The video is already few weeks but helps me to enjoy my vacation :D There will be fresh content on sunday. Hope you all had a good start into 2022. Thanks!
Schöne Ferien herr Bauer! ^^ Basically I'm seeing a 2080Ti reach the same clocks now that my 2070 boosts to with its many fewer cores :o thanks for sharing! (my gpu just gave out a couple days ago by the way-- I hope people realize stores only give 2yrs warranty but manufacturers often extend to 3 years. One call and an e-mail saved me a *year* of working for cash I don't have to throw at a new card!!)
Why didn't you make a pin connector on the variable resistor side too? Would be much easier to adjust resistance (GPU power) without disassembling the whole videocard
Tips: if you're "counting turns", make sure you buy/use _linear_ pots, *not* logarithmic ones. Also if you're assembling this and it will see a lot of use, buy a couple of regular panel pots. These trim pots (like the little blue used here), especially the cheaper ones, are usually only rated for ~20 cycles. They'll deteriorate faster with every adjustment cycle, and are liable to fail during operation if jogged around enough, which will ruin your day in the middle of a glorious full pot session. Having to tear down to solder in a new trim pot is gonna suck a lot more than the $2-3 extra for an ok panel pot.
If you connect your ground to the wiper of the pot you should be able to measure it in circuit by reading from center to the other (open) pin and subtracting from 50k. It should also be possible to run both MOSFETs off a single pot so they adjust simultaneously.
Also it might be advisable to use the 3rd pin of the potentiometer to create a proper voltage divider circuit instead of a 2pin rheostat (edit: apparently it's already a complete voltage divider on the PCB) and to put a limiting resistor between its input and output to prevent any unwanted limit leverage if the component goes short/open or if the contact of the wiper stutters when adjusting during operation. Limiting the range of a pot inside a safe zone is a common practice used on most schematics as those kind of pots are prone to wear and failure, especially when bought cheap. Also, using a semiconductor as a resistor is tricky as its resistance is non-linear and more importantly its value will vary with temperature.
You could have a couple of pots in series, you adjust one pot on the lower limit you feel safe to use and leave it set, the second pot you play around with and there isn't any risk of going to low. it might give you a peace off mind, at least untill you learn the MOSFET isn't temperature stable, it will turn harder oon or for lower resistance when it heats up...
I've nearly finished a project...led light ring for a microscope and doing just that. Set max voltage with one pot then use another to adjust brightness.
You can measure the value of the pot while powered-on by using the third pin. Just measure between the third pin and the middle pin (wiper), which will give you a reading inverse from what the MOSFET is seeing. So you can set a reading of 42K, which would mean the FET is seeing 42 - 50 = 18K.
Might be interesting if Elmor Labs, or someone, could replace the VR with a I2C controlled digital resistor so that you can have more reproduceable results and perhaps even programmatic changes to the values?
Ignoring the concern of MOSFET temperature coefficients, I'm not sure what the need for both POTs is... one should do the trick, no? Unless it's important to adjust each rail independently, which isn't mentioned in the video, so I assume you would want them to always be adjusted identically. Do any GPUs perform power deviation monitoring between the two rails? From my naive (inexperienced) perspective, I would worry about adjusting them independently without otherwise knowing that that's "safe" from a GPU stability perspective. Might be a moot point though if the MOSFETs aren't a matched pair. I definitely agree with others that "spoofing" the sense lines directly is the way to go. This trick seems too finicky for accuracy and repeatability.
Maybe an addition to this design would be to use digital potentiometers instead of physical ones, that would allow you to know the resistance you're changing it to on-the-fly because it changes in steps
One issue I could see with that is fine tuning - most digipot chips have between 100 and 256 discrete steps, where a fine-adjust analog pot can be tweaked much more slowly to creep up on the exact value you want.
You could measure the resistance in between the middle pin and unconnected pin, and by subtracting from max resistance you would know the resistance on the circuit side.
Come a long way since doing power mods with a huge fencepost soldering iron for sure! very interesting that modder ways are becoming cleaner and nicer.
