Overclock.net banner

2080 Ti - Working Shunt Mod

101K views 301 replies 53 participants last post by  GhOsT662 
#1 · (Edited)
I finally got a chance to test some other shunts on my FE 2080 Ti.

When using 5 mOhm resistors, as I did on my Titan X (Pascal), the 2080 Ti goes into emergency mode with a lot of loads (stuck at 300 MHz). My first try adding a shunt above 5 mOhms, 8 mOhms, did not trigger the cards safety mode. I used Panasonic ERJ-M1WSF8M0U current sense resistors, which match the original resistors on the card.

This is not bad, 1/5 + 1/8 = 1/3.077, for a final power limit of 1.625 * 260 = 422.5 watts (520W at 123%). :D
Edit: My Kill A Watt's reported power draw tracks well with what hwinfo reports for power use (x1.625) so I believe my mod worked as expected and I can use hwinfo to get a fairly accurate value for the card's powder draw.




Sadly my card doesn't OC that well, 2115 MHz core and 8100 MHz memory, but no throttling is nice.
Superposition 1080p Extreme: 10310
Superposition 8K Optimized: 6033
Time Spy Graphics Score: 16395
Time Spy Extreme Graphics Score: 7768
Port Royal: 9215
 
See less See more
2
#2 · (Edited)
I finally got a chance to test some other shunts on my FE 2080 Ti.

When using 5 mOhm resistors, as I did on my Titan X (Pascal), the 2080 Ti goes into emergency mode with a lot of loads (stuck at 300 MHz). My first try adding a shunt above 5 mOhms, 8 mOhms, did not trigger the cards safety mode.

This is not bad, 1/5 + 1/8 = 1/3.077, for a final power limit of 1.625 * 260 = 422.5 watts (520W at 123%). :D




Sadly my card doesn't OC that well, 2070 MHz core and 2000 MHz memory, but no throttling is nice.
Time Spy Graphics Score: 16017


Nice work! My card is a PALIT 2080Ti DUAL with a Non-A GPU with 112% bios. What will a shunt mod give me in Powerlimit? And did you do it on a Referens design PCB? :)

Would a resistor like this be usable? I have alot over from my old days fiddeling with 780Ti Volt modding.

https://www.ebay.com/itm/1500pcs-1-...h=item4b499da5be:g:SLIAAOSw-xNbWDUf:rk:1:pf:0
 
#3 ·
Nice work! My card is a PALIT 2080Ti DUAL with a Non-A GPU with 112% bios. What will a shunt mod give me in Powerlimit? And did you do it on a Referens design PCB? :)
Thanks :)

Yes, this is the reference PCB, it is a Founder Edition GPU. If you also use 8 mOhm resistors you would get 280 * 1.625 = 455W

Would a resistor like this be usable? I have alot over from my old days fiddeling with 780Ti Volt modding.

https://www.ebay.com/itm/1500pcs-1-...h=item4b499da5be:g:SLIAAOSw-xNbWDUf:rk:1:pf:0
Those resistors would not work, they are 1 - 10 ohm and we need 0.008 ohm resistors. They would also be harder to solder on.
I used these from mouser: ERJ-M1WSF8M0U
 
#4 ·
Does this really benefit you if not using sub ambient cooling?




I don't think the card hitting power limit will lose that much fps in real time scenario either.
 
#7 ·
Hi,
So if your card fries how do you remove evidence of the solder you added to the original shunts.
 
#8 ·
#11 ·
Nice to hear, but you could just have used the modded NVflash to flash the 380W KFA2/Galax Firmware to the FE Card. Getting stable 2160MHz with it and only reaching the power limit extreamly rare.
Saadly my card still doesnt get close to the Performance of my Titan V [emoji26]
I finally got a chance to test some other shunts on my FE 2080 Ti.

