Attempt to repair AMD FirePro

[Askic]

Hello guys,
inspired by this excellent tutorial made by dicky96 (this thread), I have decided to try to diagnose and repair faults AMD FirePro v4800 graphics card. This card is about 10 years old and I have another fully functional one, so this should make repair attempt a bit easier.

First of all, I have checked for shorts and compare as much measurements as I could (multimeter in diode mode) to the working one and couldn't find any significant difference.
So, when plugged in the motherboard, card is not detected and there is no picture on the screen.
Since I don't have an extension cable, I need to make all voltage measurements when card is plugged in the motherboard, so not all places are available.

First a bit history. These cards are known for its fan as a weak point. If operated in a dusty environment, the fan fails after two or three years, causing the overheating problem. This card had overheating issues and it worked without spinning fan for who knows how long before it simply failed.

After card is plugged in the motherboard and power ON, I was able to measure 12 VDC on two smaller black coils. They are part of the input circuit and together with green electrolytic caps they represent input filters. On two bigger coils I have measured stable 1.1V, and on the coil far right (marked with 1R0) I have measured 0.9V. So, it seems that all voltages are present, but there is no picture. I also noticed that GPU develops a lot of heat, and it is noticeable more hot than working one (even with replaced fan).
Is it possible that GPU is faulty because of overheating?

Is it necessary to try to find 1.8V when voltages are present on all coils?

[momaka]

Originally Posted by Askic

Is it possible that GPU is faulty because of overheating?

100% YES.

Don't take this offensively, but this is like driving your car and letting your engine overheat. Then it broke and you ask, could it be because it overheated? What do you think?

In this case, you have history of the GPU and you know it was overheating. So no point in wasting time trying to check any components or voltages. GPU chip is probably dead. You can try to bake it / reflow it / re-heat it / burn it with a piece of hot charcoal / whatever other creative heating method you can think of to try to revive it. But if you do, just beware that chances it will come back alive are small. And chances that it will last any decent length of time is pure luck... though generally, the cooler you keep it (provided the reflow even worked), the more likely it is to last longer... though still no 100% guarantee. By keeping cool, I mean keeping under 50-55C max under load. So that certainly means ditching the original cooler and going with something better. And for this kind of card, that probably isn't even worth it, unless you're doing it purely as a learning experience or to have fun.

All that said, though... don't get your hopes too high the reflow will work. IME, I've gotten about 15-20% success overall with getting a card to come back to life (most cards won't), and only a handful to still continue to work (after cooler mods and whatnot.) In other words, ask yourself, do you feel lucky.

[Dan81]

Considering it's a HD5000 series card (HD5670 to be exact) it should last quite a time. Assuming you use the right thermal paste of course (at least MX4 or MX5 is recommended for those), it should last pretty well. Add a bigger cooler as usually those cards come with poor heatsinks that aren't suitable for anything past HD3200 or 3450 max.

[dicky96]

Originally Posted by Askic

Is it necessary to try to find 1.8V when voltages are present on all coils?

No - as you understood from the video, 1.8V must be there by default, if you have PEX, Vcore, Vram - which obviously you do

Thank you for your kind words - I seem to have inspired quite a few people with that particular video (and even have some reported success stories by viewers who watched it on my YT channel then had a go themselves) so now it is time to blush


Ahh that's better
Rich

[Askic]

Hello guys and thank you for the replies. I feel I need to add more background information.
I have two DELL workstation that work as graphics servers with these type of cards. The PCs are about 10 years old. About two years ago, one PC started frequently to shut down by itself. After opening the case I saw that fan on the GPU card is not spinning at all. It was stuck with dust and dirt. The computer could start about used to shut down unpredictably. So, I cut out the existing fan and installed a bigger one on the side. It did the job and this server is today still in function without any issues. Have a look at the attached image.

Few weeks ago problem occurred on the second server and I also noticed that the fan on the GPU is not spinning, but this time the symptoms were different. Computer didn't detect GPU and there was no picture at all. So I knew that the problem is different and I knew there was a lot of overheating since fan was out of operation for unknown period of time. (The PC is periodically cleaned from dust by compress air once per year or so).

Anyway I tried to reheat and to re-flow and it didn't work. It was a good and fun experience anyway.

Do I feel lucky? Almost all the time

Thank you very much dicky96, because your video tutorial explained the most basic and important things, such as power pins, what voltages are there and similar. I'm sure t will be very helpful, because in future, I'll try to at least measure the voltages on faulty GPU cards instead of immediately replacing them.