BTW... the product page for the Elmor labs Power Limit Modifier is showing as discontinued. And specifically for the exact MOSFET he has chosen (the EMF02P02H) it is chinese mfr therefore from LCSC. those ones are currently only showing 2 in stock ATM. so what about alternatives eh? well surfing the digikey parametric search... the nearest similar IC which may not be exactly identical, but should be ok / good enough... this seems to be the the AON6411 (which has over 6000 inventory stock as of right now). they are about $2 usd each = $4 usd for a pair of two. of course you might also find other suitable mosfet from either lcsc, mouser, or arrow. whichever is your favorite supplier you are preferring
Fascinating stuff. Way out of my league but, very cool to observe. Never know, a year from now I may find the courage to tinker with an old GPU like that. Best wishes in 2022 to you too.
Was just about to retire my GTX1070, may have to give this a go. Elmor's new tools are awesome, seems like there is something new every week. Just saw a RaspberryPi USB-C switch too. Cant wait to see what comes next.
I think the buildzoid method is the most elegant technique by far. (In this method you bypass the resistor completely and give the sense pin on the shunt whatever voltage you need to make it think the power is good)
i imagine there is an easy way to build a microcontroller module to take measurements and display resistance and power information while it is running. perhaps even control tuning with a phone app if you want to go completely nuts!
The right way to use a potentiometer with a mosfet would be to put the terminals where it is always 50kohms on the positive and negative and the central terminal on the mosfet gate, because of the mosfet functioning it may go into a state where it will always be on because of that the gate acts as a capacitor and when charged the mosfet is always on
I was confused how you're able to drive a MOSFET in "analog" manner with a single resistor (gate resistance is virtually infinite), but the PCB does indeed have a second resistor which forms a divider. I wonder if MOSFETs that were sold on these PCBs were manually binned; there's always some variability in process. Or maybe resistor value was tuned. Anyway, this still is a hacky solution, just because MOSFET Rds and Vth vary with temperature. It seems that because sense connections of the shunt are separate, maybe a PCB which divides sense voltage with a potentiometer could work. Sense connections do not carry any current.
It works because changing gate drive voltage changes the MOSFET's resistance, no binning required for that. However the most effective solution but slightly harder to do would be to look for the sense lines going to the controller and doing a voltage divider there, that way it doesn't matter how much current is actually going through it it will always show the same power. For the power path it would be just putting wire so that there is low resistance. Note: the PCB of the GPU uses regular resistors but the PCB is meant for ones with sense line, that would make it extremely simple to get the sense lines since they got a easy and semi large soldering point. For MOSFET shunt temperature MUST be considered, at 90°C Rds on usually increases by 30-50% compared to 25°C and that can screw things up. Buildzoid did that controller manipulation some months ago.
I think it would make sense to use a digital potentiometer rather than the analogue wiper. Then they could set the resistance to particular values, reset them to stock and switch around programmatically without the guesswork on the number of turns.
12:05 And also, don't touch metallic stuff connected to live wires from a running system, if you're not der8auer and know exactly what you're doing 🤓😉🧑💻🧑🔧⚡
@@paskowitz I was about to say the same thing, evga kingpin and the Galax hall of fame cards. They are the only ones I know that are fully unlocked and ready for ln2 benching. That is the only reason I buy the kp card so I don't have to do anything besides go buy ln2. I have always wanted a Galax hof card but they are difficult to get here.
Could also wire the ground wires to a switch. So when you want to measure resistance, flick the switch, measure resistance, then switch back on and good to go
Just out of interest how many of you outside of Germany were taught practical uses for transistors? Saying that, when I went to school in the 1970's, the desks still had wee holes in them to put your ink pot, lol.
Would be nice for a bios mod to be invented (nvidia encrypt their bios, j just want the old days when I can flash a bios mod and voltage mod). Worried about warranty reasons of shunt modding… I would like something that doesn’t leave a trace behind, but guess I’ll keep waiting for that day.
