Graphics Card Necromancy - Sapphire Radeon R9 390 Nitro

[@coccoborg]

yay! I'm early Edit: I have a sapphire 290 vapor-x with the same crappy power stages: they heat up to 100+ °C under a synthetic load. The RX 480 I replaced it with runs much cooler. Your content is great, very clear and easy to follow for anyone trying to do similar repair, thank you so much for posting :)

[@AaronAverett]

Thanks! Are you sure your Vapor-X has these same power stages? The photos I've seen of those show the same IR6811s and 6894s that the regular 290/X has. Those are actually pretty stout, but they're not very efficient, so they run super hot. The Vapor-X is a pretty neat card, though. I've always wanted one.

[@coccoborg]

@@AaronAverett yup, pretty sure. R9 290 Vapor x Tri-x, here's a nice picture of the weird VRM images.bit-tech.net/content_images/2014/06/sapphire-radeon-r9-290-vapor-x-oc-review/sap290vxoc-13b.jpg

[@patrickvinas]

Hi Aaron, great video, it helped me isolate the short on my 390 Nitro to the memory controller's 270 uF capacitor. I'm getting ready to order replacement parts from mouser, but I need to replace the fuses and I'm not sure what to search for. Can you help me out?

[@AaronAverett]

This is the part I used. www.digikey.com/en/products/detail/littelfuse-inc/0476010-MR/5048216?s=N4IgTCBcDaIGIDYEBZkGEAqBaAcgERAF0BfIA

[@elpre11]

@@AaronAverett digikey vs aliexpress?

[@AaronAverett]

@@elpre11 Digi-Key.

[@guarand6329]

The resistance changing, is the power from the meter slowly charging a cap or caps. That's a good sign as far as looking for shorts goes.

[@blazeblake2920]

Wow! What a great tutorial I was to change the phase but I have melted exactly the same capacitor, did you know what capacitance it had? did you find the change piece?

[@richm7707]

The caps you're replacing are 10µF and the current rating for the white fuses you're replacing are 10A. Do you know what the current rating is for the brown fuses labeled R? NF1206 & NF1203.

[@duhnboa5447]

Great video! Really helpful in narrowing down the cause of my blown PCIe fuse, thanks! I seem to get 3.3 Volts on my test setup, but no 5V (78M05), no Vcore and no memory voltage (last two probably because the PCIe stuff doesn't work?) What I could gather so far: - the only blown fuse is the one next to the PCIe connector - the small one (F1801) next to the 78M05 on the card's backside measures around 1.6 Ohms (not sure if that's normal) - Vcore and memory resistances seem fine - memory controller to ground measures ~110 Ohms - you said 69 Ohms was high, what do I make of that number then? - the chokes you measure around 13:00 are roughly the same as yours I do own a cheap hot air station, but I'm not experienced in using it, so I'm hesitant about just swapping the memory controller's IR3553 (already have a few spares lying around).

[@AaronAverett]

When you have a failed fuse, you most likely have a short circuit somewhere. Check the resistance to ground on the card side of the fuse - that is, the side that's connected to the rest of the card, and not the PCI-E connector. Finding the short will be a challenge if there's no visible damage, especially with the fuse blown. The memory controller resistance is probably fine. That's higher than any I've seen, but not so high that it's outside the realm of plausibly normal.

[@duhnboa5447]

​@@AaronAverett Ah jeez. Sometimes you have to do the same thing five times over to see where you went wrong :D The resistance of "blown fuse (PCIe) to ground" was roughly the same as memory controller to ground. I measured time and time again, always within 10 Ohms of one another, but it just didn't occur to me that those were suspicious until I rewatched that part of your video (not for the first time either!). Turns out, it's not supposed to be input = output of a DrMOS! I kept searching and searching so frantically for a short to ground that I missed that. Next step, gotta grab some practice boards and bring my hot air skills up to speed so I can remove it. Hopefully nothing more is fried. Again, thank you so much! (And sorry about my linguistic acrobatics, I'm just a happy German, and a tired one at that :) EDIT: Oh, and another question: What temperature do you use for hot air work?

