MV1B VRAM issue

[bodgit]

I recently built some of @ack's excellent Neo Geo diagnostic PCBs just as a useful thing to keep handy.

Testing them out on my MV1B I wasn't expecting to have any issues. However the first time I tried them with the latest v0.19a02 diagnostic BIOS, I got a VRAM 32K error on power up which I stupidly didn't write down. When I power cycled, it didn't occur again when I left the tests running in a loop.

I've just tried again today and it's failing again semi-repeatedly on the VRAM 32K loop although sometimes the tests can run for ages with no issue, my longest run was around 0x1600/5632 passes without issue.

Address-wise, it's failed on a wide range of addresses, sometimes multiple times on the same one, (0x20 being the most common), anywhere from 0x20 all the way up to 0x70a0, I presume anywhere up to 0x7ff0 is fair game.

From my understanding the VRAM is comprised of RAM 5 & 6 and is accessed as 16-bit values. When the tests fail, it has always been (so far) the most significant byte that is wrong, for example:

ADDRESS:  001000
ACTUAL:   D7FF
EXPECTED: FFFF

As the two KM62256CLG-7 chips for RAM 5 & 6 are individually accessed as 8-bit values, am I correct in thinking that one chip holds the MSB and the other holds the LSB? This would point to just one of the chips being bad. Any idea which is which so I can just replace the right one?

I can source W24258S-70LE chips slightly easier than the original KM62256CLG-7, I presume these are a direct swap here?

I had an issue on this MV1B that I never got to the bottom of here: https://www.arcade-projects.com/threads/graphics-corruption-on-mv1b.20180/

I did run the diagnostic BIOS at the time and the VRAM error never showed up, perhaps this is the cause?

 

[ack]

As the two KM62256CLG-7 chips for RAM 5 & 6 are individually accessed as 8-bit values, am I correct in thinking that one chip holds the MSB and the other holds the LSB? This would point to just one of the chips being bad. Any idea which is which so I can just replace the right one?

That is correct.

https://github.com/jwestfall69/neogeo-diag-bios/blob/master/docs/ram_locations/mv1b.md

can source W24258S-70LE chips slightly easier than the original KM62256CLG-7, I presume these are a direct swap here?

Those should be fine.

 

[bodgit]

That is correct.

https://github.com/jwestfall69/neogeo-diag-bios/blob/master/docs/ram_locations/mv1b.md

From reading that, am I correct in translating upper = MSB, lower = LSB? That would point to RAM 6 being the culprit.

The only reason I ask is I have a NeoBiosMasta VMC fitted over the 68K CPU and it partially covers RAM 5. I remember it being quite stiff to fit and I'd rather not have to try and pry it off again if I don't need to touch RAM 5.

Those should be fine.

I ordered the W24258S-70LE from eBay and the seller refunded me this morning as they clearly didn't have the right thing. I found some W24258S-70LL instead which just look like a lower power version, still 5V supply, so I'll try those.

 

[ack]

From reading that, am I correct in translating upper = MSB, lower = LSB? That would point to RAM 6 being the culprit.

correct.

 

[bodgit]

Finally got around to replacing RAM 6. Whipped the old chip off:

I tried booting to confirm that was the right chip to remove but I just got a green screen and some R2-D2 warbling that I assume is code for "you're missing half of your VRAM", so I installed the new chip:

The error seems to have gone, I left the VRAM 32K test looping for about 15,000 passes with no error. It also seems to have fixed the glitching graphics issue I had with the Hypernoid homebrew.