When using 5 mOhm resistors, as I did on my Titan X (Pascal), the 2080 Ti goes into emergency mode with a lot of loads (stuck at 300 MHz). My first try adding a shunt above 5 mOhms, 8 mOhms, did not trigger the cards safety mode.

This is not bad, 1/5 + 1/8 = 1/3.077, for a final power limit of 1.625 * 260 = 422.5 watts (520W at 123%). :D




Sadly my card doesn't OC that well, 2070 MHz core and 2000 MHz memory, but no throttling is nice.
Time Spy Graphics Score: 16017
Gesendet von meinem H8166 mit Tapatalk
 
#12 ·
Nice to hear, but you could just have used the modded NVflash to flash the 380W KFA2/Galax Firmware to the FE Card.
I would probably have done that if the modded NVFlash was available when I first took a soldering iron to it. I do have the stock bios with an even higher power limit now though, and I can hit it. Though the difference in performance between 380W and 450W is probably negligible. I don't expect too many people to do this mod but it might be helpful for those using more extreme cooling. :)

I have seen it say 271W in hwinfo, which should be about 440W. With the system idle, it pulls 205W (GPU probably pulling 30W) and with madVR playing it pulls 620W. madVR uses very little CPU with these settings so I think all the change in power is the GPU, for ~445W total GPU power draw. With madVR at these settings I do see the power limit trigger sometimes when at 100% but not when at 123%.


 
#14 · (Edited)
I have a Palit 2080 Ti Dual and am looking into this option. I've soldered small components before so precisionwise I should be good but just to be safe I thought I'd ask about getting the right kind of soldering iron (one that lends itself well to this kind of work, the right size and heat) and any other equipment to get this done right. Also, I assume I will be needing 2x 8 milli-Ohm resistors for this so if there are certain ones that are better than others for this I'll go for them to make sure this works properly. Do I use 60/40 or leadfree or some other option for soldering it?

Oh, and could you just as a reference show exactly where those resistors are on the PCB?

Just out of curiosity, will the raised power limit reduce the lifetime of the card in any way?
 
#17 · (Edited)
I used a very cheap soldering iron with a small flat tip and the mystery solder it came with. It needs to get hot enough to melt the solder under the top resistor, which is pretty hot. It has a knob to adjust temps but no reference to know what it is set to, I had to turn it up almost all the way. I should have taken it to work, I probably would have done a much cleaner soldering job.

These are actually pretty big surface mount resistors so it isn't too small of a job. I did link the resistors I used in the original post, they match the shunts used on the card originally too (except being the 8 mOhm version).
Edit: oops, no I didn't. :eek:
I did a few posts down but they are Panasonic ERJ-M1WSF8M0U current sense resistors.
More details on how I did the soldering in my Titan X (Pascal) Shunt Mod thread.

The power limit will only reduce the life of the card if you also over-volt it and/or run it really hot. I edit the voltage curve to keep it at or below 1025mV and water cool it so I don't expect unlimited power to lower the life. However, we don't have any historical data, of course.

I have highlighted the resistors:

 
#19 ·
Overvolting the RTX cards requires hardware modification of some sort though, right? I was under the impression that for instance the slider in Afterburner didn't actually do anything due to these cards having some sort of limitation put in place.
 
#20 ·
Oh yeah, can I use nailpolish as a safetyprecaution for the stuff around the resistor in the first picture? It seems to have quite a few other bits of solder that probably shouldn't be touched around it and I don't want any to leak onto them, you know?
 
#23 ·
I will never try to RMA my cards after doing this mod. :)
 
#24 ·
I have no 2080Ti to compare, but on my Titan RTX cards I just removed the 5MO resistors and replaced them with 3MO.

No fancy aftermarket higher TDP BIOS for me, but with 112% power limit in Afterburner both cards are pulling nearly 430w each, and I'm only heat limited as far as I can tell (still in top 10 HoF for the new Port Royal ray tracing benchmark).
 
#25 ·
I have no 2080Ti to compare, but on my Titan RTX cards I just removed the 5MO resistors and replaced them with 3MO.
Very nice.

