Overclock.net banner

301 - 319 of 319 Posts

·
Facepalm
Joined
·
9,883 Posts
As someone who's attempting to shunt mod their 3090 FE, this thread has been a very helpful read. I gave it a shot this weekend with the 842AR Pen and was a pretty clear failure, no change in power. At that point I had only read through the first few pages so I realized I missed quite a lot, namely: the pen sucks, it takes multiple coats, and the FE has weird shunts that are annoying in particular to paint because of the different heights.

Your replies in particular have been helpful @Falkentyne, so thank you for that!

I'm trying to figure out what to change for my next--hopefully more successful--attempt.

I'm seeing two ways forward:
  • Potentially rescrape to make sure all conformal coating is off (feel confident that i did already, tbh)
  • Get the 55ml container of 842AR and paint several additional layers onto all 6 shunts with a toothpick being careful to do it evenly
  • Use super 33+ tape to feel comfortable doing a thorough job
Alternatively:
  • Ditch the paint method entirely and go for a soldered stack of 5 mOhm shunts
Based off what @Falkentyne's experience it seems like 842AR paint isn't as effective, especially long term. I'm a little confused about this because I'm not understanding why the paint would degrade over time, if done properly.

I'm considering soldering instead but would really prefer to go with the paint since it's all-round safe/more convenient. Only issue soldering is that is that I've never done it (although I'm familiar with the process) and would need to buy a soldering iron like the TV100, flux, solder, 5 mOhm shunts, etc. Although I've got reasonably steady hands and an old GTX 960 I can practice the solder bridge on, I guess I wanna avoid spending the $ on a one-off GPU mod. If soldering is really the only way to do this right, I'm willing to learn and do it--but I guess I'm looking to get nudged in either direction.
Elmor himself told me that he would never use paint for anything except temporary work and testing. Paint like this is useful for restoring contact to a severed trace, however, like someone who had a scratched PCB by the PCIE slot, over some traces, causing the card to run in x8 mode. When he painted the traces carefully (insulting around it with Super 33+ tape) with the 842AR, his card ran in x16 again, so that was a job well done.

On the shunts, the MG 842AR works best if you scrape the shunt's edges down to where the copper starts to show, and then apply it. Then you need to apply several layers thick and 'work it in' to the edges while connecting through the middle, which requires PERFECT protection around the shunt with Super 33+ tape (or 3M Kapton tape). But even then that isn't going to last long term.
I honestly don't know why but I suspect it's the edges of the shunts being displaced from the black housing, preventing the paint from creating a very solid adhesion. One thing I noticed, when removing the paint, the paint seems to create a PERFECT bond to the black housing in the middle and to the parts of the edges connecting the middle. But when you scrape the outer layers off and start working at the lower parts directly, the part of the paint right on the lower silver/copper parts literally just comes off in a chunk. I'm just a gamer so I don't know enough about any of this to know why.

Anyway you asked for how to solder? It's rather easy.
First the easiest way to practice PERFECTLY for 3090 FE depressed shunts is just to get either an old video card PCB with shunts, or--even better-- a card that has exposed solder bumps (unoccupied) on it (or otherwise--create your own! :) and then solder a 2W (not 1W) shunt to the board. 2W shunts have the depressed edges. Then, take a 1W shunt (1W shunts have flush edges, flush all around) and then practice 'building a solder bridge' and then attaching the 1W shunt to the 2W shunt. Fluxing every single step (before each solder passover). That's basically how you solder the actual FE card with stacking.

Here are some links.

5 mOhm 1W shunts.

5 mOhm 2W shunts (if you want to practice solder stacking a 1W shunt on top of a 2W shunt by "building a solder bridge", fluxing each step.

3 mOhm 1W shunts (for desoldering the originals (VERY HARD TO DO) and replacing them. Note: out of stock.

Ok so....Here is what you need. This is all you need for doing a good soldering job.
1) essential: 3M high temp polyimide tape.
Or

The "92" one has a higher temp rating.
use this tape and you won't have to worry about having the soldering iron touch something next to the shunt you're working on--the kapton tape will protect it.

2) essential: Rosin flux. Flux is magic.

3) useful but non essential: desoldering wick. in case you got solder on something and need it removed, or maybe there's extra solder on top of the shunt, shorting the backplate and you need to siphon it off. Chemtronics 80-1-10 Braid, DESOLDERING, Rosin SD, 10FT: Amazon.com: Industrial & Scientific

4) TS-100 temp regulated soldering iron. 65W. The bare minimum you need for a job like this as the PCB absorbs so much heat from the iron when you're trying to heat the shunt for the solder to flow to it. Please avoid ANY crappy beginner starter soldering iron!