I know people who can solder an entire smd pcb and if they use the same type of solder you simply cannot tell which one was handmade and which one machine made. Maybe they can help lol
You can always cross flash a bios from another gpu of the same model. You just need to override in nvflash. I put kingpin 3090 bios on my crappy 3090 and got a nice boost. Just make sure if it's 30 series you flash lhr to lhr and fhr to fhr.
@@falcie7743 I have an Asus strix 2080ti with an EK vector block flashed to a matrix. Asus GPU tweak 2 on screen display shows the core between 2145 to 2175.
People do that, but you need to have connections to get any BIOS modding software. I know a guy that made a working BIOS editor for Ampere and they got a cease and desist from Nvidia so they cant publicly release the software anymore. Also, the "encryption" that is put into BIOSes are fairly simple. First there is the NVflash signing issue, which can be bypassed via external hardware BIOS flashers. Then there are the checksums and UID lock in the vBIOS, which you can change if you know how to do it. That's pretty much it, no actual encryption like people say there is.
I love videos like this! Mods and anything else which I can do my self but perhaps either don’t know about or know how to do well enough is my favorite Der8auer content! Well the extreme content as well…almost all of Romans content is enjoyable come to think of it…🥴
It could be a good idea to use a stereo resistor that is 2 separate resistors on the same adjustable knob. It will keep the power draw balanced between the 2 PCI power inputs.
I modded my cheap 298€ KFA2 RTX2060 long time ago by soldering additional 20mOhm shunts in parallel, then flashed Asus Strix bios. It never hit power limit again, runs at 2000MHz all the time. Shows 65% power reading although I'm running 125% :D. A Vmod would let me OC more, but that's pretty decent, I won't be upgrading the cooler either, it's just fine like it is, around 75°C when gaming, hits 85 when running OCCT.
This type of adjustment requires absolutely no voltage sag or it changes the effect. Also, the RdsON of the PMOS is highly dependant on the temperature as well...
I shunt mod Electric Speed Controllers, just a big blob of solder does the trick for me, but a variable shunt would be much better for days I'm not trying to burn down my electric scooter wouldn't be a bad idea.
Lots of other things use shunt resistors to control power limits, for example LED drivers. Could this Elmore module be used to fudge with those too? We've always just changed the resistor value but that can have you end up with a lot of work for too much or not enough drive current when modding drivers, adjustable would be really nice and even open up having a two wattage option for running LED's as DRL or in a lower power mode and then having a high power mode for flash to pass or high beam use.
Why didn't you make a pin connector on the variable resistor side too? Would be much easier to adjust resistance (GPU power) without disassembling the whole videocard
This got me interested. I have an 2070 card with pretty low powerlimit. I think it may go way higher than it goes now but it hits the roof of the PL. I have to try this!
Why not just bypass/short the shunt, cut the trace from the shunt to the controller ic, then use a voltage regulator to send a voltage to the controller so it's completely independent from actual power?
Once you figure out the proper values for resistance, you could actually use a dial with selectable resistances so you don't have to worry about adjusting the potentiometer.
Maybe a dumb question, but could you not create a simple 0ohm bridge that can be engaged via an external switch, after the card POSTs up, thereby avoiding the safety check at the beginning? Or is the card too "smart" for that?
I know it'd be niche, but wouldn't it be a cool product idea to hook the mosfets up to a microcontroller with USB, so you can adjust from the PC itself.
I would really love to do this to my card, not just for extreme overclocking but for steady volt increase, howeeverI just don't have the courage right now to open up what is even a 400USD used card lol If I had a spare handy I might, but I'm too scared with current GPU prices to mess mine up. I saw another video where a guy hot glues another resister for the typical shunt mod, instead of soldering. I might seriously consider doing that because I really wouldn't trust my soldering skills.
some what related, there seems to be some odd behavior on 3070ti/3080ti cards in respect of TDP limit. TDP perf cap is provided but TDP is not at 100%. this issue seems to be present on AIB cards. could be worth looking into as a content piece?
