Have you done the upgrade according this manual ?

You have those chips installed backwards. If something is getting that hot, you should be stopping and questioning everything.

Google pictures of the motherboard, and you'll see that the semi-circle marking on the cache chips should be facing the riser slot.

You should do a good bit more research as far as how these chips should be positioned. they'll need to be rotated 180 degrees, but if you've started the motherboard, it's probable that the chips have been irrevocably damaged. I think that you'll want the 2 rows of pins towards the riser card empty.

View this thread for more information. The post I've linked shows how the cache should be situated.
Cache Upgrade for 486 PB

fitzpatr wrote:

I think that you'll want the 2 rows of pins towards the riser card empty.

Thanks, I appreciate the help. To be honest, in my 30 some years of computing, I've never tried to install cache, so even this is new for me. If you look at the link Roman555 provided, it shows how, when, where, etc. the cache should go. Since the text was rightside up, I went with what I saw there and never even thought the chips would be placed the opposite as all the other chips on the board.

If you'll see the portion of your post I quoted, are you talking about the 32kx8 and the 64kx1 right above the processor lever? Why would I want them empty? The cache upgrade page calls for those sockets to be populated by the chips that are there. Should they be rotated 180 degrees as well, just like the four 128kx8 chips that I have inserted incorrectly?

No problem. 486 and earlier is a very different beast from modern systems. Take your time, and do lots of reading. Most importantly, assume nothing!

All 4 cache chips (U16, U19, U29, U32), the TAG RAM (U55) and the DB RAM (U57/58) are all installed incorrectly.

Cache (SRAM) came in a DIP-28 package, which means that it has 2 rows of 14 pins, totaling 28. This was fine for 32kx8 bit chips (256 kilobit), but when larger capacities arrived, the package needed to be extended so that all of the data could be addressed. This was extended to 2 rows of 16 pins, or DIP-32. If you're using 128kx8 chips (1024 kilobit), it will take up the entire socket, but if you're using 32kx8 chips, it will need to have 2 rows of pins left empty. The semi-circlular notch in the chip indicates orientation. There will always be some kind of mark on the socket or the PCB silkscreening beneath it to indicate which direction the chip must be inserted.

Hate to bring a thread back to life here, but, on my Packard Bell Pack-Mate 28 Plus (same board revision: PB450), I installed the cache in the correct orientation (watched my dad install L2 cache in our old family computer in 1994), it detected the cache, but, once it gets to the OS, I get all kinds of errors or lock ups when attempting to run the OS, such as MS-DOS 6.22 or Windows 95C.

None of the pins are touching each other, I've straightened them out carefully, tested them one-by-one using my EEPROM programming tool and software on my laptop, which passed with flying colors, thankfully, and still nothing.

I have the following cache chips:

4x Winbond W24257AK-20 28-pin DIP 32K x 8 (256Kb) SRAM
1x WinBond W24257AK-15 28-pin DIP 32K x 8 (256Kb) SRAM TAG
1x AT&T ATT7C187P-15 22-pin DIP 64K x 1 (64KB) - Rebranded Cypress CY7C187-15PC*

* I will test this one when I get home and see if that's the issue, or if there's another issue, such as compatibility issues with Cypress cache chips. A mismatch between the Cypress 22-pin and Winbond 28-pin should work regardless, right? I cannot seem to find a Winbond 22 pin 64K x 1 15ns SRAM DB anywhere online.

So I'm here trying to find out about cache on this pb450 motherboard I bought on eBay a few years ago. And I'm noticing these melt marks on my cache sockets. I start comparing the manufacturing dates on the chips and then realize I have the exact op board. I guess he sold it when he couldn't figure it out? LoL

haha nice, does it still work?

So the W241024AK-20 seems to be hard to locate. However the 15ns and 12ns variants are available. Any downside to using the faster cache? The onboard ram has issues but can be disabled. Could any damage beside the OP cache and melting have occurred from thier backwards cache?

Yes, I would imagine plugging cache backwards might kill it in a way it temporarily passes VCC to all of its pins first, things on a memory bus might not like that. If you are lucky its only ram and not the chipset/cache controller.

Any downsides to trying fresh right way round on the damaged board? I can verify it functions on another PB450 first.

I would verify on fully working board.