Why shunt mod kills graphics cards ? 

Thank you I wanted to know what that Mod does.

@@senditall152 It increases/deletes the power limit set by Nvidia.

@@senditall152 The shunts are used to measure current by measuring the voltage drop across it as current flows through it (ohm's law and allat). The power management circuitry is pre-programmed with a known (set by Nvidia) ratio of voltage drop to current. So, when you decrease the resistance of the shunt (by adding a second one on top of it, covering it in solder or liquid metal, etc), you are reducing how much voltage drop will occur for a given current flow. This fools the card into thinking it's drawing less power than it actually is. If you cut the resistance of the shunt in half, you may be pulling 500W through it, but as far as the PM circuit is concerned, you're only pulling 250W. A valid method to increase power limits without messing with firmware (for the purposes of extreme OC, usually), but you sure as hell better know what you're doing or something like what's shown in the video might happen :D

Unless you're a complete moron, shunt mods are pretty easy to pull off. Problem is the person here shuntmodded an RTX A2000, which is PCIe power only and is already pulling 70W out of the factory. You'll either fry the motherboard or like we see, blow a mosfet

"you probably shouldn't be using one". There you go, fixed!

Thank you.

I'm so glad eth gpu mining ended forever! But it has damaged a perfect 2019 PC market that will never return. On top of that now unaware buyers can be scammed on the used market with all the ex-mining cards almost beaten to death and/or with flashed bad custom bios.

​@@Bsc8 Many youtubers already did prove that mining doesnt really damage the gpus, might need some new fans and repaste but thats it IMO. I dont think anyone would be getting "scammed" by buying a mined on gpu. I myself use a mined gpu and its perfectly fine, gpus during mining run no where close to the max power and the temps are also lower compared if the card was being gamed on. Also the A2000 i think was the only card that was being physically modded for mining so that might be a card you woudn't want for use just because it was modded and running above the specified power limit by nvidia. Also bios modding for mining is not a thing after RX5700XT series... nvidia and amd have it all very locked down and the only thing that would be possible is to flash other vendor vbios (like asus bios on a msi card or asus ROG bios on a TUF card) so not much that could happen there.

@@xpatrikpvp yeah im prefer ex mine cards imo because i know they have been running at like 10% of its intended power draw for its lifetime, the biggest issue id watch for is bios issues on the card bc they flash low power low clock bios

​@@iananderson8392and if the card survived mining, that's basically a stress test and it will probably not die on you because of a manufacturing defect.

Kids don't care.

@@Look_What_You_Did This channel is for enthousiasts, kids are here to learn, not get mindless entertainement

@@ZERARCHIVE2023 I'm twelve. Been doing similar work for two years now. The scale means nothing to everybody.

the tip of the iron is your reference. now back to some mindless desoldering entertainment.

That makes sense, I had no idea why A2000s where being modded for OC. For some OC competitions GPU's like the GT 1030 are used as there cheep used, pro cards like the A2000s are never relay used for OC competitions.

I was considering shunt modding, but was afraid that the VRM won't be able to handle the increased power long term. Now we know.

@@ZaPirate we did the wrong shunt mod tho

@@ZaPirate Normally you don't mod a low power Quadro successor card that is barely spec'd out to handle stock operation, but a fullsized card with an actual beefy VRM instead

Believe it or not but this was indeed easy.

It is like a lot of handcrafts - you learn it by starting simple and improving & challenging yourself more and more over time. Im doing very simliar SMD soldering like northwestrepair in my job.

The most crucial bit is sorting out magnification good enough where you don't just THINK you can see what you're doing, but which is good enough to where you will actually SEE what you're doing. At this scale which goes down to 0402 i think no sort of worn loupe is going to be enough, as opposed to a watchmaker you can't just bring the work to your nose, you actually need a microscope. Not seeing your tool well enough prevents you from acquiring stable hand posture and correcting your movement. If you're otherwise good with fine motoric action and soldering, after just a handful hours you might have enough dexterity to actually do this at this scale for real with a medium high degree of success. Or if not after some hours, then after some hundred hours. Even people with neurological disorders which affect their hand muscle control learn to successfully microsolder, so odds are, you can learn to do it. Human brains and hands are amazing! And best of it is, it's not surgery operating on a human. It's just stuff. If you kill it, nobody dies! There's plenty of landfill junk out there to practice on.

@@SianaGearz Wise words.

Undervolting can be a nice and necessary thing to do if you got good silicone but the problem today is that many things have locked power limit and because of this it won´t overclock well because even if undervolted it gonna want more current with higher speeds. I probably gonna do a shunt mod on my 3070 the day my current overclock is not enough to get some more useful power out of it but it depends, but the thing is don´t overdo it, to just add add a few % to the power limit and it can make a huge boost to performance. I been overclocking everything I can since the 90s now.

Actually he was just using the fuse during the repair, to avoid losing another coil by a short circuit or anything like that during the tests... at the ending (8:52) he said that he put the coil on the place

@@felipesb2 thanks, I missed that!