This is about the best you can do without getting into the hundreds of dollars.

5) applicators and wipers.
These are great. Not only useful for cleaning flux after soldering, with alcohol, but these are also splendid for liquid metal applications--better than the Grizzly Conductonaut applicators. Lint free.

6) 60/40 Kester solder. Highly recommended, even though the TS-100 kit comes with solder, this is better.

7) solder tip cleaner, use this, along with the kit's included sponge (make sure the sponge is wet) for cleaning your iron tip (you do this while the iron is hot, of course).

And that's basically it.
You can follow this video for 'building a solder bridge', even though he has flush shunts, you need this method on 2W shunts like the FE (you'll be stacking 1W shunts on top of them. note: don't stack with 2W shunts--no need for more headache :).


16 min and on in oldirdey's video shows excellent and clean shunt stacking, but these are flat 1W shunts. So this method won't work with the FE (you need to build the 'solder bridge').

When you're finished make sure the shunts are fully flat and not un-level, you want them all the way down so they clear the backplate!! You also want to make sure there are no solder points on top of the shunts on the backplate side (e.g. GPU Chip Power, PCIE Slot, PWR_SRC), as there's not much clearance. Make sure there are absolutely no 'solder sharp points" above the stacked shunt. If there are, you need to get it off with the wick, or use flux on top of the shunt and have the cleaned iron 'wipe' the solder off, from the shunt to the iron.

For extra security, put a strip of your 3M Kapton tape right above where the shunt sits, on the backplate, just in case it ends up touching (that obviously isn't necessary on the GPU Core side).
 
  • Rep+
Reactions: tefla

·
Registered
Joined
·
838 Posts
Based on my own journey, and my MG 842AR cleaning mishaps, as well as paint degradation, I'd say had I went the soldering method from the get go, I wouldn't have had so many heart attacks. 😅

The soldering method, personally, is so much cleaner and generally safer if you choose to protect the surrounding areas with kapton tape, as well as proper flux. No issues with nigh-invisible particle residue when cleaning up.

If you're having difficulty getting your shunts flush, an interesting alternative method I've tried and tested is to get your shunts more flush with one another is to only build a solder "ridge" on one side of the stock resistor and stack the new one on top. The other side will be completely flush as there is nothing under. Then you simply add solder to both edges after this initial joint (it should fill the gap enough on the opposite end too to make a solid connection).

Otherwise, soldering tweezers like this ensure both sides have the solder melted simultaneously (though I have never used one of these before).

 
  • Rep+
Reactions: tefla

·
Facepalm
Joined
·
9,883 Posts
Based on my own journey, and my MG 842AR cleaning mishaps, as well as paint degradation, I'd say had I went the soldering method from the get go, I wouldn't have had so many heart attacks. 😅

The soldering method, personally, is so much cleaner and generally safer if you choose to protect the surrounding areas with kapton tape, as well as proper flux. No issues with nigh-invisible particle residue when cleaning up.

If you're having difficulty getting your shunts flush, an interesting alternative method I've tried and tested is to get your shunts more flush with one another is to only build a solder "ridge" on one side of the stock resistor and stack the new one on top. The other side will be completely flush as there is nothing under. Then you simply add solder to both edges after this initial joint (it should fill the gap enough on the opposite end too to make a solid connection).

Otherwise, soldering tweezers like this ensure both sides have the solder melted simultaneously (though I have never used one of these before).

Unfortunately, this won't work on the FE shunts, or on the Gigabyte shunts. Since the edges are lower than the middle (2W shunts are like this), you have to build a solder bridge on both sides. If you don't, the second side won't have proper contact at all. Even if you try to apply solder around the side you didn't build a bridge on, you would have a giant gap to cover, with the PCB absorbing all the heat from the solder and it not sticking to anything except your iron or the top of the top shunt :(
 

·
Registered
Joined
·
16 Posts
Woah, thanks for the in-depth help @Falkentyne, you as well @ArcticZero. Guess I'll be practicing some shunt stacking on an old gtx 960 once i order everything.😄

Do either of you have any tips for getting the MG 842AR paint off in the best way possible? Seems like it's mostly scraping then using a q-tip dipped in IPA. Then using some compressed air on the PCB to get any potential left-overs off. I'm just not sure how thorough I need to be before it's safe to solder on. I only have one crappy layer of paint so hopefully it doesn't take too long.

