VOGONS


First post, by imi

User metadata
Rank l33t
Rank
l33t

edit: I'm just going to keep using this thread for my problems with this build, the cache issue is solved, see below for new issues ^^

Ok, I'm at a loss here... I got a nice little 386DX40 board from a box of stuff:
https://www.arvutimuuseum.ee/th99/m/S-T/32636.htm

HOT327.jpg
Filename
HOT327.jpg
File size
352.89 KiB
Views
877 views
File license
Fair use/fair dealing exception

The board had two W24257AK-15 cache chips on it, one in U10 where the TAG cache sits according to the layout, and one in U11.
I did not further investigate this at first (as in look up the board or cache setup) I just tried turning it on, and it would not post, I held my finger on the CPU and gpu to see if they're warm... and when I wanted to turn it off again I accidentally slipped with my finger over the cache and almost burnt myself, they were scorching hot.

I had a really hard time taking the cache out because they were jammed in the sockets and there was no space really between the ISA and ram slots.

After I removed the cache the board bootet up.
So far so good... cache defective? can they really get that hot if they're defective?

So after looking up the board and reading a bit it says there should be one 8K x 8 in the TAG socket and 4 32K x 8 in the cache sockets
the chips that were in it are both 32k x 8... so now I got one W2464AK-20 8K x 8 for TAG and 4x W24257AK-20 32K x 8, this should be the correct setup?

This time I put some cut cable ties underneath the chips so there would be a little bit of wiggle room to get them out again and put them in.
...turned the power on... and saw smoke rising 😵 immediately turned it off again, and yes... the cache was scorching hot again, I couldn't even touch it, almost burnt my finger again just trying, it was maybe on for a second.
I assume the smoke came from the plastic of the cable ties underneath and not the cache chips themselves...
so did I just kill a bunch more cache chips? or are all of them defective? 😵

I can't see any obvious fault or damage on the board apart from the missing battery.
does anyone know what could cause the cache to just burn up immeaditely upon turning on the system?

Luckily the board did still turn on after removing the cache again, phew.

I don't want to kill the board or more cache chips by trying out stuff, so I thought I'd ask if anyone knows what could be the culprit here?

Thanks in advance.

edit... okay call me crazy but... uh I just measured the sockets with my multimeter, and VSS/VDD seem to be entirely the other way around
are they just inserted the wrong way?
if so, the picture on TH99 is wrong, as it shows the notch towards the ram slots, and apparently the old cache was also inserted this way.

Last edited by imi on 2020-05-20, 23:01. Edited 7 times in total.

Reply 1 of 12, by dionb

User metadata
Rank l33t++
Rank
l33t++

Fire up the board without cache, get out the multimeter and check whether 5V is really coming out of the pins it should be coming out of. This almost sounds like the sockets are the wrong way round or something...

Reply 2 of 12, by imi

User metadata
Rank l33t
Rank
l33t

thank you, I just did that while you replied, that seems to be exactly it.
unfortunately the image on TH99 is wrong apparently (or they are not the same boards 😕 ) and there is no indication on the board about orientation, but apparently whoever had the board before me did run into the exact same issue.

so can I assume those cache chips are dead now? or can I still try them the RIGHT way around?

edit: it would post now, and the cache doesn't burn, but kept hanging during boot and didn't recognize all ram.
that either means that yes, the cache is busted now... or could it be that the tag chip is at the wrong socket, that those are flipped too?

Last edited by imi on 2020-05-20, 23:01. Edited 1 time in total.

Reply 3 of 12, by imi

User metadata
Rank l33t
Rank
l33t

Ok, so I put a fresh set of cache in the board, now it boots up and recognizes 16000KB RAM, which is the maximum cachable amount with 128k I guess? but why 16000 and not 16384? ^^
Without cache it detects the full 32768MB, is there any setting to get the full amount with cache enabled, or should I just stick to 16MB anyways?
Guess I need to find two 8MB sticks then to not waste all that RAM ^^

Reply 4 of 12, by Tiido

User metadata
Rank l33t
Rank
l33t

in the first 1MB, top 384KB is the upper memory which is used up by system BIOS, video BIOS and video memory in most configulations. All option ROMs and whatever else goes there and most motherboards do not count that area in the memory count.

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 5 of 12, by imi

User metadata
Rank l33t
Rank
l33t

so upon searching through vogons I came upon this thread featuring the same board:
Some questions about 2 386DX-40 mobos
and at least according to this picture the tag cache seems to sit on the opposite side, so the whole thing is 180° around... I currently have the 8k x 8 chip sitting in the outer most socket as per the image from TH99, I don't know what the expected behaviour is with the tag cache sitting in the wrong socket, but I guess I should try flipping them around?

