Sapphire HD 7870 possible short

[Anzezaf]

Re: Sapphire HD 7870 possible short

Hey guys, i'm back. Finally got the hot air gun and went to work. Removed all the tantalum caps which were all ok - showing open circuits. I then removed both the RAM and GPU lower fets and they were also fine, every single one of them.
Next, I measured the following resistances:
-GPU lower fet's source to GND ~0.6ohm
-GPU lower fet's drain to GND ~0.9ohm
-RAM lower fet's source to GND ~0.6ohm
-RAM lower fet's drain to GND ~1.5ohm

It really bummed me out when I saw every fet is OK . Is there anything else worth checking?

[momaka]

Re: Sapphire HD 7870 possible short

Those resistances suggest shorted GPU chip to me. Not sure what else could be causing them. Perhaps the VRM controllers. I guess you could remove them to see if that really is the problem. But at this point, it's starting to look pretty complicated.

[Anzezaf]

Re: Sapphire HD 7870 possible short

Nope, removed the buck controllers, and resistances are the same.
Looks like i'm gonna have to chuck in the trash, don't really know what else could be causing the problem? I think we pretty much isolated the whole power section of the card, and the short is still there.

EDIT: Say, momaka, are you willing to have a look at the 2nd card? Looks like the first one is a goner, but there might still be a chance for this one. Comparing to the first one, I think this one is a bit more promising - some higher resistances measured. It's a r9 290, the one that suffered the short only once - I removed the external connectors after the first pop.

The strangest thing happened when i first started measuring resistances - Pulled the card out of the box and the first thing i measured was between GND and 12V, and it showed a short circuit. Some time after fiddling around with the FETs, i tried measuring the 12v and GND short again, and it wasnt there - the instrument showed a solid open circuit - no ohm reading. I tried grounding the card, around the FETs, all the caps near them etc. but the short was still gone. Can't really explain this to myself

Anyhoo, heres the findings:
GPU FETS:
-Lower (biggers ones - IRF6894)FETs DRAIN to GND resistances are sub 1ohm
-Upper (smaller ones - IRF6811)FETs DRAIN to 12V resistances are sub 1ohm
BUT, there are exceptions to this:
-ONE lower fet, circled red on the picture, reports a good 200ohms between its drain and gnd
-TWO upper fets, circled blue, show a high resistance between its drain and 12v - 7k and 8k ohm

RAM FETS
-Lower (biggers ones - IRF6894)FETs DRAIN to GND resistances are a healthy 4ohm
-Upper (smaller one - IRF6811)FET DRAIN to 12V resistance is sub 1ohm (leads to pcie slot 12v rail)

Do I go ahead and try running in a PC and measuring voltages or..? I forgot if this card lets the PC boot up or not. :\

[momaka]

Re: Sapphire HD 7870 possible short

Looks like i'm gonna have to chuck in the trash, don't really know what else could be causing the problem? I think we pretty much isolated the whole power section of the card, and the short is still there.

Well, don't throw it in the trash. Just keep it for parts - especially if you ever plant to repair more stuff in the future. I've had many junk boards help me save good hardware. Probably over 20 things.
If nothing else, at least save the polymer capacitors.
(And also check them for short-circuit, while you're at it. I think those are the only parts that you haven't pulled away to check for short-circuit, yes?)

EDIT: Say, momaka, are you willing to have a look at the 2nd card?

Of course!

The strangest thing happened when i first started measuring resistances - Pulled the card out of the box and the first thing i measured was between GND and 12V, and it showed a short circuit. Some time after fiddling around with the FETs, i tried measuring the 12v and GND short again, and it wasnt there - the instrument showed a solid open circuit - no ohm reading. I tried grounding the card, around the FETs, all the caps near them etc. but the short was still gone. Can't really explain this to myself

Was the short-circuit there for more than several seconds? If not, then it could have been the caps charging and making the multimeter beep before going open-circuit.

Anyhoo, heres the findings:
GPU FETS:
-Lower (biggers ones - IRF6894)FETs DRAIN to GND resistances are sub 1ohm
-Upper (smaller ones - IRF6811)FETs DRAIN to 12V resistances are sub 1ohm

The sub 1-Ohm readings on the lower FETs is a bit worrisome again. But let's hope that's normal static resistance for the GPU

The upper FETs, you test between Drain and Source. Drain is connected to 12V rail, and Source is connected to GPU V_core and also the Drain of the lower FETs.