Just as an aside, I think it would be awesome if you eventually got around to making a new thread @Falkentyne. You've shared a ton of amazing info/learnings spread out across this and another thread (about shunt modding in general, paint vs solder, and the 90FE in particular). It would be a great resource to have that all collected in one place!
 

·
Facepalm
Joined
·
9,883 Posts
Woah, thanks for the in-depth help @Falkentyne, you as well @ArcticZero. Guess I'll be practicing some shunt stacking on an old gtx 960 once i order everything.😄

Do either of you have any tips for getting the MG 842AR paint off in the best way possible? Seems like it's mostly scraping then using a q-tip dipped in IPA. Then using some compressed air on the PCB to get any potential left-overs off. I'm just not sure how thorough I need to be before it's safe to solder on. I only have one crappy layer of paint so hopefully it doesn't take too long.

Just as an aside, I think it would be awesome if you eventually got around to making a new thread @Falkentyne. You've shared a ton of amazing info/learnings spread out across this and another thread (about shunt modding in general, paint vs solder, and the 90FE in particular). It would be a great resource to have that all collected in one place!
Forgot to say, use 385C on the iron for good results making the solder bridge. It takes several seconds of putting the iron tip (tinned with solder of course) and applying back and forth motions to get the silver shunt edge hot enough for the solder to flow and then stick to the edge and create your bridge.

Please buy some super33+ electrical tape first. You want to use a small mini flat blade screwdriver very carefully to scrape. 3M high temp polyimide tape (the same stuff you will use for soldering protection) works also but is more expensive. Use the tape completely around the shunt at least an inch around because flakes will get all over the place, and the tape will help protect the PCB and keep the paint in a controlled area. Wipe/blow the area of the shunt off frequently. I would NOT use alcohol until the very end, because the liquid mixture with paint will go everywhere it can, into every little tiny crack and around every tiny SMD component, making cleanup extremely difficult, and even those lip applicators I linked last post won't be able to fit in some areas. Just scrape and scrape with LIGHT pressure and take your time. Go very slow, and you will be thanking that super 33+ tape for protecting the PCB and components around the edges of the shunts. This is a very time consuming process.

Of course, it goes faster with 99% alcohol, much faster, but then the cleanup afterwards will take up time. The issue is that the liquid mixture gets into the adhesive underside and loosens the tape grip, allowing more liquid to get down there too.
 
  • Rep+
Reactions: ArcticZero

·
Facepalm
Joined
·
9,883 Posts
Based on my own journey, and my MG 842AR cleaning mishaps, as well as paint degradation, I'd say had I went the soldering method from the get go, I wouldn't have had so many heart attacks. 😅

The soldering method, personally, is so much cleaner and generally safer if you choose to protect the surrounding areas with kapton tape, as well as proper flux. No issues with nigh-invisible particle residue when cleaning up.

If you're having difficulty getting your shunts flush, an interesting alternative method I've tried and tested is to get your shunts more flush with one another is to only build a solder "ridge" on one side of the stock resistor and stack the new one on top. The other side will be completely flush as there is nothing under. Then you simply add solder to both edges after this initial joint (it should fill the gap enough on the opposite end too to make a solid connection).

Otherwise, soldering tweezers like this ensure both sides have the solder melted simultaneously (though I have never used one of these before).

BTW you don't need a solder bridge at all on the flat / flush shunts. Just flux each step, that's all, and use oldirdey's method (starting at 16 min) in his video (the second video).
But on the FE shunts (Gigabyte too, or any 2W shunts with lower silver edges than the black middle), you must build a bridge---see first video--just with the depressed edges it's harder to get the solder to flow---you have to heat the silver part for several seconds to get the shunt hot enough for the solder to stick (remember flux BEFORE the bridge, and flux AFTER the bridge also and then flux again when putting the new shunt on top (on the sides of the new to old shunt, not the top of the new shunt). Flux every step.
 
