Emboldened by the last repair, I ordered the second card from the seller.


In my naïvety, I was expecting the second card to have been better treated. Alas, even more scratches.





Checks
Compared to the previous card, the trace damage is more extensive but surprisingly, traces have continuity.
Another interesting aspect, the boards are virtually identical, even their BIOS have the same checksum. The only difference is the FCC ID, this one having FCC ID: J6QGD5422DM2.
I proceeded to test the card and no surprises either, 1024Kb of RAM detected. But, even before starting the benchmarks, something seemed off. The text display was slower. The sister card text display was fast, blocks of text appeared all at once, while with this one, it felt more sluggish.
Why is it slow ?
The benchmarks confirmed my suspicion, same cards, but this one was slower:
How come two identical cards have different video performances, so widely apart ?
I didn’t know where to start. But I had an idea to fiddle with the BIOS settings, to see if they impact the card performance. I disabled the wait states on the board (with the jumper), 3DBench 1.0 score was now down to ~14 instead of the already slow ~20. Disabling the L2 cache had an impact too.
The thing that didn’t have an impact is toggling the Video BIOS Shadow. With or without BIOS Shadow, the 3DBench 1.0 result stayed the same, just as with the bios shadow on. Hm!
Why would this BIOS not be shadowed by the motherboard ? I started looking very closely around the BIOS ROM area on the component side of the board. And then I spotted it:

Two of the traces had no continuity, the ones going to these two pins of the ISA slot:

Fixing
I removed the “Q.C. Passed” label with a bit of label removal – as it was too closed to the affected area and I would have dirtied it with flux. Scraped the solder mask to expose a bit of the copper and proceeded to bridge the gaps. Under the microscope, I saw another affected trace.






The fix works and board has now the exact same results as its previously fixed sister card.
Before I applied the solder mask, I had one last look at the fix under the microscope. Looks a bit messy at this magnification.



And two more shots of the card:


This card (as well as its sister) have problems driving my LDC monitor, text is wiggly. The OAK OTI077 card has a more stable image.
Leave a Reply