VOGONS


First post, by vstrakh

User metadata
Rank Member
Rank
Member

The story begins here 486 motherboard L2 cache misconfigured on purpose?
I found that the actual cache size is 64KB, while the motherboard has soldered SRAM chips that should give 256KB of cache, and the bios was altered to misreport cache size as 256KB.

I've finally received the Renesas chips I've ordered from Mouser, so it's time to fix this issue.

First removing the fake chips. I don't have any sophisticated equipment (no fancy vacuum pump soldering iron), so I've preheated the relevant part of the motherboard with spare heating bed from 3D printer, then used hot air gun to heat the pins, chip by chip. When solder has melted, a slight knock on the mobo - and the chip falls out under its own weight. Idea was to not apply any force to avoid tearing/lifting the traces.

Desoldered chips, both sides:

The attachment fake_front.jpg is no longer available
The attachment fake_back.jpg is no longer available

It went very well and quite easy. The pads are in great condition, nothing damaged, no tears or burns.

The attachment desoldered1.jpg is no longer available
The attachment desoldered2.jpg is no longer available

The silkscreen labels suggests the board expects Cypress CY7C185 (8K) or CY7C199 (32K).

Last edited by vstrakh on 2022-10-16, 21:09. Edited 1 time in total.

Reply 1 of 7, by vstrakh

User metadata
Rank Member
Rank
Member

Now cleaning the holes from the solder with solder wick - that's a tedious process. The most time was wasted in this step:

The attachment holes1.jpg is no longer available
The attachment holes2.jpg is no longer available
The attachment holes3.jpg is no longer available

Now solder the sockets:

The attachment sockets_populated.jpg is no longer available

New chips to be installed:

The attachment new_chips.jpg is no longer available

Reply 2 of 7, by vstrakh

User metadata
Rank Member
Rank
Member

The final assembly:

The attachment assembled.jpg is no longer available

And finally the result - 256KB cache size, as it should be 😀

The attachment cache_256k_ok.jpg is no longer available

I did test all desoldered chips, they all pass tests as 8K x 8 sram, and none could pass the test as 32K x 8.
The date code at the top of the chips is the same on all 9 chips, but there are clearly at least two different batches when looking at the bottom of the chips, so they were sanded and relabeled.

Reply 3 of 7, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie
The attachment new_chips1.jpg is no longer available

They put this sticker there for the lulz.
And they labeled the ISA slots "SOLT" instead of SLOT. That there is one funny mainboard 😉

Congrats for finally solving this cache-mystery!

Reply 4 of 7, by mr-spain

User metadata
Rank Newbie
Rank
Newbie

Love stuff like this! GJ

Reply 5 of 7, by amadeus777999

User metadata
Rank Oldbie
Rank
Oldbie

Well done!

Reply 6 of 7, by AlexZ

User metadata
Rank Oldbie
Rank
Oldbie

It's amazing this board finally got the cache it deserves.

Pentium III 900E, ECS P6BXT-A+, 384MB RAM, GeForce FX 5600 128MB, Voodoo 2 12MB, Yamaha SM718 ISA
Athlon 64 3400+, Gigabyte GA-K8NE, 2GB RAM, GeForce 9800GT 512MB, Sound Blaster Audigy 2 ZS
Phenom II X6 1100, Asus 990FX, 32GB RAM, GeForce GTX 980 Ti