BUT, there are exceptions to this:
-ONE lower fet, circled red on the picture, reports a good 200ohms between its drain and gnd

The more at look at that picture with the circled FETs, the more I am starting to think there are several different GPU and RAM rails.

The "lower" FET you circled in red may be another rail related to RAM or GPU communication with the RAM - this I don't know for sure. It's a bit hard to see the traces on the board due to it being black, but I think I can see a different trace for that lower FET and the "upper" FET above it. You can see that trace going near the four ceramic capacitors on the left: C716, C717, C719, C720.

-TWO upper fets, circled blue, show a high resistance between its drain and 12v - 7k and 8k ohm

I now think those upper FETs (Q511, Q510) could also be part of a different rail, and perhaps derived from the GPU V_core rail. See if these "upper" FETs you circled in blue show continuity/low resistance from their Drain to the Drail of the lower MOSFETs above them (Q502, Q503, Q504)

RAM FETS
-Lower (biggers ones - IRF6894)FETs DRAIN to GND resistances are a healthy 4ohm
-Upper (smaller one - IRF6811)FET DRAIN to 12V resistance is sub 1ohm (leads to pcie slot 12v rail)

Again, that low 4 Ohm resistance on the lower FETs is a bit worrysome. I have my fingers crossed.

Do I go ahead and try running in a PC and measuring voltages or..?

Yes, I think that would be the best thing to do.

But before you do that, I think we need to break down what all of the voltage rails are, as it seems there is more on this card than "GPU" and "RAM" rail FETs. So I will do that by listing what *I think* are some possible rails. You will need to confirm this with your multimeter resistance, though.

---- RAM rail 1
* Q513 (upper FET), Drain tied to 12V rail
* Q512, Q500 (lower FETs), Drain tied to Q513 Source
* L500 ("R15" inductor), both ends should appear tied to Q513 Source and Q512/Q500 Drains.
* C616A, C617, C616, and C617A (polymer capacitors), filter for rail output. The (+) leads on these caps should be tied to Q513 Source and Q512/Q500 Drains

** check the validity of my statement above by measuring resistance.
** to measure the voltage on this rail, measure at one of the metal ends of inductor L500 with respect to ground.

---- RAM rail 2
* Q702 (upper FET), Drain tied to 12V rail
* Q703 (lower FET), Drain tied to Q702 Source
* first inductor (L501?) at top with "R23" middle line: both ends of this inductor should appear tied to Q702 Source and Q703 Drain.
* two polymer capacitors to the left of this inductor: filter for rail output. The (+) leads on these caps should be tied to Q702 Source and Q703 Drain

** check the validity of my statement above by measuring resistance.
** to measure the voltage on this rail, measure at one of the metal ends of the "R23" inductor with respect to ground.

---- GPU V_core rail
* Q505, Q506, Q507 (upper FETs): Drain tied to 12V rail
* Q502, Q503, Q504 (lower FETs), Drain tied to Q505/Q506/Q507 Source
* L502(?), L503(?), L504 ("R15" inductor), both ends should appear tied to Source on upper FETs and Drain on lower FETs.
* C504, C511, C516, C520 (polymer capacitors): filter for GPU V_core rail output. The (+) leads on these caps should be tied to Source of upper FETs and Drain on lower FETs

** check the validity of my statement above by measuring resistance.
** to measure the voltage on this rail, measure at one of the metal ends of inductors L502/L503/L504 with respect to ground.

---- GPU rail 2
* Q510, Q511 (upper FETs): Drain tied to ???
* Q508, Q509 (lower FETs): Source tied to ground, Drain tied to Q510/Q511 Sources
* L505(?), L506(?) ("R15" inductor): both ends should appear tied to Source on upper FETs and Drain on lower FETs.
* C505, C507, C512 (polymer capacitors): filter for GPU V_core rail output. The (+) leads on these caps should be tied to Source of upper FETs and Drain on lower FETs

** check the validity of my statement above by measuring resistance.
** to measure the voltage on this rail, measure at one of the metal ends of inductors L505/L506 with respect to ground.

Sorry if these are inaccurate. I can't pinpoint these rails exactly, because I have not yet had a chance to work on new hardware like this. My newest cards are a GeForce 9800GT and HD3870 .