With this kind of information, there migth be new GPUs in future with a pot built into them. if not that exactly, some headers to create a pluggable module.
How does a genius like you increase performance of laptop GPU's of Maxwell and Kepler with no boost algorithm, where adding voltage and/or current wouldn't achieve anything alone?
1935 MHz, 226 Watt and 80C is what my stock 2060S Runs at with Afterburner Power Limit Increased to 130%. It may be an architectural limit you're hitting. Going forward it will be interesting to see how the lower tier cards clocks compare to the larger Dies. For example the 6500 XT is supposed to hit 2.8 GHZ+ out of the box.
From my experience, the lower tier cards on average do hit higher clocks. But then again because of silicon lottery, some top end cards will run higher clocks than lower end.
@@digitaltactics9234 It's just a comparison, the 64-bit bus sucks. There has to be a "cheaper" design in their portfolio, 64 bit is easier to implement. Less traces, less chips, less power. Believe it or not many people buy a PC and seldom game with it, such a card being installed in a prebuilt would raise the profile to "gaming PC" and sell no problem making money for AMD and OEMs.
You can safely ignore the wire resistance. When you have a pot that is in several kiloOhm range at least, and i forget what flavour he used, but he's definitely in that order of magnitude, and near the ends of the range a pot tends to be a little rough anyway. The wire resistance is around 0.15 Ohm/m.
Not really, if not doing xoc but still want to see what cards can do, I'd suggest trying new vbios. But need to do research first on which ones work with which cards.
I hate these titles with a passion. I know they work really well for the general public and mass entrainment, but I really have to wonder if they're still required for relatively niche technical channels like this one. Every time I see such titles on actual content, I die a little inside.
told treat electricty like water it flows the same resistors are small dam walls to reduce flow thats how i see it. Akway something new & interesting on here Redkneck science i'll never do any of these things But its fascinating to watch it all . *Great channel de8rauer thought it was debrauer .
The video is already few weeks but helps me to enjoy my vacation :D There will be fresh content on sunday. Hope you all had a good start into 2022. Thanks!
Schöne Ferien herr Bauer! ^^ Basically I'm seeing a 2080Ti reach the same clocks now that my 2070 boosts to with its many fewer cores :o thanks for sharing!
(my gpu just gave out a couple days ago by the way-- I hope people realize stores only give 2yrs warranty but manufacturers often extend to 3 years. One call and an e-mail saved me a *year* of working for cash I don't have to throw at a new card!!)
With the tattoo missing, we knew right away! :D
Why didn't you make a pin connector on the variable resistor side too? Would be much easier to adjust resistance (GPU power) without disassembling the whole videocard
@@ToTheGAMES exactly, I seen it also before I read the comments 😀
Btw Elmor Is awesome guy with a lot things for hardcore hardware fans.
You forgot a link to this Ten Dollar mod :- )
Tips: if you're "counting turns", make sure you buy/use _linear_ pots, *not* logarithmic ones. Also if you're assembling this and it will see a lot of use, buy a couple of regular panel pots. These trim pots (like the little blue used here), especially the cheaper ones, are usually only rated for ~20 cycles. They'll deteriorate faster with every adjustment cycle, and are liable to fail during operation if jogged around enough, which will ruin your day in the middle of a glorious full pot session. Having to tear down to solder in a new trim pot is gonna suck a lot more than the $2-3 extra for an ok panel pot.
If you connect your ground to the wiper of the pot you should be able to measure it in circuit by reading from center to the other (open) pin and subtracting from 50k.
It should also be possible to run both MOSFETs off a single pot so they adjust simultaneously.
This! Was going to comment this myself.
Also it might be advisable to use the 3rd pin of the potentiometer to create a proper voltage divider circuit instead of a 2pin rheostat (edit: apparently it's already a complete voltage divider on the PCB) and to put a limiting resistor between its input and output to prevent any unwanted limit leverage if the component goes short/open or if the contact of the wiper stutters when adjusting during operation. Limiting the range of a pot inside a safe zone is a common practice used on most schematics as those kind of pots are prone to wear and failure, especially when bought cheap.
