128K SRAM cache on AM386DX-40 board not working \ VOGONS

128K SRAM cache on AM386DX-40 board not working

Topic actions

First post, by luRaichu

Posted on 2025-03-11, 02:34

luRaichu Offline

Rank Newbie

Rank: Newbie
Posts: 79
Joined: 2024-12-16, 18:22

I installed four 32Kx8 SRAM chips from eBay (ISSI IS61C256AH-15N) on my 386 board to obtain a total of 128K processor cache, removing eight 8Kx8 chips (64K cache). There are two SRAM banks so I filled bank 0 and left bank 1 empty as the board’s booklet says, set the jumpers correctly. The BIOS prints “128K CACHE MEMORY” and boots to MS-DOS correctly, but Windows 95 crashes. It will start normally if cache is disabled in BIOS settings. I’ve experimented with various BIOS parameters and yup, the 128K cache is what’s killing Windows.
In a stroke of genius I decided to remove the 32K chips and revert to the original 64K cache configuration which was working with Windows, leaving the 8K chips loose in the sockets for EZ removal… And that’s how I let the magic smoke out of one of the original 8K cache chips in bank 1. They’re Samsung KM6865BP-15’s.
Can I use a different 8K chip with faster/equivalent access time (15ns or less) to replace the fried one? I’m thinking of buying the one pictured.

The attachment 1330824A-6EBB-4223-8067-1FCED3ED5AF0.jpeg is no longer available

I hope I haven’t damaged my motherboard.

Reply 1 of 8, by Disruptor

Posted on 2025-03-11, 03:00

Disruptor Offline

Rank Oldbie

Rank: Oldbie
Posts: 1823
Joined: 2018-03-22, 18:31
Location: European Union

Both SRAMs have same pin layout.
But this SRAM has a 20 ns access time (H-5).
I'm not sure whether this is fast enough for a 40 MHz clocked 386.

Reply 2 of 8, by jakethompson1

Posted on 2025-03-11, 03:08

jakethompson1 Offline

Rank l33t

Rank: l33t
Posts: 2033
Joined: 2015-11-17, 04:16

If you don't have a TL-866 II (or similar) already, you may want to get one to identify dud cache chips, and save you a lot of frustration.

Reply 3 of 8, by MikeSG

Posted on 2025-03-11, 10:29

MikeSG Offline

Rank Member

Rank: Member
Posts: 478
Joined: 2023-02-14, 08:31

You could try adding a wait state to the cache or RAM if it works in DOS but not windows.

Reply 4 of 8, by stef80

Posted on 2025-03-11, 10:39

stef80 Offline

Rank Member

Rank: Member
Posts: 402
Joined: 2020-04-04, 16:34
Location: EUSSR

jakethompson1 wrote on 2025-03-11, 03:08:

If you don't have a TL-866 II (or similar) already, you may want to get one to identify dud cache chips, and save you a lot of frustration.

100% agree. Also good for un-bricking boards/BIOSes.

Reply 5 of 8, by luRaichu

Posted on 2025-03-11, 15:51

luRaichu Offline

Rank Newbie

Rank: Newbie
Posts: 79
Joined: 2024-12-16, 18:22

MikeSG wrote on 2025-03-11, 10:29:

You could try adding a wait state to the cache or RAM if it works in DOS but not windows.

How would I do this? It doesn't appear to be an option on my Phoenix BIOS.

Reply 6 of 8, by MikeSG

Posted on 2025-03-11, 18:54

MikeSG Offline

Rank Member

Rank: Member
Posts: 478
Joined: 2023-02-14, 08:31

Any options related to CAS delay?

May be best to try different cache chips though

Reply 7 of 8, by vstrakh

Posted on 2025-03-11, 20:36

vstrakh Offline

Rank Member

Rank: Member
Posts: 423
Joined: 2021-04-09, 12:31
Location: Ukraine, Dnipro obl.

MikeSG wrote on 2025-03-11, 18:54:

Any options related to CAS delay?

These are the asynchronous static ram chips, they do not multiplex the address lines (no CAS/RAS) and they do not spend clocks waiting for data to populate, so no latency clocks.
There's only one parameter - the access time - the delay from the change in the address lines values to the valid data on the output lines, it's an absolute value in nanoseconds, and the smaller that number - the better, faster chip will always work in place of slower chip. The 12ns access time is a possibility for chips in DIP packages, and 20ns is what typically can be found on old motherboards.
Putting faster chips in place of slower chips makes no difference at all if you do not bring system clock higher - faster chips will continue functioning at higher speeds, and slower chips can't go higher.
Although, there are motherboards that will insert wait cycles when accessing asynchronous cache, so putting faster chips will allow selecting lower latency in bios.

Reply 8 of 8, by luRaichu

Posted on 2025-03-23, 15:13

luRaichu Offline

Rank Newbie

Rank: Newbie
Posts: 79
Joined: 2024-12-16, 18:22

I installed Micron Technologies MT5C6408-10 to replace the previous burnt 8k chip, and the 386 is booting to Windows correctly in 64k cache memory configuration.
Will try a different set of 32k cache chips later.

Go to top of page Go to top of page

Back to General Old Hardware