That inductor is on the input side. Which is why it's next to the shunt resistor. If it has blown an output inductor, the core would be dead.

It will have a negative impact on the circuit, the coil is an energy storage and filtering device, the fuse is just a piece of wire, check the outputs on an oscilloscope of one with the coil vs one with the fuse and you’ll see the differences. Terrible idea to replace the coil with the fuse. Yes the card works, but how noisy and how much ripple have you now introduced onto that power rail?!

It's in the 12v input indicator. Will do nothing to how the card works.

​@@motomario8795look at the schematic of that card, the schematic tells you what that circuit will see.

Thanks!

Thanks!

and do it to something with more than a 3phase vcore xd

I'm pretty sure the PCIe power consumption is tracked separately from the shunt circuit, which usually measures the 12V power connector(s) only. Some cards don't use the PCIe power at all.

@@BladeScraper The A2000 doesn't have any extra power connectors. It draws all its power from the PCIe slot. Nvidia cards have a shunt on both the PCIe slot power and the additional power connectors.

@@gawrbage gotcha, I don't know about this specific GPU. I was just talking in general: most of the time the power is tracked separately for the PCIe and for the 12vin connectors. In the case of this card, yeah yikes - definitely shouldn't do a shunt mod on a card that only uses PCIe power. Cryptominers are insane 😂

You're not wrong, but many boards (and seperately-powered mining risers) are 'safe' well-beyond that 75W limit. IIRC, it was once common for ATI/AMD cards to draw more than spec. Typically, only PCIe 1.x and 'OEM PC' motherboards are designed for 75W Absolute Max (or less). Related: PCI-E 1.0 (no 'a', etc.) had a lower max current specification on the x16 slot. (You can find proof in *archived* PCI-SIG documentation.) Early-adopters of PCI-E sometimes followed that early spec, and many boards were not built for PCI-E 1.0a and 1.1's specification. (Personally, I've ran across at least half a dozen HP/Compaq, Dell, etc. motherboards with some variant of this problem) For a short period of time, eVGA (and others, probably) were making a molex->PCIe slot 'power injectors' to give a low-resistance current-source +12V to your PCIe 1.1 or newer GPU. The adapter plugged into an x1 slot's +12V and Gnd pins, giving the *adjacently* slotted GPU more-stable slot power (across the short traces between slots)

@@gawrbage Doesn't cryptominers use some accesories, where this card might be pugged to some connector where the PCIe power really comes from power suppply 12v directly?

list of equipment used should be under older videos

thanks !

Thanks 👍

Litterally noone has ever said a shunt mod is good for reliability. If you want to shunt mod for benching then do it. If you want the card to be daily useable afterwards install the shunt mod via a switch

I learned about shunt modding from LTT and debaur. maybe it is not too bad as long as you start to realize how high risk it is. risk killing the gpu for a slight uptick in performance instead of having reliability. Riding on the edge. My biggest gripe are the people who do it for money are passing the risk to the customer without any risk for themselves. Thus lazy work and greed will make them go dangerous shunts for the highest gains, only for the card to fail after some time after it leaves the "shop"

RSRS?

He's cheap. His words.

Those cards went for 700 USD each just a year ago and are still going for 440 to 720 USD a pop, because the model is basically still brand new. The repair of all three cards probably took less than 2 hours in total, so even if the hourly rate was expensive, you'd still rather spend 1500 to 2000 USD to replace the cards instead of may be the 200 bucks it cost to repair them?

​@@pseudonym3690these cards have been available for $250 for a few months

He's affordable, many gpu fixes don't take long. The short easy repairs are the gravy, the odd harder repairs are videos,.win win.

Circuit overload.

ahahahaha, he's such a chatty Cathy. I knew he wouldn't explain much, if anything about why they did this. I had to read the other comments to find out.

No am not 🚫

There was nothing on there, simple linear regulators, one switchmode buck converter at most or just one beefier linear, the video processor, the VRAM, some VGA port protection against static, a crystal oscillator, and a parallel EEPROM firmware IC. But think back to early 90s the typical VLB card, that was long. Or in the 80s there were many cards made from discrete low integration ICs which were full enclosure length. Remember back in the day there was a rack at the front of the enclosure to slot your super long cards into, so they are secure and don't hang through, don't flop around. We lost that useful retainer when in the early 2000s, long cards died out, would have been good to have nowadays would it not. Of course that front retainer only directly works for ISA and VLB PCBs but PCI PCBs which are that long would come with a thing that looks like a handle attached to the end, that would offer some diagonal offset upwards and slot into the same retainer as well. But those were rare devices which needed that, none of them were for commodity office/home PC. I did have one card which used the retainer, a CT3980, but ISA cards that long were actually more common.

They really are fantastic little cards, too bad a symptom of them using cherrypicked silicon is the massive increase in price

Basically a workstation 3060. We use these too.

They keep voltage spikes at bay. Good PSU should do that on its own.

@@northwestrepair Ah, so it's a trade off between securing against overvoltage or overcurrent. Fhanks for the answer.