Also, using a semiconductor as a resistor is tricky as its resistance is non-linear and more importantly its value will vary with temperature.
You could have a couple of pots in series, you adjust one pot on the lower limit you feel safe to use and leave it set, the second pot you play around with and there isn't any risk of going to low. it might give you a peace off mind, at least untill you learn the MOSFET isn't temperature stable, it will turn harder oon or for lower resistance when it heats up...
I've nearly finished a project...led light ring for a microscope and doing just that. Set max voltage with one pot then use another to adjust brightness.
You can measure the value of the pot while powered-on by using the third pin. Just measure between the third pin and the middle pin (wiper), which will give you a reading inverse from what the MOSFET is seeing. So you can set a reading of 42K, which would mean the FET is seeing 42 - 50 = 18K.
Might be interesting if Elmor Labs, or someone, could replace the VR with a I2C controlled digital resistor so that you can have more reproduceable results and perhaps even programmatic changes to the values?
That would be awesome if Elmor or anybody else can do it.
I remember having adjustable Vmod on Radeon 9500@9700 :) Those times were fun
Ignoring the concern of MOSFET temperature coefficients, I'm not sure what the need for both POTs is... one should do the trick, no? Unless it's important to adjust each rail independently, which isn't mentioned in the video, so I assume you would want them to always be adjusted identically. Do any GPUs perform power deviation monitoring between the two rails? From my naive (inexperienced) perspective, I would worry about adjusting them independently without otherwise knowing that that's "safe" from a GPU stability perspective. Might be a moot point though if the MOSFETs aren't a matched pair. I definitely agree with others that "spoofing" the sense lines directly is the way to go. This trick seems too finicky for accuracy and repeatability.
bro with this kind of experiments I feel that I want to be your assistant, this is incredible
Maybe an addition to this design would be to use digital potentiometers instead of physical ones, that would allow you to know the resistance you're changing it to on-the-fly because it changes in steps
One issue I could see with that is fine tuning - most digipot chips have between 100 and 256 discrete steps, where a fine-adjust analog pot can be tweaked much more slowly to creep up on the exact value you want.
You could measure the resistance in between the middle pin and unconnected pin, and by subtracting from max resistance you would know the resistance on the circuit side.
Come a long way since doing power mods with a huge fencepost soldering iron for sure! very interesting that modder ways are becoming cleaner and nicer.
I do this with the power tubes on guitar amps. Totally never thought of it for a motherboard before.
BTW... the product page for the Elmor labs Power Limit Modifier is showing as discontinued. And specifically for the exact MOSFET he has chosen (the EMF02P02H) it is chinese mfr therefore from LCSC. those ones are currently only showing 2 in stock ATM. so what about alternatives eh? well surfing the digikey parametric search... the nearest similar IC which may not be exactly identical, but should be ok / good enough... this seems to be the the AON6411 (which has over 6000 inventory stock as of right now). they are about $2 usd each = $4 usd for a pair of two. of course you might also find other suitable mosfet from either lcsc, mouser, or arrow. whichever is your favorite supplier you are preferring
Fascinating stuff. Way out of my league but, very cool to observe. Never know, a year from now I may find the courage to tinker with an old GPU like that. Best wishes in 2022 to you too.
Nice! Thanks man! Greetings from Bulgaria! 🇧🇬
Was just about to retire my GTX1070, may have to give this a go. Elmor's new tools are awesome, seems like there is something new every week. Just saw a RaspberryPi USB-C switch too. Cant wait to see what comes next.
I think the buildzoid method is the most elegant technique by far.
(In this method you bypass the resistor completely and give the sense pin on the shunt whatever voltage you need to make it think the power is good)
that is total genius elmoor. using the RDSon like that. the perfect solution 👍
i imagine there is an easy way to build a microcontroller module to take measurements and display resistance and power information while it is running. perhaps even control tuning with a phone app if you want to go completely nuts!