I wish I had a manual for this board ^^

edit: uh yeah, sure enough, now it detects all the ram again with cache enabled, the TAG goes in the inner socket like on the other photo, the whole cache setup on TH99 is flipped 180°
that cost me 5 chips 😢 +the 2 probably already dead ones that were on the board.

now if only I could get my gotek to work 🙁

Last edited by imi on 2020-05-20, 23:02. Edited 2 times in total.

Reply 6 of 12, by Thermalwrong

User metadata
Rank Oldbie
Rank
Oldbie

That's good it's working though 😀

I actually had a couple of SRAM chips start to emit smoke. I can't remember if they were in backwards or in the wrong position for a larger slot, they actually still worked after that though. Maybe test them again to see if they still work?

Reply 7 of 12, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t
Tiido wrote:

in the first 1MB, top 384KB is the upper memory which is used up by system BIOS, video BIOS and video memory in most configulations. All option ROMs and whatever else goes there and most motherboards do not count that area in the memory count.

The memory located between 640kb and 1MB isn't normally used by the system BIOS, but the address space is. However, you want that memory overlapping the option ROMs, because the empty gaps in between can be used as UMB memory, freeing up more conventional memory in DOS if you load drivers in there. A lot of motherboards have an option in the BIOS to relocate that area above 1MB, but as far as I know that option is only somewhat useful on systems equipped with 1MB of RAM.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 8 of 12, by imi

User metadata
Rank l33t
Rank
l33t

I'm just going to keep using this thread for all my issues with this build:

step2, my gotek emulator is not working and I can't find aynthing to solve my issue.

the S1 jumper on the gotek is set and I have to set the JA jumper too apparently or otherwise the system just stops after posting and doesn't do anything
if it is set to S0 it says "A: drive error"
the drive is connected after the twist.

if S1 and JA is set it just says this after post:
DRIVE NOT READY ERROR
Insert Boot diskette in A:
Press any key when ready

the green LED lights up on drive access

now as soon as a usb stick is in the gotek, I cannot change the disk number anymore, if I press the right button it briefly flashes "d0" and if I press the left button it very very shortly flashes something I can't make out... this is not normal behavior I guess?
video of what is happening: https://www.youtube.com/watch?v=kJQ02915z4M

I tried formatting the stick by holding both buttons while turning on, this didn't work apparently, it just displays the firmware version "F31" "13A", so I used the "USB Floppy Manager 1.40i"
I tried moving MS DOS 6.22 on 00, 01, 02 manually, and I tried writing it as images.

I tried a different I/O controller, but no changes.
I tried the gotek on a newer motherboard and the behavior is exactly the same.
I tried two different USB sticks still the same.

I have a SFR1M44-FEL-DL (I didn't really pay attention when purchasing it apart from that it is a 1,44MB version) ...does that simply not work on PC? what is the difference between the different gotek versions?

I also can't switch the disk number on the gotek with a usb stick plugged in even when the drive isn't connected to the PC at all.

Reply 9 of 12, by imi

User metadata
Rank l33t
Rank
l33t

so... the version of the Gotek does matter after all, after I couldn't get mine to work I ordered another one... this time a SFR1M44-U100K that gets mentioned the most, plugged it in, and it just works.

the second jumper is set differently on this one ootb, it bridges JA and J5, I only found one picture searching around with it jumpered like that and that was talking about reflashing the drive... but even if I set the jumper like this on the other one it won't change it's behavior.

anyways... new gotek, correct version, and it boots up DOS now 😀

next step, get the IDE-CF to work.... I don't think the bios supports 2GB (which is the smallest CF card I have atm) ^^ any tips on what cylinders/heads/sectors settings to use?
I plan to get a 128-512MB CF in the future.

Reply 10 of 12, by appiah4

User metadata
Rank l33t++
Rank
l33t++

On your old Gotek, flash it to Flash-Floppy, set the jumper to S0 and and using FF.CFG set the type to IBM PC (instead of Shugart as the original firmware is normally configured) and it will work. I've had that happen before.

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 11 of 12, by imi

User metadata
Rank l33t
Rank
l33t

thanks, I assumed reflashing it would probably fix it and I'm planning to reflash them in the future anyways, I just wanted to get it up and running now so I can actually go and start to test stuff 😀

Reply 12 of 12, by imi

User metadata
Rank l33t
Rank
l33t
imi wrote:

there is no indication on the board about orientation, but apparently whoever had the board before me did run into the exact same issue.

quoting myself here...

I stumbled upon the picture I took for this thread yesterday while looking through stuff... and actually saw that the cache sockets are indeed marked with an orientation, there is a '1' and a '28' on the socket indicating pin orientation, idk how I totally missed that the first time around... but IT IS WRONG... the sockets themselves are mounted on the board in the wrong orientation.

so yeah I can see how someone made same mistake before me, and how the error on the drawing of the board came to be.