  • Rep+
Reactions: ArcticZero

·
Registered
Joined
·
838 Posts
Do either of you have any tips for getting the MG 842AR paint off in the best way possible? Seems like it's mostly scraping then using a q-tip dipped in IPA. Then using some compressed air on the PCB to get any potential left-overs off. I'm just not sure how thorough I need to be before it's safe to solder on. I only have one crappy layer of paint so hopefully it doesn't take too long.
Consider my case a prime example of what Falkentyne is saying. I cleaned off the MG 842AR with alcohol dipped Q-tips before soldering the shunts on. My GPU would randomly fail to be detected on the BIOS, and even when it did would crash randomly. I was blaming my soldering job and was considering undoing the shunt mod completely until I waa pointed in the right direction (paint flakes)

Don't use alcohol for wiping the paint off your shunts. Paint flakes will mix in with alcohol and spill all over your PCB, which obviously can cause shorts you don't even see. No matter how much manual clean up you do, this is bound to happen if you go down that path. The only thing that solved it for me was a complete 99% isopropyl alcohol bath, which despite my prior investigation, yielded several shiny paint flakes floating in the liquid after.
 

·
Registered
Joined
·
16 Posts
Well, I did something! Took 7 hours to practice, scrap paint off, solder, cut new thermal pads, and assemble but i didn’t break anything so that’s great. Some were better than others but none were terrible to my untrained eye. @Falkentyne would love to get your take on the results. What testing protocol do you usually do? I saw you prefer heaven bench and gpu-z max power readings. Any other specific settings?
2487011

2487013

2487016
 

·
Facepalm
Joined
·
9,883 Posts
Well, I did something! Took 7 hours to practice, scrap paint off, solder, cut new thermal pads, and assemble but i didn’t break anything so that’s great. Some were better than others but none were terrible to my untrained eye. @Falkentyne would love to get your take on the results. What testing protocol do you usually do? I saw you prefer heaven bench and gpu-z max power readings. Any other specific settings? View attachment 2487011
View attachment 2487013
View attachment 2487016
Please be VERY CAREFUL with those solder bumps on top of the shunts!!!!
If you are using the original backplate, bumps that large WILL 100% touch the backplate! There is already not a lot of space between the original shunt and backplate. If the solder bump sits even 1/5th of a mm higher than the MLCC caps right next to it, it will touch the backplate and short it. This is not a problem on the GPU Core side but is on the back side.

You can do a test by putting a trimmed strip of Kapton tape right behind the shunt location on the backplate, put a very small amount of flux (very small, spread out thin) on top of the shunt and then screw the backplate (all torx and screws) on completely then unscrew and remove, then inspect the kapton tape. If you see any flux on the tape, or any indention on the tape, it means the shunt is touching it.

I highly recommend removing all the excess solder from the top of the shunts on the backplate side and having it as flat as possible there. And when finished, leave a strip of kapton tape above each shunt's location to be safe.

Shorts will cause very severe stuttering/hitches or even the card to shut off or possible damage.
 

·
Registered
Joined
·
16 Posts
Please be VERY CAREFUL with those solder bumps on top of the shunts!!!!
If you are using the original backplate, bumps that large WILL 100% touch the backplate! There is already not a lot of space between the original shunt and backplate. If the solder bump sits even 1/5th of a mm higher than the MLCC caps right next to it, it will touch the backplate and short it. This is not a problem on the GPU Core side but is on the back side.

You can do a test by putting a trimmed strip of Kapton tape right behind the shunt location on the backplate, put a very small amount of flux (very small, spread out thin) on top of the shunt and then screw the backplate (all torx and screws) on completely then unscrew and remove, then inspect the kapton tape. If you see any flux on the tape, or any indention on the tape, it means the shunt is touching it.

I highly recommend removing all the excess solder from the top of the shunts on the backplate side and having it as flat as possible there. And when finished, leave a strip of kapton tape above each shunt's location to be safe.

Shorts will cause very severe stuttering/hitches or even the card to shut off or possible damage.
I was mindful of that before re-assembly as I had read discussion here earlier about it. I also did place super 33+ tape as a safety net where the shunts could touch. No issues so far either, I think. However I take your concern and I’ll probably open it up in the next day or two to clean up. Should be simple enough with a desolde

Heres a port royal run and power limits. At stock (@114%) was 14500ish, so 362pt improvement. I haven’t had time to dig into results and see what may need touching up, what do you think?

2487112
 

·
Facepalm
Joined
·
9,883 Posts
I was mindful of that before re-assembly as I had read discussion here earlier about it. I also did place super 33+ tape as a safety net where the shunts could touch. No issues so far either, I think. However I take your concern and I’ll probably open it up in the next day or two to clean up. Should be simple enough with a desolde

Heres a port royal run and power limits. At stock (@114%) was 14500ish, so 362pt improvement. I haven’t had time to dig into results and see what may need touching up, what do you think?