[@AaronAverett]

Yeah, you should have a few thousand ohms to ground on the 12V of the IR3553s, and a few tens of ohms on the output side. If you have near zero ohms from the input side to the output side of them, then that's a giveaway that one of them has failed and is shorting the 12V input to the output, which is a pretty common mode of failure. Regarding the temperature, I usually set the hot air station to 400 or 450, but it depends heavily on what I'm doing. For the IR3553s on this card, you'll likely have to use a lot of heat, as there's a lot of copper in that part of the board that you have to heat up before the solder will melt. The truly "proper" way to do this is to use a preheater to get the board up to ~170C, and then use the hot air station at 300C or so to heat it up the rest of the way. The real tricky part of this isn't going to be removing the failed IR3553. That's pretty easy to do with enough patience. The real hard part will be soldering a new one on there and getting all of the pins soldered correctly.

[@ktrontrontron]

Awesome video! What size footprint caps were they? Was V-Core labeled on the board?

[@AaronAverett]

Vcore isn't usually labeled on the board as such. You can identify it in a few different ways, but the easiest is to just look for the biggest group of VRM phases. Generally, you only have one rail with six monster power stages. Regarding the capacitors, I they're an 0805 size. These are the caps I used to replace them. www.digikey.com/en/products/detail/murata-electronics/GRM21BR61E106MA73L/4905533

[@ktrontrontron]

@@AaronAverett Thanks for providing a link! Awesome work!

[@xGxPhantomZzz]

So my card also kinda died (always crashed when getting warm i.e after hours of gaming or sometimes youtube) and now is completly gone. It went from crashing after minutes to not even turning the fans on anymore. Measured the 12V Resistance like you did and well, why is my resistance rising and not falling like yours did? It starts at 0 and just rises and doesn't slow down or stop. Also it shorted the case after crashing, so if you touch the case after it crashes it zaps you o.O

[@xGxPhantomZzz]

The resistance you measured next to the PCI Pins doesn't even exist, its just nothing.

[@AaronAverett]

Don't worry about the exact resistance values. If you don't get either a zero or a totally open circuit when doing that test, you're probably good. The reason it rises is that the meter is charging up all the capacitors on the board just a little bit, and as it does, the current it's providing, which is used to measure the resistance, slowly drops. Given that your card deteriorated slowly, there's a pretty good chance that your failure is actually in the GPU itself, much like we used to see with the Xbox 360. Nevertheless, it would be worth pulling the heatsink off and checking for continuity across the fuses. If you find any of the fuses are open, then you can troubleshoot just that part of the board.

[@xGxPhantomZzz]

@@AaronAverett I've got the board infront of me and i'm pretty sure it died a heat induced death. Thermal paste is not existing anymore, its just crumbling off and well uhm while cleaning the GPU it kinda got scratches to I think I killed it, or are slight scratches not a 100% death sentence

[@AaronAverett]

The processing logic is actually on the underside of the silicon. Scratching the top won't really hurt it unless you scratched it so badly that it cracked. Some competitive overclockers actually lap the dies to improve mating with the heatsink, since they sometimes come from the factory with scratches or unevenness. It's possible that you have an actual failure of some kind if you weren't getting a picture at all, but it's absolutely worth a shot to clean it off and replace the paste and any disintegrated thermal pads.

[@xGxPhantomZzz]

I just measured the resistances and fuses, they are all good except for the one near the PCI-Pins, it measures nothing, even if you use the "buzzer" setting.

[@MTS_IT]

also, i have the same card for repair... nothing is blown... just shows black screen 10 secs into usage... :/

[@AaronAverett]

So like, you get a picture on the screen, windows start booting, and then the screen just goes black? Or, you get into Windows, 2D stuff works OK, but you start a game, and the screen goes black after 10 seconds?

[@MTS_IT]

@@AaronAverett boots info windows but screen goes black after some like 10 secs

[@AaronAverett]

If you use DDU to remove the driver, and set the "don't download drivers" flag, does it still do that?

[@MTS_IT]

@@AaronAverett yup. even with new windows (clean install): I've even tried at my friends house (we actually have the same GPU)

[@AaronAverett]

Have you tried all the normal steps, like cleaning the heatsink, replacing the thermal grease, et cetera? If not, try that. If yes, try this: 1. Remove the OS drive, so there is no bootable volume attached to the system 2. Remove the metal shroud from the back of the card, so you can get a finger on the back side of the area where the VRM components are 3. Plug the card in, start the system up, and let it idle for 10-15 seconds 4. Feel the back of the card in the VRM area, and see if you have any spots that are getting really hot. 5. If you find one, troubleshoot the VRM phase located in that area. WRT step 4, warm is normal - what you're looking for is, like, burn your skin hot. When troubleshooting this, look at the bootstrap capacitors first. I have a video where I dealt with a problem like this on a 290X, which is very similar. If you don't find a phase that is way hotter than the others, your problem is most likely a BGA component failure, but you can rule out a power issue by checking which VRM rails are still running after it crashes. To do that, it really helps to test on a bench with the card on a riser like I showed in the video. It would be a real pain with it in an actual case.