[Anzezaf]

Re: Sapphire HD 7870 possible short

If nothing else, at least save the polymer capacitors.
(And also check them for short-circuit, while you're at it. I think those are the only parts that you haven't pulled away to check for short-circuit, yes?)

Will remove them and report back.

Was the short-circuit there for more than several seconds? If not, then it could have been the caps charging and making the multimeter beep before going open-circuit.

I didn't measure it for more than a few seconds, so it could be some caps filling up, but i checked again this morning (surely any caps would empty in ~12 hours)and the multimeter again showed an open circuit.. A mystery indeed

The "lower" FET you circled in red may be another rail related to RAM or GPU communication with the RAM - this I don't know for sure. It's a bit hard to see the traces on the board due to it being black, but I think I can see a different trace for that lower FET and the "upper" FET above it. You can see that trace going near the four ceramic capacitors on the left: C716, C717, C719, C720.

Here's the pic of the area mentioned.

Notice there is only one differential trace to be seen,and it leads to/under a RAM IC/chip. Everything else is routed through vias on the three big pads. It might just be power delivery for the RAM, due to the different resistances and also the different choke used.
But again, this is coming from a noob..

I now think those upper FETs (Q511, Q510) could also be part of a different rail, and perhaps derived from the GPU V_core rail. See if these "upper" FETs you circled in blue show continuity/low resistance from their Drain to the Drail of the lower MOSFETs above them (Q502, Q503, Q504)

Nope, they do not show continuity to the DRAIN of Q502/503/504.

---- RAM rail 1
* Q513 (upper FET), Drain tied to 12V rail
* Q512, Q500 (lower FETs), Drain tied to Q513 Source
* L500 ("R15" inductor), both ends should appear tied to Q513 Source and Q512/Q500 Drains.
* C616A, C617, C616, and C617A (polymer capacitors), filter for rail output. The (+) leads on these caps should be tied to Q513 Source and Q512/Q500 Drains

-q513 drain is tied to 12v rain on the PCIe slot
-I cannot check FET source connections without removing them first, and quite frankly I wouldn't want to start removing them so soon into the analysis . They're quite a hassle to mount again, but of course, I will do it if it's needed .
-L500 is indeed connected to Q512/500 drains.
-And again correct, the caps mentioned are tied to Q512/500 drains

---- RAM rail 2
* Q702 (upper FET), Drain tied to 12V rail
* Q703 (lower FET), Drain tied to Q702 Source
* first inductor (L501?) at top with "R23" middle line: both ends of this inductor should appear tied to Q702 Source and Q703 Drain.
* two polymer capacitors to the left of this inductor: filter for rail output. The (+) leads on these caps should be tied to Q702 Source and Q703 Drain

-Yes, Q702 is tied to 12v rail on external power connectors
-Again, cannot say, but give a go-ahead and i will remove them to check source.
-Yes, this R23 "L700" inductor is connected to q703 drain.
-Correct on the two polymer caps.

---- GPU V_core rail
* Q505, Q506, Q507 (upper FETs): Drain tied to 12V rail
* Q502, Q503, Q504 (lower FETs), Drain tied to Q505/Q506/Q507 Source
* L502(?), L503(?), L504 ("R15" inductor), both ends should appear tied to Source on upper FETs and Drain on lower FETs.
* C504, C511, C516, C520 (polymer capacitors): filter for GPU V_core rail output. The (+) leads on these caps should be tied to Source of upper FETs and Drain on lower FETs

-Correct on everything, although i cant check the upper fets source.

---- GPU rail 2
* Q510, Q511 (upper FETs): Drain tied to ???
* Q508, Q509 (lower FETs): Source tied to ground, Drain tied to Q510/Q511 Sources
* L505(?), L506(?) ("R15" inductor): both ends should appear tied to Source on upper FETs and Drain on lower FETs.
* C505, C507, C512 (polymer capacitors): filter for GPU V_core rail output. The (+) leads on these caps should be tied to Source of upper FETs and Drain on lower FETs

-Q510,511 drain tied to 2nd external power connector
-Again, cant say
-L505,506 tied to lower fets drain, except for the top most FET.
-Same with C505,507,512.

