VOGONS


First post, by st31276a

User metadata
Rank Member
Rank
Member

I have a Gemlight GMB-P54SPV motherboard that I got for free in 2004. It was hit by lightning through an attached modem; the serial line driver IC on the board shot a hole in its top.

I installed a SY062 pentium (120) in it and it worked, except that the thing hung when it booted if L2 cache was enabled in bios. I used it like this with a separate serial card ever since.

Yesterday I was poking around on the retro web and found an actual manual for this board (actually its DTK PAM-0036S equivalent) that is not truncated somewhere in the middle and could see all the jumper configurations for the first time.

The two yellow jumpers by the cpu socket control cache size, which is a strange provision, because it has 256k soldered-on pipeline burst sram and no sockets soldered in for the async srams the chipset can also make use of. It can also only use either the one or the other, not both simultaneously. The cache size is therefore rather fixed unless you whip out some soldering equipment, with the exception of jumpers that are provided to setup the (fixed) cache size - either correctly or incorrectly.

It turns out that the jumpers on this board were configured for 1M cache and not 256K.

When I moved the jumpers to the proper position, it worked!

Cue the sad trombone sound...

While cacheable limits were not an issue with no cache working and with the 5511 memory controller able to take loads of ram, this thing ended up having 256MB of it. Now suddenly I have cache working, so naturally I was getting curious which part of it is cached.

According to the SIS 5511 datasheet, it caches 64MB with 256K cache. Strange thing is, it refers to two tag chips, one 8 bits wide and another 1 bit. This board only has place for the 8 bit wide chip, would that cut it in half? I see no write through / write back optios in bios, only cache timings, so no gains are to be made by switching to write through.

Nonetheless, the cached part makes a difference in how it boots red hat 8 (2.4.18 i586 kernel) - I can hear it giving the hard drive beans at times instead of the lazy tickety-ticking it used to do, and it boots slightly faster fwiw.

Hope somebody finds this useless info interesting 😀