Two faulty Radeon 5970 cards

[oleq83]

Hello.

I've been searching a bit on the net and there seemed to be a lot of people with with good technical insight on this forum, so I was wondering if anyone could help me with two faulty graphic cards that I purchased recently.

So. I recently purchased two used Radeon 5970 cards for cheap.
I was told by the previous owner that the cards had stopped working after his PSU had all of a sudden started sending out 16v on the 12v lane, causing his motherboard and both the graphic cards to stop working.

He suggested that the problem was propably due to broken Volt Regulators on the cards and it shouldn't be too difficult to replace. I have a hot air station and have replaced mosfets and a few smds earlier, so I thought it wouldn't be too hard to replace the Vrm chips.

When I received the cards I first tested the cards in my computer, to see what happened.
The first card got no signal on the screen and the fan was running at 100%. But even though the card didn't output any signal, the computer continued to boot into windows, as if everything was working just fine

The second card also did not output any video signal. This card however, the fan was running at stock speed. But here also, the computer continued to boot as if nothing was wrong.

So, I dissassembled the cards, to see if there was anything visibly broken on the cards. But as far as I could see, nothing seemed broken.

I then noticed these Vrm chips and they did not look at all like what I was expecting. I was expecting to see something like mosfets, or transistors or something, but in stead, there where nine tiny chips on the right side and at the top of the card.

So I realise that I've propably gone in way over my head here and that these cards should propably be looked at by proffessionals, but I was just wondering, could it be as easy as to just replace these chips and the cards would be up and running again?
Cause if so, then I think I should be able to replace them with my workstation as long as I can find the needed parts somewhere

[momaka]

Re: Two faulty Radeon 5970 cards

So I realise that I've propably gone in way over my head here and that these cards should propably be looked at by proffessionals, but I was just wondering, could it be as easy as to just replace these chips and the cards would be up and running again?

It's possible, but not highly likely.

Do you have a multimeter? If so, start checking the MOSFETs for short circuit between Source and Drain. Keep in mind that some MOSFETs may apart shorted due to surrounding components in the circuit.

If you're not sure what I'm talking, just post some pictures of the cards here (use the "Manage Attachments" forum feature for that) and we can tell you what to check.

[severach]

Re: Two faulty Radeon 5970 cards

Since all regulators have a capacitor on the output, checking the capacitors for shorts is sufficient. Short testing is done on the diode or low resistance scale with the equipment unpowered. Capacitors are easier to identify and less confusing to test than MOSFETs. Read the meter carefully. Many meters beep when the diode drop voltage is less than 0.075v because that would usually be considered a short. On a high current low voltage GPU or CPU that is a perfectly normal reading in some places. Shorts tend to be 0.008v or less and some DVM do not measure that low properly.

Voltage regulators are tough and hard to burn out. It's their job to take an uncertain input voltage and put out a definite output voltage. 16v into the many linear and PWM regulators many of which start with 12v shouldn't be a problem.

The big ATI 5970 chip is the high risk. It is fed by many distinct voltages ranging from 3.3v to 0.85v that come from the many regulators around the board. Many of the chip's voltage inputs are very sensitive and a bit too much is enough to short the internals out. There isn't much on a video card that 16v on the 12v line would damage because 12v is used entirely for voltage regulators. None of the card runs at 12v.

Since the cards run, you can also test the capacitors for voltage with the equipment powered taking care to not short anything with the leads. Write the voltages down along with the capacitor voltage rating and you may see patterns emerge. A shorted regulator MOSFET would result in a capacitor with zero volts like a shorted GPU or linear regulator would result in one of the lower voltages shorting to one of the higher voltages.

I think the power supply did much more than jump to 16 volts before the end. A highly unstable 12v can trick the many regulators into producing voltages they aren't supposed to.

[oleq83]

Re: Two faulty Radeon 5970 cards

Ok. Thanks a lot for the reply guys.

Hmm. I guess there's a lot of different components that could be broken then.

I took some pictures of the card. I decided to start of with the card which had the fan at stock speed.

When I looked over the card, I noticed that one of the components (I think it is a capacitor) looked a little dirty, so I used my cheapo usb microscope to get a clooser look at it. As you can see in the pictures that I have included, the capacitor on the right side, had a thine "line", running across it. I tried to pick at it with a tweezer, to see if it was just a piece of hair, or something else, that had been placed on the component. But when I picked on it with my tweezer, I discovered that it was in fact a small crack.


So, I tested these three capacitors, or components, using the diode function on the multimeter.
Using the positive probe from the multimeter on the left side of each component, I got these readings. Starting from the left: 0.648v, 0.000v, 0.217v.

I compared these readings with the other card. On the other card I got these readings. Starting from the left: 0.669v, 0.669v, 0.702v.


Could this mean that the voltage regulators for the two capacitors which had different readings are broken?
And could this also mean that the Gpu's could have gotten too high voltage, causing a short inside the gpu?

[momaka]

Re: Two faulty Radeon 5970 cards

When I looked over the card, I noticed that one of the components (I think it is a capacitor) looked a little dirty, so I used my cheapo usb microscope to get a clooser look at it. As you can see in the pictures that I have included, the capacitor on the right side, had a thine "line", running across it. I tried to pick at it with a tweezer, to see if it was just a piece of hair, or something else, that had been placed on the component. But when I picked on it with my tweezer, I discovered that it was in fact a small crack.

That ceramic capacitor is indeed bad. Ceramic capacitors tend to go short circuit when they go bad. Remove it from the board.

So, I tested these three capacitors, or components, using the diode function on the multimeter.
Using the positive probe from the multimeter on the left side of each component, I got these readings. Starting from the left: 0.648v, 0.000v, 0.217v.

Looks like the middle and rightmost ceramic capacitors are bad. However, these may not be the only bad ceramic capacitors on those rails. If you look above them, you will see other smaller tan-colored ceramic capacitors - those can be bad too.
What to do: remove the ceramic capacitors that are showing 0.000v and 0.217v. The use the diode test function on your multimeter again to check the spots where those ceramic capacitors were located. If you still get the same readings, then you haven't found the shorted capacitor(s) yet. In that case, keep looking with your multimeter for capacitors that show the same 0.000v and 0.217v readings. Once you find all (or as many as you can) capacitors like that, see which ones look the most dodgy and burned and remove them.
Also, don't throw away the capacitors you remove yet. Check them with your multimeter on diode test function instead. If they show open-circuit (multimeter acts as if you have nothing connected across its test leads), then the ceramic cap in question is likely fine.

Could this mean that the voltage regulators for the two capacitors which had different readings are broken?

Can't tell from the measurements you did. First remove the ceramic capacitors that appear shorted, then we will see.

And could this also mean that the Gpu's could have gotten too high voltage, causing a short inside the gpu?

Maybe, maybe not. Most likely no.
This is a bit like the lottery, yes, but you get some learning experience from it.

[skeeter105]

Re: Two faulty Radeon 5970 cards

Hello, I have several Radeon 5970s with this exact same issue. I was looking at replacing the capacitors as they appear shorted when testing them with my DMM. The issue i'm having is i'm unable to figure out what size and type of capacitors I need to replace them with as they have no identifying marks. Dose anyone know what caps I need to order to replace these or any suggestions on how to identify what size these caps are?

Thanks,

[bravo-tel]

Re: Two faulty Radeon 5970 cards

10nf

[chrle]

Re: Two faulty Radeon 5970 cards

hello,I have radeon 5970,when I power up my pc, fan is blowing 100% ,no signal ,and 2 red dieodes are lighting. what could be problem?

I have 1000w corsair power supply.