Sorry if these are inaccurate. I can't pinpoint these rails exactly, because I have not yet had a chance to work on new hardware like this. My newest cards are a GeForce 9800GT and HD3870 .

HD3870 is a golden card, trust me. Mine is pushing 9 years old now. Used it for gaming from 2007 until last year when GTA5 came out, and when it got replaced i saved it for any emergencies or comp builds, and it saved my ass again when I shorted these two GPUs. Oh did I mention it broke 3 or 4 times and I saved it every time by baking it in an oven?

[ChaosLegionnaire]

Re: Sapphire HD 7870 possible short

not sure if my info is accurate or reliable or not but all modern gpus should have at least 3 or 4 voltage rails. the gpu core and vram voltage are the known common ones. but the third one is called the vddq reference voltage (used for signalling through the pci-express bus or powering the vrm controller ic, not sure) and then there is also the vtt or termination voltage for the gddr5 vram. there could be more rails or other rails but i think this covers it for now.

[Anzezaf]

Re: Sapphire HD 7870 possible short

Right, I now checked all polymer caps on the first GPU, and the only ones showing a short were those 4 blue lined ones near the RAM chokes. Removed them, but they all turned up normal.

[momaka]

Re: Sapphire HD 7870 possible short

Here's the pic of the area mentioned.

Notice there is only one differential trace to be seen,and it leads to/under a RAM IC/chip. Everything else is routed through vias on the three big pads. It might just be power delivery for the RAM, due to the different resistances and also the different choke used.
But again, this is coming from a noob..

Nope, note a noob anymore for sure .
Yes, the different choke does indicate a different rail.

-I cannot check FET source connections without removing them first, and quite frankly I wouldn't want to start removing them so soon into the analysis .

No problem.

In that case, just try to take the voltage measurements I mentioned in my previous post above (i.e. the voltages at the inductors for the 4 different voltage rails).

Make sure the heatsink IS on the video card and that there is thermal compound. If you can't take the voltage measurements because the heatsink is in the way, then solder a piece of wire to the inductors you want to measure, and measure from the wire when the card is running (with respect to ground, of course.)

HD3870 is a golden card, trust me. Mine is pushing 9 years old now. Used it for gaming from 2007 until last year when GTA5 came out, and when it got replaced i saved it for any emergencies or comp builds, and it saved my ass again when I shorted these two GPUs. Oh did I mention it broke 3 or 4 times and I saved it every time by baking it in an oven?

That's pretty impressive!
I actually have 2 of these cards now. I got them from eBay for very cheap - less than $15 each. And in general, I've always had better luck with reviving ATI stuff - at least when it comes to reflowing. I'm surprised the caps didn't cook on yours when you baked it in the oven. Or maybe they did, but the VRMs just have enough ceramic caps to work even with the bad electrolytic caps? IDK.

Right, I now checked all polymer caps on the first GPU, and the only ones showing a short were those 4 blue lined ones near the RAM chokes. Removed them, but they all turned up normal.

That means they are probably good. Just save them in case you ever need them for another video card. A lot of people know about the oven "trick" and try to bake their card, and a lot of times that ruins the caps too. So if you ever get your hands on a card like that, you can replace the caps with these spares .

not sure if my info is accurate or reliable or not but all modern gpus should have at least 3 or 4 voltage rails. the gpu core and vram voltage are the known common ones. but the third one is called the vddq reference voltage (used for signalling through the pci-express bus or powering the vrm controller ic, not sure) and then there is also the vtt or termination voltage for the gddr5 vram. there could be more rails or other rails but i think this covers it for now.

Well, for desktop RAM, there are always two rails: Vdd/Vddq and the termination voltage Vtt. On graphics cards with up to G/DDR3, though, I've only seen Vdd/Vddq. But maybe my information is outdated or this is changed now with G/DDR5.
Then there is the main GPU V_core, of course. And finally, like you mentioned, there is another rail for PCI-E bus communication - or at least that's what I think it is for. Usually around 1.2-1.25V, IIRC.
So I don't know what the 4th rail could be. Perhaps for the shader core on the GPU? Or some other part?

I think I need to come out of the stone age and get a contemporary broken GPU to play with so I can figure these . But even broken GPUs still cost a lot .

