Sapphire HD 7870 possible short

[mikkhh]

Re: Sapphire HD 7870 possible short

I tested everything else on the card and if somebody wants a picture about the shorted mosfet, well there it is. But the gpu memory mosfet itself could be just fine at the worst case of scenario. The short could be caused by GPU memory itself (atleast half the rams are dead) or the memory controller within the GPU die. I will solder the mosfet off and report back then. Let´s hope for the best. BTW that shorted mosfet is very very common AMD cards and most likely it is the faultty memory controller within the gpu die .

[AMD_Ripper]

Re: Sapphire HD 7870 possible short

I didn't read all but did you fix the 7870 ?

[AMD_Ripper]

Re: Sapphire HD 7870 possible short

What's the sense is signing up to a forum if you can't post anything.

[CroTech]

Re: Sapphire HD 7870 possible short

Hi guys,

So my XFX R9 380 was working fine but I wanted to put it to my cheaper/weaker PC since I got upgraded, everything was working fine, but I think my generic PSU killed it somehow.

I installed the GPU and it turned on ok, started playing a game and suddenly black screen.

I restarted PC but the GPU seems to be as if not pluged in.

Took the heatsink if and saw the connector terminals are shorted, spotted these 2 MOSFETs shorted and replaced them with indentical, nothing changed.

There is a short somewhere but I dont know where.

[momaka]

Re: Sapphire HD 7870 possible short

I installed the GPU and it turned on ok, started playing a game and suddenly black screen.

I restarted PC but the GPU seems to be as if not pluged in.

Took the heatsink if and saw the connector terminals are shorted, spotted these 2 MOSFETs shorted and replaced them with indentical, nothing changed.

There is a short somewhere but I dont know where.

So with the video card in the computer and the PCI-E power connectors plugged in, can you turn On the computer without the PSU shutting down instantly? If yes, then you might not have a short-circuit. A short-circuit on the GPU V_core or RAM Vdd usually pulls down the whole PSU and causes it to shut down. If that's not the case, you probably have a GPU BGA/bumpgate problem.

[CroTech]

Re: Sapphire HD 7870 possible short

With the video card plugged in the PCIe, and both connector from the PSU to the Video card the PC turns on, works well on the integrated GPU and it boots normaly even thoguh Ohm meter shows 1 Ohm on the marked terminals.

The GPU did not have any problems so far with the BGA and has not been "oven repaired".

Also I would like to add that while the PC is on, the heatsink is cold and the GPU is not developing any heat.

So I would say that the GPU chip is not shorted, however I do not understand how can the PC boot and work if there is a short anywhere in the PC, in my caste the video card connector is shorted (or is it)?

[momaka]

Re: Sapphire HD 7870 possible short

With the video card plugged in the PCIe, and both connector from the PSU to the Video card the PC turns on, works well on the integrated GPU and it boots normaly even thoguh Ohm meter shows 1 Ohm on the marked terminals.
...
however I do not understand how can the PC boot and work if there is a short anywhere in the PC, in my caste the video card connector is shorted (or is it)?

If you had a shorted MOSFET like you say, then the PC would NOT turn On. Reason being is the GPU (and sometimes also the RAM) has a very low resistance - typically less than 10 Ohms (usually around 1-3 Ohms on recent GPUs). Thus, a shorted upper MOSFET will cause the GPU chip to try to draw an extreme amount of current from the PSU. This will almost always trigger the PSU's short-circuit protection (the few cases where that didn't happen, I saw the GPU shoot up smoke like a Diesel truck, but only white in color and with a TERRIBLE smell ).

So check that you are measuring the pins in the PCI-E power connector properly. On a 6-pin and 8-pin PCI-E power connector, the row of 3 (or 4) pins on the side of the lock/latch are ground. The 3 (or 4) opposite of that are 12V. Have a look on Google for a diagram.

Also, note that on an EPS12V connector (those 8-pin CPU connectors), the ground and 12V pins are completely the opposite to that of the PCI-E power connector: the row of 4 pins closer to the connector lock/latch are 12V and the other side is ground. So keep that in mind while measuring.

Finally, make sure you keep your multimeter probes connected long enough to measure the resistance. Due to capacitors in the circuit, your multimeter might show a low resistance of a few Ohms for the first 1-2 seconds. But if you leave the probes connected long enough for the caps to charge, the resistance should rise (typically it will be more than 1 KOhm with the video card removed from the motherboard and disconnected from the PSU). And last but not least: make sure you are not misinterpreting your multimeter (especially if it's a manual-range one) - a "1" on the left side of the screen does NOT mean the resistance is "1 Ohm". It means "over range of the current scale". I've seen people make this mistake, so I am hoping that's not what's going on here.

