EVGA Video memory repair [770 GTX / 980TI]

[coquinho]

I have 2 EVGA boards, both have similar problems - but is a little more complicated than that. I tried theses repairs with interval of months, boards that came some time apart to my repair shop. Now I decided to share that I have done, maybe someone here have some idea about it.

1. The two boards had memory short circuit, but on the 980TI, the short was very visible, memory VRM had exploded a high Mosfet, shunt resistor, 12V inductor, total short on memory bank. On the 770, the VRM was fine, giving 1.5/1.35V normally, but the board had artifacts all over the screen, memory bank was measuring 2-3 ohms, instead of 70, 80+ ohms that is more usual.

2. I fixed the 980TI VRM, and replaced all memory modules with identical chips. I did the same with the 770, replaced all the memory chips.

3. Both video cards now can work normally without the driver, but when running the driver, massive artifacts on the screen, and crash. I discovered that with lower memory clocks, below 810mhz, both boards become stable. I tested benchmarks and games, no crash at all.

4. I reballed the 980TI chip - no result, same problem.

5. I changed the 770 GTX chip by other - no result, same problem.

6. I compared measures points on the 770 GTX with other board identical, and the measures was very similar between then.

Anyone have any idea why with high memory freqs the boards become unstable? I tried to read some schematics of notebook that uses discret GPUs (AMD Venus, etc) but I didn't find much to help. I just want to find what it is, I think that is some knowledge that could help in the future with other cards...

Anyway, Thanks for the attention 

[Rasz]

where did you get new memory? did you pull them yourself from dead cards, or sourced from china? chinese chips will most likely be remarked to look like faster parts

[wraper]

2. I fixed the 980TI VRM, and replaced all memory modules with identical chips. I did the same with the 770, replaced all the memory chips.

Seems that you don't value your time. If there was low resistance, it should be quiet easy to figure out which chip is bad as it will heat up a lot more than others when power is applied (from lab PSU if VRM is blown). I would guess you got chips with either remarked speed grade, model or even remarked chips made by different manufacturer (can check manufacturer with GPU-Z). IMO you can either replace those chips or try different GPU bios which is compatible with your card as it may have support for chips you used or have more loose vram timings.

[TiN]

Did you use same memory model/part number parts for replacements. Different memory chips often require different timings/settings to run properly on full-speed mode (which you get once driver is initialized).