day use? It's certainly necessary to get good benchmark runs under LN2 (let's not forget the Untouchables!) but my impression was that voltage scaling was pretty poor at ambient temps on modern nodes.
Click to expand...
The way the Boost table works, it can. The shunt mod makes the voltage read lower than it actually is, effectively offsetting your position in the Boost table lower. You can get a higher Boost clock for the same operating conditions with no other changes, but if you don't control the resistance, it is easy to burn up your card without proper cooling.
 
Hey, this is [H] and all.. so I think the idea (of shunt modding) is cool. However, the performance gained vs risk and effort doesn't really make sense IMO. Actual 3090s are going for pretty cheap now, and they have twice the VRAM. If you're patient you can score a used 3090 Ti for ~$850 that will outperform the modded 3080Ti. Lots of other options... but if it's for the sake of [H] factor then I'm all in!

Or solder resistors in a stack on top of the existing resistors.....
Click to expand...

Exactly. If you're going to do this, do it right. If you don't have a good iron and the skill to to SMT work, then it probably doesn't make sense to do... Risk vs Reward.
 
Last edited:
Azzkiker said:
Hey, this is [H] and all.. so I think the idea is cool. However, the performance gained vs risk and effort doesn't really make sense IMO. Actual 3090s are going for pretty cheap now, and they have twice the VRAM. If you're patient you can score a used 3090 Ti for ~$850 that will outperform the modded 3080Ti. Lots of other options... but if it's for the sake of [H] factor then I'm all in!



Exactly. If you're going to do this, do it right. If you don't have a good iron and the skill to to SMT work, then it probably doesn't make sense to do... Risk vs Reward.
Click to expand...
Doing it properly is cool. Doing it with liquid metal is dumb.
 
There's no moisture (that is, water) in the thermal pads. What you're seeing is mostly like the oil the pads are impregnated with.

The difference matters because water will eventually become conductive and cause horrific corrosion, whereas the oil generally does not.
 
I hit the hard wall of 600w limit
 

Attachments

  • 132873DD-646E-4041-8F56-1A0A706F9E3C.jpeg
    132873DD-646E-4041-8F56-1A0A706F9E3C.jpeg
    1.1 MB · Views: 0
nice results, this is the stuff I like to see. Cool the silicon down and you will gain higher boost clocks. May I ask what method you finally settled on going with?
 
LM + isolator
Click to expand...

do you know what the final resistance you ended up with was by chance? from that we/you would be able to calculate out what your final/possible gains would be with that mod... would be interested to see.
 
You must log in or register to reply here.
Back
Top