The result of my mod is very similar, adding the 8MO on top of the 5MO is equivalent to replacing it with a 3.077 mOhm resistor but I didn't need to de-solder the original. The difference with Turing is that it goes into emergency mode if you use a 2.5 mOhm resistor while Pascal was fine with it.
 
#28 ·
@Asmodian are you saying 7mo resistors would be better for the shunt than 8mo.
I tried 5mo,s last time and card failed, I don’t know if the shunts caused it or natural nv failure.
 
#29 ·
I am saying I believe they would work, and they would allow more power, but I did not test them to be sure they do not cause the same failure as 5MOs. :(
 
#30 · (Edited)
Thanks for the info and photos in the opening post. Great job, bro.

I had done this mod to my 1080 Ti and it worked great, and now I am planning to do the same with my 2080 Ti. I was able to run the 1080 Ti at 2240 on core with 1.200V and no dynamic clocking nonsense, so I am hoping to be able to do the same or better with the 2080 Ti. Without the mod, my max core clock is 2205 with frequent dropping to 2175 in spite of the fact that I keep it below 30°C when benching it.

Question: Why no shunt mod on RS16? I am wondering if not doing that is what is triggering your low clocks with the 5 ohm resistors and possibly limiting your max core overclock. I did the shunt mod in all three power circuits on the 1080 Ti.
 
#31 ·
Question: Why no shunt mod on RS16? I am wondering if not doing that is what is triggering your low clocks with the 5 ohm resistors and possibly limiting your max core overclock. I did the shunt mod in all three power circuits on the 1080 Ti.
That is for the power from the PCIe slot and I didn't want it pulling more from the motherboard. The power balancing circuity only considers the two 8 pin power connections and the card is designed to get most of its power from them anyway. Missing a resistor definitely didn't cause safety mode, the easiest way to fix that was to desolder the resistor I added to the back of the card, with only one it never went into safety mode. Adding more shunts is more likely to cause it to go into emergency mode (except I think it only monitors the two 8 pins?). I never see the power limit flag trigger when running any benchmark or game, only voltage or load limits. My GPU simply doesn't clock very well, even with light loads. Heaven 1080p will crash sometimes at 2130 MHz at any voltage. The lottery is what it is. :(

I usually run a very stable 2070MHz at 1.025V and I usually see loads that don't go much above 90% of the new power limit (422W). I leave it on 123% anyway. :)

With a better GPU a higher power limit would probably be more important, the watts used goes up quite a bit at 2200MHz compared to 2100.
 
#32 · (Edited)
Just to confirm, the second image shows the resistor to be replaced that's on the... well, back of the PCB? The side covered by a backplate, that is? Also, I have a equipment on the way now and the soldering station can run between 150 to 450 c (300 to 850 f) and I just thought maybe there was a sweet spot for this kind of job? Is hotter preferable so it's done fast?

Also, removing the original solder, is a wick best for that since it's relatively small?
 
#33 ·
I did not replace any resistors, I added those 8 mOhm resistors on top of the original 5 mOhm ones. There is one on the front and one on the back. If you want to replace them you need to use 3 mOhm shunts, using 8 mOhm ones would lower the power limit a lot. There was plenty of room for the double stacked shunts under my water block and backplate (both from EK).

Check out my Titan X (Pascal) mod (link in the first post) to see a photo of the stacked shunts from the side, it worked exactly the same way on my 2080 Ti (except both shunts were on the front of the Titan X).

I think 350°C is about where I found it to work well, too hot is bad because we don't want any other components falling off but at the same time it needed to be pretty hot to melt the solder between the two shunts. I tinned the new resistor, held it on top of the original one with tweezers, and heated it from the top to melt the solder between the two. This created a good connection between them without using extra solder which might get where it shouldn't. Surface mount can be tricky with very small components near by while soldering larger things. I found that I needed to turn up the temperature until the solder melted, just a little bit cooler and it wouldn't flow no matter how long I heated it, while as soon as I got the soldering iron hot enough it would flow within a few seconds and the entire process was pretty easy and safe feeling.
 