That said, just because the video card didn't have BGA/bumpgate issues up to this point and was never oven-baked doesn't mean that it couldn't have just developed them. Given the symptoms you describe and the fact that your computer can run without the video card shorting out the PSU strongly suggests BGA/bumpgate issues.

[CroTech]

Re: Sapphire HD 7870 possible short

Momaka, you were right, after leaving it for a 2-3 min the caps charged and the measurements are different, here are the photos, the same thing applies to the second connector.

So I am looking for a blown fuse/component that is connected to the second pin on each power connector? Since the resistance is infinity on those.

I did not upload the second connector photos since they are exactly the same.

[CroTech]

Re: Sapphire HD 7870 possible short

SO for the fuses, the only 2 I found are those by the HDMI connector and they are probably there for protecting the HDMI device, they checked out fine on the multimeter, in diode mode it beeps, in resistance mode they are short.

[momaka]

Re: Sapphire HD 7870 possible short

Momaka, you were right, after leaving it for a 2-3 min the caps charged and the measurements are different, here are the photos, the same thing applies to the second connector.

Yup, that's a pretty high resistance. That means you *don't* have shorted MOSFETs.

So I am looking for a blown fuse/component that is connected to the second pin on each power connector? Since the resistance is infinity on those.

The middle 12V pin (and sometimes also middle ground on certain video cards) is used to tell the video card that the 6/8 pin PCI-E power connector is connected. That pin (or those pins) goes to a different circuit - hence the different (infinite) resistance. This is normal.

SO for the fuses, the only 2 I found are those by the HDMI connector and they are probably there for protecting the HDMI device, they checked out fine on the multimeter, in diode mode it beeps, in resistance mode they are short.

Yes, those are most likely to protect the HDMI device and also the video card, in case the HDMI cable or connector gets damaged.

That said, have you tried testing the video card on a different output connector, like the DVI-D or DisplayPort (if the card has one)?

[CroTech]

Re: Sapphire HD 7870 possible short

Yeah. Tried both DVIs and the HDMI. Dont have a display port device so... But the card is not shown in the device manager. Are there any more fuses on the card I should check? Thanks for the reply Momaka.

[momaka]

Re: Sapphire HD 7870 possible short

It's not going to be a fuse issue. Check the GPU V_core and memory voltage outputs with the card/PC turned On. If they are normal, you probably have a BGA/substrate issue. These cards fail all the time like this. A reflow is probably in order (but I *don't* recommend an oven reflow, though).

[CroTech]

Re: Sapphire HD 7870 possible short

I have to check where should I measure those voltages since I havent done this before but I will have it a go. I really doubt it is a BGA problem since the card worked flawlesly in my main rig but I upgraded to gtx 1070. My main fault was putting a molex converter to 6 pin since the old PSU had only 1 6pin power connector and even on first boot GPU was working fine but when I streesed the gpu with a game then it died.

[CroTech]

Re: Sapphire HD 7870 possible short

Back with more info.

Found out that you can measure Vcore on the middle SMD capacitor on the back of the gpu, while PC was on, and everything connected I got 0.02V so that is saying the core is not getting the voltage.

Also started looking the VRMs, I have 4983NF

Tried this method

Diode mode beeps on DRAIN TO SOURCE no matter which probe is on which, also tried discharging the gate to ground, also beeps, and that is on every MOSFET

[CroTech]

Re: Sapphire HD 7870 possible short

UPDATE:

Took one IC off a board and tested it and it checks out fine.

Checked for shorts on the PCB where the pads where, YUP all three SOURCE short to DRAIN.

Sorry for low quality photo.

Could it be that if one of the VRMs is making a short between SOURCE and DRAIN that they all get shorted?
I wonder what would happen if I took them all down.

[CroTech]

Re: Sapphire HD 7870 possible short

Aaaaaand removed all of them, SOURCE AND DRAIN still shorted...

[Th3_uN1Qu3]

Re: Sapphire HD 7870 possible short

The GPU chip died when the mosfets shorted. Sorry.

[CroTech]

Re: Sapphire HD 7870 possible short

Yeah thought so,
when watched this video