The right way to use a potentiometer with a mosfet would be to put the terminals where it is always 50kohms on the positive and negative and the central terminal on the mosfet gate, because of the mosfet functioning it may go into a state where it will always be on because of that the gate acts as a capacitor and when charged the mosfet is always on
This was extremely clear and interesting, thank you!
EEK but the warranty. LOL Fun to watch and educational. Something I would never do. Fear rules my life!
I was confused how you're able to drive a MOSFET in "analog" manner with a single resistor (gate resistance is virtually infinite), but the PCB does indeed have a second resistor which forms a divider.
I wonder if MOSFETs that were sold on these PCBs were manually binned; there's always some variability in process. Or maybe resistor value was tuned. Anyway, this still is a hacky solution, just because MOSFET Rds and Vth vary with temperature. It seems that because sense connections of the shunt are separate, maybe a PCB which divides sense voltage with a potentiometer could work. Sense connections do not carry any current.
Basically a transistor is a current/voltage driven resistor.
It works because changing gate drive voltage changes the MOSFET's resistance, no binning required for that. However the most effective solution but slightly harder to do would be to look for the sense lines going to the controller and doing a voltage divider there, that way it doesn't matter how much current is actually going through it it will always show the same power. For the power path it would be just putting wire so that there is low resistance.
Note: the PCB of the GPU uses regular resistors but the PCB is meant for ones with sense line, that would make it extremely simple to get the sense lines since they got a easy and semi large soldering point. For MOSFET shunt temperature MUST be considered, at 90°C Rds on usually increases by 30-50% compared to 25°C and that can screw things up.
Buildzoid did that controller manipulation some months ago.
I think it would make sense to use a digital potentiometer rather than the analogue wiper. Then they could set the resistance to particular values, reset them to stock and switch around programmatically without the guesswork on the number of turns.
@@BobHannent not the same circuit nor simplicity though.. He dosn not sleep alredy don't add pain to his endeavor!
@@goldnoob6191 step by step
12:05 And also, don't touch metallic stuff connected to live wires from a running system, if you're not der8auer and know exactly what you're doing 🤓😉🧑💻🧑🔧⚡
I'm surprised a manufacturer has not worked with you to make a fully customized gpu. Fully unlocked, with all the controls adjustable.
That's pretty much the EVGA KINGPIN
Nvidia wouldn't like that i assume.
@@paskowitz I was about to say the same thing, evga kingpin and the Galax hall of fame cards. They are the only ones I know that are fully unlocked and ready for ln2 benching.
That is the only reason I buy the kp card so I don't have to do anything besides go buy ln2. I have always wanted a Galax hof card but they are difficult to get here.
Very interesting! Thank you!
Could also wire the ground wires to a switch. So when you want to measure resistance, flick the switch, measure resistance, then switch back on and good to go
Just out of interest how many of you outside of Germany were taught practical uses for transistors? Saying that, when I went to school in the 1970's, the desks still had wee holes in them to put your ink pot, lol.
When step by step perfect tattoo removal video is coming out 😝
I would never do this myself, but interesting and fun to watch!
Would be nice for a bios mod to be invented (nvidia encrypt their bios, j just want the old days when I can flash a bios mod and voltage mod). Worried about warranty reasons of shunt modding… I would like something that doesn’t leave a trace behind, but guess I’ll keep waiting for that day.
I know people who can solder an entire smd pcb and if they use the same type of solder you simply cannot tell which one was handmade and which one machine made. Maybe they can help lol
@@tetryds Worried about warranty reasons if you have this worry then you don't shunt mod.
You can always cross flash a bios from another gpu of the same model. You just need to override in nvflash. I put kingpin 3090 bios on my crappy 3090 and got a nice boost. Just make sure if it's 30 series you flash lhr to lhr and fhr to fhr.
@@falcie7743 I have an Asus strix 2080ti with an EK vector block flashed to a matrix. Asus GPU tweak 2 on screen display shows the core between 2145 to 2175.