#34 ·
Ok, very good thing I asked about this. So I just need the solder to connect the two shunts then. I suppose I can adhere the one end with just a tiny strip of tape or maybe apply pressure with a precisiontool of some sort... or both... and then apply the heat on top of the other side of the 8 mOhm shunt. Then poke it until it flows and give it a enough to connect them. Once one side is good, the other is a breeze. I have 1 mm leadfree tin that I hope will be good enough. There was also 0.8 mm that I probably should have gotten but whatever...

Also, I'll apply some nail polish on the stuff I wanna protect and this should honestly be a relatively painfree process, provided I can still my heartbeat and work past my sweaty palms...
 
#35 ·
Sounds good but I would keep anything the melts/burns away (tape or similar). The 1 mm solder will be great, tinning the new shunt is really easy and you don't need to add solder on the card. A precision tool of some sort to hold the new resistor would be great, my hands are pretty steady but soldering that first connection while holding the top resistor in place with tweezers was definitely the tricky part, you can see that I didn't get the front resistor perfectly aligned. Getting the soldering iron temperature high enough was key so I didn't need to hold steady for too long.

Personally I wouldn't use nail polish on anything like that, you really shouldn't be spraying solder anywhere and nail polish melts/burns at solder temps so it isn't great protection. It probably just creates a false sense of security and a mess. With liquid metal nail polish makes a lot more sense.
 
#36 ·
Ok, got it. No nail polish and I'll see about how to keep the shunt in place more securely. Would you advise getting isopropylalcohol applied first to clean everything up both on the original and the new shunt?
 
#40 ·
Yes, isopropyl alcohol is always a good idea. Some solder flux isn't a terrible idea either, but not required.
 
#38 · (Edited)
I finally got a chance to test some other shunts on my FE 2080 Ti.
Thanks again for creating this thread and posting your mod. It was very helpful.

OK, my shunt mod is also complete now, and it has worked with a major caveat: voltage is still the same... TOO LOW. It is not drawing any more watts because it was already getting all of the watts it needed access to based upon the stupid low voltage limit. So, I cannot get a higher overclock with the same voltage. It was a success in that it is functioning correctly by grossly under-reporting power draw and there are no more voltage and power limit reasons showing in HWiNFO64 and GPU-Z. There is a little less fluctuation in the power utilization so it might return slightly higher benchmark scores, but it won't be earth-shattering improvement. Testing that theory is pending. So, I am going to have to do something to get the voltage higher to take this GPU where I want it to go. So far, I have not found any vBIOS to allow more than 1.093V max (all are identical), so we either need a hardware mod or a special vBIOS like the Strix XOC 1080 Ti vBIOS that allows adjustment up to 1.200V. It doesn't need to be signed firmware as long as it works. I'll use a "sledge-hammer" approach to flashing if I need to. (Already have been tinkering with that, but no vBIOS with voltage unlocked yet.




Here is a side-by-side video showing power draw before and after the mod with the same vBIOS, CPU BIOS settings and same GPU overclock profile. HWiNFO/OSD was reporting about 380W maximum and 120% of TDP during 3DMark 11 Test #1 and now it reports around 250W and 65% of TDP.



Here are some before and after photos... As you can see, the first 0.005 ohm resistors I tried also didn't work for my GPU, so I had to decrease the resistance using 0.008 ohm resistors.










 
  • Rep+
Reactions: Asmodian
#39 · (Edited)
OK, my shunt mod is also complete now
Nice job! Great photos too.

It is very interesting that even an 8MO resistor on top of RS16 caused it to go into safe mode, I would not have expected that. Nvidia must have tighter restrictions on the power draw from the PCIe bus.
Edit: Or maybe even with 8mOhm shunts added the total measured powder draw is really close to being below the lower limit, so adding the third shunt pushes it below the limit?
 