[Anzezaf]

Re: Sapphire HD 7870 possible short

Eh, I measured at the points you wanted, but there isnt any voltage on any of them, all 0V. May be due to the fact that i ran it on a only 420W PSU (and as a result, the PSU didnt supply the GPU because its too weak) but i doubt it - consumption while not under load should be small. Also, the fan of the cooler didnt spin up, but there is an LED on the board thats shining.

Goddamn, was I pissed when trying to solder anything on the mentioned spots, couldnt get a nice solid joint no matter how I tried. I think I must have spent at least an hour for those 4 wires, and in the end I didn't even try to make a nice joint, did everything with semi-cold solder (It might also be the cause for 0V everywhere - not good enough contacts, but I highlydoubt it.). Never cursed so much while soldering .

[momaka]

Re: Sapphire HD 7870 possible short

Eh, I measured at the points you wanted, but there isnt any voltage on any of them, all 0V. May be due to the fact that i ran it on a only 420W PSU (and as a result, the PSU didnt supply the GPU because its too weak) but i doubt it

The PSU power rating wouldn't make the card not boot. If the PSU isn't adequate, it will either shut down or explode (typically that happens to low-quality PSUs). So I doubt that was the problem as well.

Also, the fan of the cooler didnt spin up, but there is an LED on the board thats shining.

Interesting.

I don't see any fuses near the PCI-E 16x power pins, so everything should be getting power in theory.

That might suggest several issues:
-- VRM controller IC(s) or MOSFET gate drivers damaged
-- GPU and/or RAM shorted, thus causing the the VRM controller IC(s) not to start
-- flash ROM dead

Is there anything on the card or in the manual that says what the LED status indicates?

Goddamn, was I pissed when trying to solder anything on the mentioned spots, couldnt get a nice solid joint no matter how I tried. I think I must have spent at least an hour for those 4 wires, and in the end I didn't even try to make a nice joint, did everything with semi-cold solder (It might also be the cause for 0V everywhere - not good enough contacts, but I highlydoubt it.). Never cursed so much while soldering .

Yes, high-end video cards tend to have big copper pads that really sink heat from your iron and make it very hard to solder to. When working on these, your iron needs to be rated for 60 Watts minimum - and that's if it has a really thick and heavy copper tip to hold enough energy (heat) to solder. Even an 80 Watt iron could struggle if it doesn't have a decent enough tip.

That said, it often helps to pre-heat the board before trying to work on it. Even the mildly warm air from a hair dryer could be helpful in some cases.

[Anzezaf]

Re: Sapphire HD 7870 possible short

Hey, sorry for the late reply. Did some googling and some guys are saying that on a MSI r9 290x, the LED shows the "powerstate?" and/or something called "ULPS".

Another guy from that same thread is saying

"I have 2 290's from MSI and saw the same green light. The light comes on because of AMD's "ZeroCore" technology. When the GPU isnt being used/outputting to a display it effectively turns off, the fan stops etc. The light must be to show that the GPU is still actually on and working but powered down to save energy."

source: http://www.overclock.net/t/1439978/msi-r9-290x-led

[momaka]

Re: Sapphire HD 7870 possible short

I guess that is more evidence indicating a fault with the GPU.

Can you boot your PC with another (working) video card and have this video card in the second PCI-E slot? I'm curious to know if the operating system can even see the video card. If no, then it's quite possible either the GPU or flash ROM could be dead.

[adymang]

Re: Sapphire HD 7870 possible short

Hy guys. I have the same card 7870 2G, and a shortcut on the chard.
1. Starting the pc the power source is beeping (shortcut detected) and shuts down.
2. measurements: GND - MAGIC R19 = 0,3 ohm
12V- GND = 600 ohm
1R0 - GND = 210 ohm

Can anyone give me a starting hint to detect the problem.
Thank you.

[adymang]

Re: Sapphire HD 7870 possible short

Also I have to mention that if I plug just one power connector the fan is spinning. If I plug the second power connector the shortcut is present. In conclusion the shortcut is on the second rail (the exterior power connector). Maybe someone can explain what is powering this circuit and haw to identify the defect. Thanx

[Anzezaf]

Re: Sapphire HD 7870 possible short

I guess that is more evidence indicating a fault with the GPU.

Can you boot your PC with another (working) video card and have this video card in the second PCI-E slot? I'm curious to know if the operating system can even see the video card. If no, then it's quite possible either the GPU or flash ROM could be dead.

