VOGONS


First post, by Frank Grudge

User metadata
Rank Newbie
Rank
Newbie

Benchmark programs are tellling me my L2 cache is not working, ive noticed 2 absent 74244 ics on my board. (Also some decoupling caps) Comparison from the retroweb.
Could it be possible L2 was never enabled?
I am thinking to fit these 2 chips . Why are they missing?
Board = 4SGS-3VL

Thanks

Attachments

Reply 2 of 8, by Frank Grudge

User metadata
Rank Newbie
Rank
Newbie
jakethompson1 wrote on 2023-11-16, 08:14:

Is your tag ram (down by the first VLB slot) populated?

Yes it is populated identical to the rest of the L2 chips w61c256a

Reply 3 of 8, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

Your mainboard is configured fpr 256K cache with 8 x 32Kx8 chips. Jumpers 1,2 are soldered so cannot select any option with 512K or 1M L2 cache.

This might be a case of "fake cache" and they spared the buffer-drivers as well.
Have a look here:
Fake cache board, reboot loop after adding real cache

Those "dt" cache chips don´t look very common and the brand "Azza"...well...

So just replace the 9 chips, add the jumpers and the two buffer-drivers 74F244. (Of course you can give the current chips a chance first after adding the buffers.)

Reply 4 of 8, by mkarcher

User metadata
Rank l33t
Rank
l33t
majestyk wrote on 2023-11-16, 14:23:

Those "dt" cache chips don´t look very common and the brand "Azza"...well...

The marking of the cache chips looks like real cache chips. The "integral" symbol followed by "dt" is a manufacturer logo of IDT (Integrated Device Technology), and I am very confident I own working chips with that branding. The IDT 71256 is a real product, and the latest datasheet for them is Renesas branded. This of course doesn't mean that this branding can not be applied to fake chips as well.

Without the 74244 address buffers, the cache definitely will not work, though. I suspect someone tried to add cache by installing the SRAM, and it never worked.

Reply 5 of 8, by CoffeeOne

User metadata
Rank Oldbie
Rank
Oldbie
Frank Grudge wrote on 2023-11-16, 08:22:
jakethompson1 wrote on 2023-11-16, 08:14:

Is your tag ram (down by the first VLB slot) populated?

Yes it is populated identical to the rest of the L2 chips w61c256a

What do you mean? The other chips on the pictures are IDT ones.

Reply 6 of 8, by Tiido

User metadata
Rank l33t
Rank
l33t

On the second photo in first post, one chip has its pin1 bent and not in the socket.

Those buffers next to sockets definitely need to be there for things to work.

T-04YBSC, a new YMF71x based sound card & Official VOGONS thread about it
Newly made 4MB 60ns 30pin SIMMs ~
mida sa loed ? nagunii aru ei saa 😜

Reply 7 of 8, by jakethompson1

User metadata
Rank Oldbie
Rank
Oldbie
mkarcher wrote on 2023-11-16, 19:42:
majestyk wrote on 2023-11-16, 14:23:

Those "dt" cache chips don´t look very common and the brand "Azza"...well...

The marking of the cache chips looks like real cache chips. The "integral" symbol followed by "dt" is a manufacturer logo of IDT (Integrated Device Technology), and I am very confident I own working chips with that branding. The IDT 71256 is a real product, and the latest datasheet for them is Renesas branded. This of course doesn't mean that this branding can not be applied to fake chips as well.

Without the 74244 address buffers, the cache definitely will not work, though. I suspect someone tried to add cache by installing the SRAM, and it never worked.

Any thoughts why this board vendor populated the jumpers to hardwire the board for 256K, double bank cache, but didn't put those chips on there?
I have seen some boards where they omitted chips and replaced them with eight diagonal resistors; could the board vendor have done that instead or is the need to be able to tri-state those outputs inherent in the design?

Reply 8 of 8, by mkarcher

User metadata
Rank l33t
Rank
l33t
jakethompson1 wrote on 2023-11-17, 01:03:

Any thoughts why this board vendor populated the jumpers to hardwire the board for 256K, double bank cache, but didn't put those chips on there?
I have seen some boards where they omitted chips and replaced them with eight diagonal resistors; could the board vendor have done that instead or is the need to be able to tri-state those outputs inherent in the design?

This is just guesswork, but you asked for "any thoughts": The vendor shipped the board in the 256K an 0K configuration, and they only omitted components in the "0K" edition that had a significant cost in the BOM. Jumper wires might be expensive to install, but they are very cheap to buy, so the vendor didn't think about removing them when they specified the 0K edition.

The '244 is a unidirectional buffer, and at that location on the board, I am very confident that it is used a address buffer, input is the processor address bus, and output is the cache address pins. As the cache chips are the only thing on the output side, and the 0K edition doesn't have the cache chips installed, the tristate option is definitely not required. As long as loading of the address bus is not an issue, the '244 chips can be replaced by 8 jumper wires each.