@@Donnerwamp Nah, it is to stop the GPU from polluting the 12V power rails, limit inrush current, etc. The GPU itself is usually better off not having input inductors, but the power supply or other components might be unhappy about the selfish GPU.

@@northwestrepair is the coil before or after the FET, in which case isn't it's purpose to reduce the switching noise as part of an LC filter? Replacing it with a fuse might turn it into a C or RC filter, which may be OK. It would be interesting to see the scope of a power supply phase with and without a coil under different GPU loads!

Yes but it's not cost effective.

@@northwestrepair in what sense? For the chips? Or both the chips and the install labor? Genuinely curious because I saw someone online added 8gb to the card, my wife has 3070ti suprim x and it already doesn’t have enough vram for high settings 1440p so was curious if it can be done at all.

One doesn't simply "add 8gb" one replaces every memory chip with 2x capacity chip

@@vvb890 This year at Computex, Taipei modded cards with double the VRAM were shown. But there's a problem with the Nvidia driver. So it's not solely a hardware problem and as northwestrepair said it's not worth it if you can't DIY and can't get the chips for cheap.

Well, for one thing, make SURE everything is getting cooled, like the coils that melted. They create quite a bit of heat under heavy overload.

@@theLuigiFan0007Productions Exactly... which is why I wonder if they could have been killed due to heat, rather than the shunt mod.

It tricks something either GPU or voltage controller into thinking it's drawing or receiving far less current then reality, meaning you can vastly increase power limits on a GPU. It's typically used for XOC, where you need extra power to push clocks higher and cooling is sufficient, however if cooling is in sufficient or VRM is not properly specced for the new load you get this

@@mercuryrising9758 Thank you very much for this explanation. Now I have an understanding of it.

@@Vile-Flesh no problem I love helping

Putty may be easy to apply but not easy to remove. Pads allow you to reassemble and they last for years. Putty lasts maybe 2 years and dries rock solid. I don't like it. I think pads are better long term.

@@northwestrepair What about copper mod?

@@key462 risky. You can easily short memory and 1.8v rails killing the entire memory stack and the core.

@@northwestrepair I thought using kapton tape handily prevented that?

@@northwestrepair Does that mean the memory die? conductive? I saw alot of people use capton tape/nail varnish to cover the exposed part of the caps and such arround the memory but not the memory itself that makes contact with it

That's making the shunt mod ineffective. Shunt resistor are in the milliOhms range, switch can be in the hundreds of milliOhms rangee (for high quality one) so in the end your shunt mod is most likely increasing TDP by 1 or 2%, which is ineffective.

@@TheQuentincc have you ever tested the resistance of a wire using a milliohm meter? Or a high quality switch? You are quite mistaken. To put it bluntly...

@@JatXoc Alright then, I might have bit a quick to answer but nertherless it's not as good as you seems to think. For your setup you will use at best AWG12 and at worst AWG16 wire, which translate to 5~13Ohms/Km so that's about 0.5mOhms~1.3mOhms for only 10cm total of wire. Then fine let's have a look at switches datasheet, depending on price (between $5 to $50 for PROPER and BIG 5~30A switches) the maximum expected resistance is about 50mOhms~20mOhms, I'd guess they won't specs a switch at 50mOhms if it's actually well below that so it's probably in the 40~15mOhms. So in best case scenario, without even including a shunt resistor, over a 5mOhms existing shunt resistance you will get a MAX 33% increase in TDP, by adding a 5mOhms resistance on your setup you'll end up at MAX 25% increase in TDP... and if we want to go the cheaper way you'll get about 10% increase in TDP. All of that for a messy setup when you can replace the shunt with a lower value, for instance a 4mOhms shunt is priced in cents while giving you 25% increase in TDP (and it's 100% sure you'll get about 25%), or mount another one on top like a 20mOhms for 25% increase in TDP. Netherless, of course shunt mod is inadapted on such power design like the A2000, it's already capped at 70W from the factory, which is quite some work to handle for little cooled mosfet (let's say core load is about 60W with 0.725v according to TPU review, that translate to ~80A total, so 27A per phase so about 90% efficiency which is ~2W of losses), shunt mod is mostly intended on higher end card with better power design like RTX 3080 and higher.

thats too much. 3090 memory temps usually follow hot spot. Check that. If your hot spot is 100+, you need to redo thermal interface asap before you kill it. Also clean it up. maybe that helps.

It’s an untouched card I purchased from eBay about a year ago. RMA will be impossible I’m sure. Yeah I know lol. I don’t want to ruin it. I purchased the coolmygpu cooling plate for it but it hasn’t arrived yet. I also purchased gelid pads from .5-3, I’m going to measure what the cooling plate doesn’t cover and replace accordingly. There’s very little info on the 90Ti ftw3.

just dont unless you really know what you're doing, the problem is less the cooler and more people not knowing the limit of their hardware. these mods will let you destroy any gpu regardless of how good the cooler is. even liquid nitrogen wont stop it welding components to the pcb if you push too much current.

I'll go with performance over size any day 😋

@@Meganomaly3 Swan Lake?

ahahahaha Excellent