TBH I'm quite fed up with these GPUs, plus i dont have such a mobo. If any other kind of possible solution crosses your mind, please say so - if not, i will just sell it and see how much i can get.

[momaka]

Re: Sapphire HD 7870 possible short

Also I have to mention that if I plug just one power connector the fan is spinning. If I plug the second power connector the shortcut is present. In conclusion the shortcut is on the second rail (the exterior power connector). Maybe someone can explain what is powering this circuit and haw to identify the defect. Thanx

I don't have that video card, so you will need to give me more details and I will try to work with the pictures that modman1337 posted.

Basically, which is the "second" exterior power connector? If you look on the PCB, you will see markings for each component. One of these power connectors has J1000 printed next to it, and the other is J1001.

Next, we look for which MOSFETs are connected to the power connector that has the short-circuit (the "shortcut", as you call it, so we are on the same page ).
To do that, first take your multimeter and set to lowest resistance scale (usually 200 Ohms on manual-range multimeters). Put the red (+) multimeter probe on a pin on the 12V rail of the connector that has a short-circuit to ground. Then take the black (-) multimeter probe, and put it on every Drain pin of every MOSFET on the board. See for which MOSFETs you get a reading of less than 5 Ohms. Note their board numbers and post them here (i.e. is it Q703, Q804, Q631.... etc.) The ones that show it *may* be connected to the shorted 12V rail. We will determine that in another round of testing, though, when you post the results.

TBH I'm quite fed up with these GPUs, plus i dont have such a mobo. If any other kind of possible solution crosses your mind, please say so - if not, i will just sell it and see how much i can get.

Yes, sometimes these things get to a point where you just need to put too much effort into them and it may not be worthwhile to you. So selling them is not a bad idea. Someone who does this type of repair for a living and has all the tools may actually find useful components on these cards (or even fix them). But I am afraid that is as far as my knowledge stretches. Sorry we couldn't get these fixed. :\

[Anzezaf]

Re: Sapphire HD 7870 possible short

Yes, sometimes these things get to a point where you just need to put too much effort into them and it may not be worthwhile to you. So selling them is not a bad idea. Someone who does this type of repair for a living and has all the tools may actually find useful components on these cards (or even fix them). But I am afraid that is as far as my knowledge stretches. Sorry we couldn't get these fixed. :\

It's just so dissappointing, we tried a lot and got nowhere. Hitting a brick wall every step of the way..

In any case, I'm really grateful that you took the time to help out. Thanks a bunch mate

[SpiderWaffle]

Re: Sapphire HD 7870 possible short

I've had a problem with my 7870 too, maybe a bit different, but caps are notoriously a weak link in electronic failure. Mine is a PowerColor brand, it worked fine for 2 years, then when I my ASUS LCD started to tank, it was causing my new BenQ LCD to flicker and the video card to humm. I since repaired the very old ASUS LCD it's working fine now. But the the 7870 card doesn't seem to be able to handle the new 144hz BenQ too well, used to only flicker when I plugged in both LCDs, but now the BenQ alone is flickering by itself. I plugged the BenQ into a different PC and the BenQ works fine.

1. So I'm thinking I might have some bulging caps on the 7870, does that sound like a highly likely problem I could fix by swapping them all out?

2. Also, these are contact terminals right? Aren't those really hard to solder with a normal solder gun? (I was able to use my gun on the 28ish through-hole caps on the ASUS alright, not easy getting some of the little ones out though)

[momaka]

Re: Sapphire HD 7870 possible short

1. So I'm thinking I might have some bulging caps on the 7870, does that sound like a highly likely problem I could fix by swapping them all out?

Post a picture. I don't think there are any modern-day video cards that use bad capacitors anymore. So unlikely that this is your problem. But as they say, A picture is worth a 1000 words.

[mikkhh]

Re: Sapphire HD 7870 possible short

Again, speaking a bit off-topic. But what about GPU memory phase mosfet shorts on AMD cards? The card displays only black image, the fans are spinning etc. The lower (speaking from memory, not 100% sure, but it is the larger directfet mosfet) mosfet is shorted to the ground. I am hoping to remove it and if I remove it and the short goes away from the pcb then the GPU and memory are OK, hopefully. Can it be fixed by replacing the mosfet? Or the mosfet itself is good, but the GPU memory/ memory channels within the GPU die are faulty?