VOGONS


First post, by pshipkov

User metadata
Rank l33t
Rank
l33t

A while ago upgraded this card Number Nine 9FX Motion 771 VL (S3 Vision968-P) from the standard 2 to 4 Mb of RAM.
I think i saw 4Mb reported by SpeedSys at the time, but recently i noticed that only 2 are reported again.
Now i am starting to doubt myself that i ever saw 4Mb.

The 2 additional megabytes are added on the back.

The attachment 20241110_111806.jpg is no longer available

Double and triple checked that all pins are properly soldered properly.
Soldered capacitors of the same properties as the ones in the front - 110mF.
The two resistors R44 R45 in the upper left are also matched as in the front (R42 R43). Their "output" voltages are slightly above the 0.53V readings in the front ones. I have them at 0.55V.
Checked voltages all around - they match the ones from the chips on the front.

There is always a chance that at least one of the newly added chips failed, but i doubt it. Wonder if somebody has an idea of what may be wrong, before i get quick with the soldering tools and potentially add damage to the assembly along the way.

Thanks in advance.

retro bits and bytes

Reply 1 of 8, by kingcake

User metadata
Rank Oldbie
Rank
Oldbie

Either the vga option rom is different, which I doubt, or there's a configuration resistor/link somewhere you're missing.

Reply 2 of 8, by pshipkov

User metadata
Rank l33t
Rank
l33t

There are only 2 resistors on the back. I tried few different combinations - both open, one open, the other open, 2-3 different swapped the resistors around.
I am hesitant to keep messing around because the pads integrity will start deteriorating.
Unless there are resistors that must be removed on the front, i am not sure what to do right now.

Anyone ever saw one of those cards with 4Mb of RAM installed ?

I found few references of the FX 771 series but in PC format - very similar layout with 2Mb RAM on the back - took the resistor types from there.
Still, best option will be to see the exact same VLB model with 4 megs and working.

Don't think the ROM has anything to do with RAM recognition/initialization, does it ?

retro bits and bytes

Reply 3 of 8, by willmurray461

User metadata
Rank Newbie
Rank
Newbie

Not sure about your card, but mine (4MB) uses TMS55160 not TMS55165. At first glance, the two chips may seem identical, but the pinout is slightly different. TMS 55160 uses dual CAS, whereas TMS55165 uses dual WE. Pretty sure that's your problem.

Also, I'm fairly certain the capacitors next to the RAM chips are not all the same value. Half of them are visibly taller than the other half, suggesting different capacitance.

In case it is helpful, I attached my card's BIOS (again, 4MB). The card doesn't work though, so there's a chance something's wrong with the BIOS.

Reply 4 of 8, by pshipkov

User metadata
Rank l33t
Rank
l33t

Hey, thanks for the response.
This has been settled but forgot to update the status of the thread.
Yes, i used the wrong chips. Corrected it later.
However, i am using the same capacitors everywhere (matched the ones on the front).
Things still work, but if you can check the capacitance of the elements on your card that will be awesome.
At minimum - can you provide a photo of the back of your card?
What are the values of the two resistors R44 and R45.
I am wondering how off my approximations are from the ground truth?

Too bad your card is busted. Do you have a hunch what could be the problem?

retro bits and bytes

Reply 5 of 8, by willmurray461

User metadata
Rank Newbie
Rank
Newbie

Here are some pictures. As you can see, your resistor choice was correct. Also, as you can see, the capacitors on the back are different heights. It's the same on the front of the card, so if you want to know the values, just measure the ones on the front of your card. The values shouldn't be super important anyway though. I'm pretty sure they're just decoupling capacitors to smooth out power delivery to the memory chips.

The attachment IMG_7376.jpg is no longer available
The attachment IMG_7374.jpg is no longer available
The attachment IMG_7375.jpg is no longer available

When I got my card, some of the pins on the 968 chip were bent in a way that made them touch each other. Before powering the card on, I straightened the pins back. I'm not sure if the previous owner could have powered the card on while these pins were shorted, which could have damaged the 968. I would think the chip is more durable than that though, and I also don't see any burn marks. I don't see any broken traces either...

Do you know of any specific BIOS settings this card needs to function? Any specific motherboard settings? I have two Asus boards, a PVI-486SP3 and a VL/I-486SV2GX4, and neither of them work with the card. Also, could you attach a dump of your card's bios? Maybe mine got corrupted somehow...

Reply 6 of 8, by willmurray461

User metadata
Rank Newbie
Rank
Newbie

Upon further inspection, I did notice one strange thing. When powered on, the 968 chip does not get very warm. It's barely hotter than room temperature. Compare that to my Mach32 VLB, which gets noticeably hot after only a few minutes.

It reminds me of a dead motherboard I had once. The CPU never got hot, and it turned out the trace between the clock and the CPU was broken. Restoring that trace revived the motherboard.

Maybe the oscillator on my card is broken or the trace connecting the clock to the 968 is damaged? If you can, can you test whether the 968 chip on your card gets hot after a few minutes of being powered on, or does it stay around room temperature?

Otherwise, it's probably time for me to get out the oscilloscope and start poking around.

Reply 7 of 8, by willmurray461

User metadata
Rank Newbie
Rank
Newbie

In case anyone is interested, I was able to revive the dead VLB 771 by replacing the 968 chip. My guess is that the original chip died due to the shorted pins. I sourced the 968 from a much more common PCI donor card.

Reply 8 of 8, by pshipkov

User metadata
Rank l33t
Rank
l33t

nice. congrats.
these vlb cards are somewhat rare, so good to hear one of them is functional again.

retro bits and bytes