View attachment 2487112
Looks perfect to me. Well done.
I'm guessing if your shunts aren't touching the backplate, the shunts you soldered may be thinner than normal. I guess the Panasonic shunts I stacked are a bit thicker than those.

I think you have a power limit warning due to those bios level MSVDD/NVVDD rail limits mentioned before, which cause a high TDP Normalized %. Unless your clocks are so high at 114% TDP that it still triggers a high normalized (at 114% TDP, you will see this rail trigger a power limit throttle at about 106% Normalized).

is your TDP at 100% or 114%? What's your core/ram clocks for PR?

Run that benchmark with HWinfo64 this time (make sure you 'expand' the power rails, the newest HWinfo64 has a bunch of rail windows collapsed on the left side), and you should see something like TDP Normalized 90% and TDP 75% if you're at 100% TDP. This causes a drop in your effective clocks however. I found that due to this issue, the TDP slider between 85% to 100% makes absolutely no difference at all (Because the "Default" MSVDD/NVVDD limits can not go below 100%).

To bypass the drop in effective clocks, set the TDP% to 114% if it isn't there already. If it's already at 114%, your effective clocks shouldn't drop too much.
 

·
Registered
Joined
·
16 Posts
Looks perfect to me. Well done.
I'm guessing if your shunts aren't touching the backplate, the shunts you soldered may be thinner than normal. I guess the Panasonic shunts I stacked are a bit thicker than those.

I think you have a power limit warning due to those bios level MSVDD/NVVDD rail limits mentioned before, which cause a high TDP Normalized %. Unless your clocks are so high at 114% TDP that it still triggers a high normalized (at 114% TDP, you will see this rail trigger a power limit throttle at about 106% Normalized).

is your TDP at 100% or 114%? What's your core/ram clocks for PR?

Run that benchmark with HWinfo64 this time (make sure you 'expand' the power rails, the newest HWinfo64 has a bunch of rail windows collapsed on the left side), and you should see something like TDP Normalized 90% and TDP 75% if you're at 100% TDP. This causes a drop in your effective clocks however. I found that due to this issue, the TDP slider between 85% to 100% makes absolutely no difference at all (Because the "Default" MSVDD/NVVDD limits can not go below 100%).

To bypass the drop in effective clocks, set the TDP% to 114% if it isn't there already. If it's already at 114%, your effective clocks shouldn't drop too much.
I used these Visay ones, they were quite thin: https://www.mouser.com/ProductDetail/71-WSL25125L000FEA18

TDP @ 114% for the one I showed in the previous comment, and this one. Here's another run from this evening (I scored 14 915 in Port Royal), slightly better score with the HWinfo64 you asked for. I was reallllly hoping to break 15k just for bragging rights, but I think I'd have to go even higher, and I imagine this is already well into diminishing returns. Just need to win the silicon lottery a bit better.

Average core clock: 2,121 MHz
Average mem clock: 1,329 MHz

2487129
 

·
Facepalm
Joined
·
9,883 Posts
I used these Visay ones, they were quite thin: https://www.mouser.com/ProductDetail/71-WSL25125L000FEA18

TDP @ 114% for the one I showed in the previous comment, and this one. Here's another run from this evening (I scored 14 915 in Port Royal), slightly better score with the HWinfo64 you asked for. I was reallllly hoping to break 15k just for bragging rights, but I think I'd have to go even higher, and I imagine this is already well into diminishing returns. Just need to win the silicon lottery a bit better.

Average core clock: 2,121 MHz
Average mem clock: 1,329 MHz

View attachment 2487129
Hi, thank you for the screenshot, but you didn't show your TDP% and TDP Normalized % on your hwinfo. it's below the bottom of your HWinfo stats, scrolled down.
 

·
Registered
Joined
·
201 Posts
More like every month or two, depending on how well you scraped off the conformal coating originally and how well you applied the paint to begin with. 1W shunts like yours (flush edges to middle) are FAR more stable than FE and MSI shunts however (2W shunts, depressed edges, those are very bad for paint because the uneven depressed edges create a lot of difficulty getting solid contact, and require an extra step (solder bridge) for solder shunt mods since the edge is lower than the middle, so a stacked shunt won't touch the silver part unless you create a gap fill solder bridge first).

Did you scrape the conformal coating off the edges of the shunts before applying the paint, as instructed and mentioned several times in both of the shunt mod threads? The better you scrape, the longer the stability will be. If you scrape the edges to when the copper starts to just barely show, might last awhile.