People do that, but you need to have connections to get any BIOS modding software. I know a guy that made a working BIOS editor for Ampere and they got a cease and desist from Nvidia so they cant publicly release the software anymore.
Also, the "encryption" that is put into BIOSes are fairly simple. First there is the NVflash signing issue, which can be bypassed via external hardware BIOS flashers. Then there are the checksums and UID lock in the vBIOS, which you can change if you know how to do it. That's pretty much it, no actual encryption like people say there is.
use of mosfet to kinda short the resistor is brilliant
I love videos like this! Mods and anything else which I can do my self but perhaps either don’t know about or know how to do well enough is my favorite Der8auer content! Well the extreme content as well…almost all of Romans content is enjoyable come to think of it…🥴
It could be a good idea to use a stereo resistor that is 2 separate resistors on the same adjustable knob.
It will keep the power draw balanced between the 2 PCI power inputs.
I modded my cheap 298€ KFA2 RTX2060 long time ago by soldering additional 20mOhm shunts in parallel, then flashed Asus Strix bios.
It never hit power limit again, runs at 2000MHz all the time.
Shows 65% power reading although I'm running 125% :D.
A Vmod would let me OC more, but that's pretty decent, I won't be upgrading the cooler either, it's just fine like it is, around 75°C when gaming, hits 85 when running OCCT.
Nice work !!!
This type of adjustment requires absolutely no voltage sag or it changes the effect. Also, the RdsON of the PMOS is highly dependant on the temperature as well...
wow nice soldering station ... do you like it?
I shunt mod Electric Speed Controllers, just a big blob of solder does the trick for me, but a variable shunt would be much better for days I'm not trying to burn down my electric scooter wouldn't be a bad idea.
Lots of other things use shunt resistors to control power limits, for example LED drivers. Could this Elmore module be used to fudge with those too? We've always just changed the resistor value but that can have you end up with a lot of work for too much or not enough drive current when modding drivers, adjustable would be really nice and even open up having a two wattage option for running LED's as DRL or in a lower power mode and then having a high power mode for flash to pass or high beam use.
Why didn't you make a pin connector on the variable resistor side too? Would be much easier to adjust resistance (GPU power) without disassembling the whole videocard
This got me interested. I have an 2070 card with pretty low powerlimit. I think it may go way higher than it goes now but it hits the roof of the PL. I have to try this!
Why not just bypass/short the shunt, cut the trace from the shunt to the controller ic, then use a voltage regulator to send a voltage to the controller so it's completely independent from actual power?
Once you figure out the proper values for resistance, you could actually use a dial with selectable resistances so you don't have to worry about adjusting the potentiometer.
It's so sad how hard I laughed when you suggested that we learned about transistors or mosfets in school.
Maybe a dumb question, but could you not create a simple 0ohm bridge that can be engaged via an external switch, after the card POSTs up, thereby avoiding the safety check at the beginning? Or is the card too "smart" for that?
i wonder if stereo/dual multiturn potensiometer better for this since you can adjust both connector power draw at the same time
Not into doing xoc myself but it sounds like a good tool for those using ln2/lhe
I know it'd be niche, but wouldn't it be a cool product idea to hook the mosfets up to a microcontroller with USB, so you can adjust from the PC itself.
When are we going to see those new minus pads??
thanks for sharing video
Why not run both off the same variable resistor?
I have no ides what that thing is, besides it's for overclocking, but sounds cool :)
Where can I find this shunt resistor online? Would love some help
I would really love to do this to my card, not just for extreme overclocking but for steady volt increase, howeeverI just don't have the courage right now to open up what is even a 400USD used card lol If I had a spare handy I might, but I'm too scared with current GPU prices to mess mine up. I saw another video where a guy hot glues another resister for the typical shunt mod, instead of soldering. I might seriously consider doing that because I really wouldn't trust my soldering skills.
love it
some what related, there seems to be some odd behavior on 3070ti/3080ti cards in respect of TDP limit. TDP perf cap is provided but TDP is not at 100%. this issue seems to be present on AIB cards. could be worth looking into as a content piece?
you should use double potentiometer it would be much better
Interesting video
Amazing, can I do this mod on the 6600xt?