[@latigoirusta9792]

Hello brother, I am from ARGENTINA, I did all your steps as is and everything is perfect, could it be the voltage controller of the memories ??? How do I measure the mosfet?

[@AaronAverett]

When you press the power button on the system, what happens? No fan spin at all? Fan spin but no post? POST stops with a graphics card related error code or diagnostic light?

[@latigoirusta9792]

@@AaronAverett The only thing that informs me, is that radeon stopped working and is ticked, it returns, it ticks and returns until the screen turns black and does not work anymore. The fans work normal as always.

[@Jonathan_2099]

came here from the [H] forum...Great video

[@blazeblake2920]

I already found and bought the fuses, but I don't have the capacitor if you know its capacitance I think I can find it

[@AaronAverett]

Watch the follow up to this, I’m pretty sure I mentioned it.

[@stefanogrillo6040]

Hello i have a defective/unstable 390 sapphire card. I just checked everything and the pcb is ok, do you think that screen artifacts might be caused by the gpu's oxidized solder or is the mem modules?

[@AaronAverett]

Cracked or corroded solder balls are a possibility. It's also possible, perhaps even likely, that the problem is not the solder balls, but the tiny interconnects that connect the silicon to the substrate. It is possible to reball a GPU or memory chip (and the memory chips can be replaced), but it's quite difficult, particularly without special automated equipment. Edit: As to whether your problem is specific to the GPU or memory chips, it's hard to say. I'd be inclined to suspect the GPU first, but I'm also not able to inspect the card for visible damage like you are.

[@stefanogrillo6040]

@@AaronAverett Thanks

[@vapi27V]

Hi in mine the 3.3 resetable fuse is at 2.7v instead of 3.3v any ideas?

[@AaronAverett]

Check the resistance to ground on each side of it and the resistance across the fuse. As I recall, you should have about 350 ohms to ground on the 3.3V rail on this card. Resistance across a fuse should be near zero. If you have high resistance across it, I'd suggest replacing the fuse. If you have low resistance to ground, you have a short that you need to investigate.

[@vapi27V]

@@AaronAverett I think I have a short around 6r8 inductor ! I check few Cms capacitor but they are not short out of circuit!

[@chunny1980]

anyone got any idea what the issue could potentially be if the fuses for the right hand side power connector are blown?. its the 2 with the letter R on?. any help would be much appreciated. I've got tools to test it but i'm not really sure as to what i should be checking first as i'm fairly new at this.

[@AaronAverett]

A blown fuse, almost without exception, means you have a short to ground. Confirm this by checking resistance between ground and the side of the fuses farthest from the power connector. After that, you need to figure out where the short is. The first step is to figure which part of the board is connected to your blown fuse. On this card, that's the bottom half, when looking at it face up on a desk. You'll need to check for continuity to the fuse to tell for sure which components are connected to it. Anything that is is a potential source of a short to ground. Second step is a visual inspection. Look for anything burned, discolored or otherwise visibly damaged. Third step is voltage injection. Supply 1.0V to the fuse and see what heats up on the board. This doesn't usually work on 12V shorts because the circuit is meant to handle much higher voltage, but it's worth a shot. Fourth step is trial and error - start removing suspect parts and check for your short. When the short goes away, you know that the last thing you removed was your culprit. Confirm by testing the removed component for a short.

[@chunny1980]

@@AaronAverett Thank-you so much for the reply, I'll keep looking into it. I bought this card broken last week from someone on Facebook in the hope of being able to fix and use in my pc for the time being. I'm noticing that the fuse is connected to the bottom 3 mosfets and I'm getting a reading of 2.9 ohms on each one if I'm doing it correctly?, the caps in that area read the same as well on one side. Do you think there's a chance the gpu itself could be dead and is there a way to check this before I go any further taking bits off the board to test?. I'm trying to learn as I go but I'm realising I'm a bit out of my depth here 😆

[@AaronAverett]

​@@chunny1980 Here's a hint: That 2.9 ohms is roughly the resistance you'd expect to see through the GPU core. Take a look at a schematic for a buck converter and see if you can think of a way that that circuit could fail that would create this situation. Second hint: It's different from the problem I had on my card.