You'll know soon enough when one power rail suddenly starts getting far different than your original readings.

My original 8 pin rails lasted a few weeks until they suddenly degraded after opening for a repaste (Probably didn't scrape enough of the conformal coating off the shunts). Once that happened I could never fully stabilize them, mainly because 8 pin #1's shunt was right up against a choke vertically (unlike yours, which is horizontal), making accessing that edge almost impossible without getting paint on the choke. The GPU Chip Power shunt lasted maybe 2 months before degrading. PCIE Slot Power seemed to last the longest.

Basically, I know something is wrong with your mod because 105W for SRC and MVDDC is close to the original readings (SRC is usually 105-120W on 2x8 pin cards and MVDDC is about 80-95W), then of course you raise them a bit for drawing more power, but you also have to reduce them for a mod being in place. In your case, MVDDC is just too high. Having all the rails expanded in HWinfo64 would give a bit more information because MVDDC has continuity with one of the MISC rails, and also with PCIE Slot Power, GPU Chip Power and SRC also linked to each other and to the 8 pins.

Here is a good idea of the power balance you should be expecting from a balanced mod. This is all six shunts with 5 mOhm soldered stacked shunts on a 3090 FE. Don't laugh at the temps. You can see the GPU Board Power multiplier in hwinfo already. Almost 600W. Was running Overwatch at 4k (1080p+200% render scale) at 114% TDP limit and uncapped FPS. TDP Normalized is high because of hidden MSVDD limits I can't shunt.

View attachment 2485264
Following up several weeks later. Thanks to some great advice and guidance from Falkentyne, I have just soldered my Alienware 3090 (which is a reference 2x8). I have it running under an EK block with the active backplate. Sharing a picture of my soldering (I am really terrible but I didn't destroy anything) and Port Royal results.

While it is still showing power limited, it's definitely pulling more than the 390W Gigabyte BIOS. It does seem to pull 20-30W less than my Strix with the Kingpin 520W BIOS by looking at my Kill-A-Watt.

Let me know if this is looking right!
2488027
IMG_2348.jpg
2488028
 

·
Facepalm
Joined
·
9,883 Posts
My numbers didn't look right and was still power throttled. Reinstalled Gigabyte Master 390W bios and got much better results. Curious from Falkentyne and others if my power draw numbers now look right?
I can't say if your numbers look right or not because you can overclock your VRAM Much higher than mine so you get higher scores.
But in the first picture you were being throttled by something not related to shunts (Look at the TDP Normalized %). Your normalized% is much lower in the second screenshot. That's alot better.
 

·
Registered
Joined
·
201 Posts
I can't say if your numbers look right or not because you can overclock your VRAM Much higher than mine so you get higher scores.
But in the first picture you were being throttled by something not related to shunts (Look at the TDP Normalized %). Your normalized% is much lower in the second screenshot. That's alot better.
Thanks -- I've run a bunch more benchmarks and seems to be little power limiting happening now. I can re-run with a lower memory OC if that helps normalize a bit?
 

·
Facepalm
Joined
·
9,883 Posts
Thanks -- I've run a bunch more benchmarks and seems to be little power limiting happening now. I can re-run with a lower memory OC if that helps normalize a bit?
You would have to ask Elmor about this. Hardly anyone here knows how Normalized works, except normalized TDP is very similar to how "Hotspot" temp works--except it is NOT an absolute "maximum wattage" at all since it's normalized, not raw. It is any single internal power rail that exceeds its default limit by the greatest amount, with respect to how much any other rail exceeds its limit. For example if rail "A" has a default power limit of 50W, rail "B" has a default power limit of 200W and rail "C" has a default power limit of 500W, if rail A's current limit was 100W max, rail B's current limit was 300W max and rail C's current limit was 600W max, your normalized TDP would be 200% because "rail A" is exceeding its limit by a higher amount (200% of its base TDP, or double), than rail B (150% of its base TDP) and rail C (120% of its stock TDP).

Neither hwinfo nor GPU-Z reports any of the "hidden" limits that TDP Normalized can see (all of the unhidden wattage rails shown in hwinfo report to TDP Normalized already), or if they are, no one here wants to tell you what they are. Elmor thinks it's either a NVVDD voltage rail, a MSVDD voltage rail or a PLL rail. I told Martin (hwinfo) about this, but he doesn't know how to access what Normalized TDP is getting its data from, if the sensors are not visible.
 
  • Rep+
Reactions: cletus-cassidy
301 - 319 of 319 Posts
Top