With this kind of information, there migth be new GPUs in future with a pot built into them. if not that exactly, some headers to create a pluggable module.
How does a genius like you increase performance of laptop GPU's of Maxwell and Kepler with no boost algorithm, where adding voltage and/or current wouldn't achieve anything alone?
Did der8auer ever find out how to unlock the 6900xt to get more voltage, or more frequency?
could u make a tutorial how to do this kind of mod for gddr6x voltage regulator? i saw one on ahoc channel and i looked very interesting
1935 MHz, 226 Watt and 80C is what my stock 2060S Runs at with Afterburner Power Limit Increased to 130%. It may be an architectural limit you're hitting. Going forward it will be interesting to see how the lower tier cards clocks compare to the larger Dies. For example the 6500 XT is supposed to hit 2.8 GHZ+ out of the box.
From my experience, the lower tier cards on average do hit higher clocks.
But then again because of silicon lottery, some top end cards will run higher clocks than lower end.
6500 XT 64 bit on the bus pcie 3.0 & 4gb on the vram why bother. 🤣
@@digitaltactics9234 imo if the market wasn't so fucked for us, I don't think amd nor nvidia would have released these cards.
@@digitaltactics9234 It's just a comparison, the 64-bit bus sucks. There has to be a "cheaper" design in their portfolio, 64 bit is easier to implement. Less traces, less chips, less power. Believe it or not many people buy a PC and seldom game with it, such a card being installed in a prebuilt would raise the profile to "gaming PC" and sell no problem making money for AMD and OEMs.
can you test it on mining ethereum ?
Does the cables add any resistance for the initial settings on the adjustable resistors?
You can safely ignore the wire resistance. When you have a pot that is in several kiloOhm range at least, and i forget what flavour he used, but he's definitely in that order of magnitude, and near the ends of the range a pot tends to be a little rough anyway. The wire resistance is around 0.15 Ohm/m.
guys no one is talking about the werbevideo sign in the upper right corner
leute keiner kommentiert an dass das da oben rechts werbevideo steht
Imagine this power mod with the unreleased 3090ti 😂🔥
The kingpin version can already use 1000W.
we didn't get to see it blowup!?
Use photo resistor instead of potentiometer.
More RGB more FPS! finally it's true.
why using so long cables ?
Yeah you can't measure resistance while it's running, but you could measure the voltage drop and determine the resistance that way...
Is there any point in this for non-XOC applications?
Not really, if not doing xoc but still want to see what cards can do, I'd suggest trying new vbios. But need to do research first on which ones work with which cards.
I can finally shunt mod my imagenery gpu.
can you make 1650 STRIX 6 GB ??
☻☻☻I fucking love you Man! enjoy your vacation!☻☻☻
this elmor labs is very cool. why doesn’t anyone make a vbios hacking tool?
And does this work on AMD cards too???
Iirc Amd doesn't use shunts like Nvidia, so it would not (there's other things you can do on md that you can't with Nvidia, like custom vbios though)
Freak out 😂
I can see the RTX 4090 Kingpin card coming with this mod already built in.
All the kingpins already fully unlocked.
Upon watching this video you get an honoree Masters Degree in electronics!
Ne. Die Dinger fangen an zu kratzen mit der Zeit.
I hate these titles with a passion. I know they work really well for the general public and mass entrainment, but I really have to wonder if they're still required for relatively niche technical channels like this one. Every time I see such titles on actual content, I die a little inside.
Just leave the GPU cards ALONE!
this is sketch AF bruv
Why the hell is this an analogue mod for a digital controller? wtf?
what are you asking exactly? about shunt mod itself?
Signals measured/modified here are not digital.
told treat electricty like water it flows the same resistors are small dam walls to reduce flow thats how i see it.
Akway something new & interesting on here Redkneck science i'll never do any of these things But its fascinating to watch it all .
*Great channel de8rauer thought it was debrauer .