[@Tony2a]

can you link me the capacitor?

[@rivanotimmerman3590]

hi i need help i think my fuse i blown nr: nf 1202 but i dont know where to get a new one do you know where i can get a new one? also i that fuse is blown how can i find the cause of it/where can i find it? edit: i think i got the parts i need NCV78M05ABDTRKG 2410SFV8.00FM/125-2 IR3553MTRPBF are these useble?

[@AaronAverett]

2410SFV8.00FM/125 Should be the fuse you need. It comes in different ratings, so you'll need to be sure you get the right one. Match up the marking to the one the datasheet lists, and order that specific one. Regarding the underlying cause, you'll need to use the methods I've discussed here and elsewhere to find the short. Your short is on a different circuit than mine was, so I would not assume that the problem is the same as mine. You'll need to figure out what is actually connected to your blown fuse, and troubleshoot each circuit until you find one that's shorted. Do you have access to a hot air station? if so, are you comfortable using it?

[@rivanotimmerman3590]

@@AaronAverett thx and no i dont have a hot air only a soldering iron, also where can i find the datasheet list because i think its nf 1207 but I'm not shure because its need 5v on one side and 3 on the other but its not its the same on both sides right?

[@AaronAverett]

You really need a hot air station to do this successfully. Soldering irons are great for doing one connection at a time, but most of the components on a graphics card have at least two, and you tend to damage the PCB when you try to desolder components with them. I've had a number of people ask me to fix their mistakes after trying something like this with just a soldering iron, and it's usually not pretty. Is there any chance you can post a photo or video of the component you're suspecting is the problem? A fuse should have a very small voltage drop across it - so small that you wouldn't notice with a handheld multimeter. If you have a component that has 5V on one side and 3V on the other, it's probably not a fuse.

[@rivanotimmerman3590]

@@AaronAverett hi its been a while and i am coming back to this and indeed the fuse is blown i have orderd new ones but if i check with my multimeter the resistans is around 75 ohms is the same place as here 12:50 in the video i dont thinks thats good. or it is bc there is resistants or is its just me being dump haha the rest looks normal any idea of this is good ima test the card if i have the fuse instaled but i wanna know the cause of it

[@MTS_IT]

lol... 1 louis rossman :D

[@m7ki390]

I got the same graphics card from my friend, he said that it would still work. I installed it on my pc and i get the beep code that the graphics card is broken.

[@AaronAverett]

Man, that's a bummer. If you get as far as a POST beep code, then whatever is wrong with your card is probably different from the card in the video. You'd get an immediate shutdown if you had a short to ground on the 12V rails like I had on this card. If you're interested in trying to repair it, the next step is to take the backplate off and test each of the power rails at the choke outputs for operation with a multimeter. You should have 0.9-1.0 on the core, 1.5 on the memory and ~1.0V on the memory controller rail (the one on the left side, near the display connectors). This won't tell you exactly what's wrong, but it will help narrow it down. Edit: Actually, I forgot about the fuses. It's actually possible that the problem is something similar. If your buddy had that card kicking around his house for a while, it's totally possible that it suffered some mechanical damage to the back that caused a short like my card hard. Try doing the same resistance checks that I did in the video, and let me know what you find. :)

[@cyranoburleson4240]

@@AaronAverett I have the same card with no display output and it doesn't show in bios. But the blue light for the second bios does turn one. NO idea were to start with this.

[@AaronAverett]

​If I made another video, where I show the steps I would take if I had your card in my hands, would that help you?

[@m7ki390]

@@AaronAverett the problem is solved recently bought a new PC

[@m7ki390]

@@AaronAverett but still thx for trying to help

[@semirodobasic5542]

hey is this card better than rx 570 xfx plese tell me

[@AaronAverett]

I have not done any testing, but I would expect the 390 to be slightly faster than an RX570. The RX570 is a better card, though. It's newer, smaller and way more efficient. If you have a choice between the two, get the RX570.

[@nikolakarovic5964]

Hawai cards are from hell.

[@AaronAverett]

You're not wrong, but apparently, Hell isn't ready to take this one back.