#43 ·
Very true... But simultaneously, I'm grateful there are people trying to figure out how to hardware mod past the 1.093 V limit since I'm stuck with my Non-A card so flashing the vBIOS is impossible for me. :c
 
#45 ·
Hey Rob w. I use liquid electrical tape to insulate everything when I do shunt mod. It covers all the solder up and anything else that might be an issue. I just used LM as its very easy to apply and remove. The liquid electrical tape takes the risk away. And it is very easy to remove when dry. Peels right off and looks brand new underneath. I never use back plates so no experience with the tight fit there.
 
#47 ·
Speaking of liquid metal, just wanted to ask since I'll be detaching the cooler for the modding, can I apply liquid metal too the GPU die? That is, is the cooler surface nickel or copper? Figured I may as well get as low temps as I can to push the card as much as possible.
 
#48 ·
I use LM on both my gpu’s (Titan v and 2080ti) I’ve found it better than paste, I just make sure surrounding components are protected, (never had a problem with it)
For shunt mods ? I tried LM on the Titan but had for better results when I replaced it with 5mo resistors.
 
#56 ·
Shunts are circled in red for the 8pin power connectors are the ones to mod,
The shunt circled in green is for the pcie, not normal to mod this one.

 
#58 ·
I finally got a chance to test some other shunts on my FE 2080 Ti.

When using 5 mOhm resistors, as I did on my Titan X (Pascal), the 2080 Ti goes into emergency mode with a lot of loads (stuck at 300 MHz). My first try adding a shunt above 5 mOhms, 8 mOhms, did not trigger the cards safety mode. I used Panasonic ERJ-M1WSF8M0U current sense resistors, which match the original resistors on the card.

This is not bad, 1/5 + 1/8 = 1/3.077, for a final power limit of 1.625 * 260 = 422.5 watts (520W at 123%). :D
Edit: My Kill A Watt's reported power draw tracks well with what hwinfo reports for power use (x1.625) so I believe my mod worked as expected and I can use hwinfo to get a fairly accurate value for the card's powder draw.
Thanks for the tips. Other modders have also confirmed that a total resistance of 3 mOhm across the shunts is about the ideal spot.

My 2 2080Ti XC Gaming Blacks will be en-route to me this Friday from eVGA step-up. I ordered the SMD resistors you linked, my NVLink bridge ($50 NIB on eBay!), and a slew of other components to upgrade my rig.

Now, you only had to do this mod to the 2 shunts pictured? There is a third for the PCIe power correct? I know on the Titan V you had to mod that one as well since it ended up having a connection back to the +12V on the 8 pin power connector.
 
#59 ·
Asmodian :specool:
Renegade I have the same resistors on my desk, I will be using the 8mohms on top of the 5mhom.
Afaik you should not do the pcie shunt, I did it on my titanv and ended up blowing a power reg on back ( luckily repaired ) so I would not recommend it.
Been trying to flash my fe a chip but it won’t flash:rolleyes: so will shunt it instead:D
I actually feel happy shunting this one:h34r-smi;)
 
#60 ·
Oh nice, sucks your card doesn't OC that well. My waterforce can hit like 2200+ a few mhz, but I don't know what kind of changes gigabyte does to it.

Out of curiosity, do you know how shunt mods like this affect the lifespan of the card? I'm curious about doing these but I want a GPU to last me at least 3 years usually since I use my older ones for algorithm testing.
 
#63 ·
Out of curiosity, do you know how shunt mods like this affect the lifespan of the card? I'm curious about doing these but I want a GPU to last me at least 3 years usually since I use my older ones for algorithm testing.
As long as temps are kept under control I don't think it hurts longevity at all. Extra voltage is what hurts longevity the most; don't go above 1.05V or so if you want to maximize longevity.
 
Top