VOGONS


First post, by neonumber6

User metadata
Rank Newbie
Rank
Newbie

Hello:

For some time I've been trying to setup a "4386-VC-V" motherboard ( https://stason.org/TULARC/pc/motherboards/U/U … -4386-VC-V.html ) with 128 MB of RAM so that I could use it with Linux.

The board is currently populated with 256 KB of external cache and a 486 DX33 MHz CPU.
According to the stason link this CPU should allow 128 MB of RAM.

But unfortunately when I run memtest-86 v3.0 any attempts to use over 32 MB of RAM results in all kinds of errors.
If I use 32 MB or less memtest-86 V3.0 completes with no errors.

I've tried both 30-pin PARITY and 30-pin NON-PARITY memory modules -- in each case 8 modules of 16 MB each.

The board appears to be somehow currently (?) limited to addressing 32 MB of RAM

On bootup I see this:
"
AMIBIOS (C)1992 AMerican Megatrends Inc.,
Version 1.02

[...]

X0-0100-00121-00101111-021993-VIA-F
".

The BIOS chip label on the motherboard says "1993".

Could there be some BIOS update I need to do in order to allow use of 128 MB of RAM?
Would anyone be able to provide the instructions on how to proceed with a BIOS update?

Thanks for any help in advance...

Reply 1 of 13, by Strahssis

User metadata
Rank Member
Rank
Member

Have you tried 64MB of RAM yet? It says on Stason that the maximum amount of RAM (64 or 128MB) depends on the CPU you have installed. Maybe with the 486DX 33MHz that is 64MB.

It also says on the page that you have to adjust the jumper settings for the amount of cache you have installed. I am assuming you already did that, but if you didn't that would be your first thing to do.

As for the BIOS update, I think this is the one you're looking for (second post on the page): 80486 BIOS image collection
I have experienced two ways with my computers and laptops for this. I have either had an executable that would update the BIOS through a Windows installer. Or either I have has an executable that made me make a bootable floppy disk with the BIOS update on it. For a 486 I expect it to be the second option. You can also usually make the floppy disk on another machine. I have done it on Windows 2000 for example.

I personally never used more than 32MB of RAM on a 486. Maybe it's just me, because I haven't really experienced a performance increase in Windows 95 beyond 16MB of RAM on a 486.

Mimi: AMD K6-2/266, S3 Trio64, Diamond Monster 3D II, Sound Blaster CT2800, 32MB RAM
Satellite 220CS: Pentium 133, SVGA DSTN, Sound Blaster Pro, 64MB RAM
Contura 420CX: 486DX4 75, VGA TFT, Roland Serial MIDI, 16MB RAM

Reply 2 of 13, by neonumber6

User metadata
Rank Newbie
Rank
Newbie
Strahssis wrote on 2022-05-10, 15:47:
Have you tried 64MB of RAM yet? It says on Stason that the maximum amount of RAM (64 or 128MB) depends on the CPU you have insta […]
Show full quote

Have you tried 64MB of RAM yet? It says on Stason that the maximum amount of RAM (64 or 128MB) depends on the CPU you have installed. Maybe with the 486DX 33MHz that is 64MB.

It also says on the page that you have to adjust the jumper settings for the amount of cache you have installed. I am assuming you already did that, but if you didn't that would be your first thing to do.

As for the BIOS update, I think this is the one you're looking for (second post on the page): 80486 BIOS image collection
I have experienced two ways with my computers and laptops for this. I have either had an executable that would update the BIOS through a Windows installer. Or either I have has an executable that made me make a bootable floppy disk with the BIOS update on it. For a 486 I expect it to be the second option. You can also usually make the floppy disk on another machine. I have done it on Windows 2000 for example.

I personally never used more than 32MB of RAM on a 486. Maybe it's just me, because I haven't really experienced a performance increase in Windows 95 beyond 16MB of RAM on a 486.

Wow thanks for such a quick response -- I was not expecting that 😀

1. Yes I tried 64 MB of RAM -- same results -- lots of errors.
2. I did carefully set all the jumpers for ALL the hardware I have so no doubts there.
3. I've carefully confirmed that all the cache chips are 32Kx8 as expected for 256 KB external cache.
4. Now that I think of it I wonder if the 486 DX 33 CPU would not be limited to 32 MB of RAM... I do have a spare 486 DX2 66 CPU on hand... I guess I should try to replace the DX33 with DX266 (and update ALL related jumpers) this week.
5. I'm very comfortable with DOS so if step #4 above does not help I think I'll try setting up a boot floppy with the AMIBIOS update.

I will let you know how #4 goes.
In the meantime if anyone else has any ideas I can definitely try a few things.
Will keep #5 as the last try.

Reply 4 of 13, by Strahssis

User metadata
Rank Member
Rank
Member

@neonumber6 you're welcome! I am not sure what Linux version you're going to use, but if it requires a minimum of 128MB of RAM, a 486DX2 66MHz is probably going to be a more joyful experience for you anyways. I hope that will fix the problem for you. 😀

Mimi: AMD K6-2/266, S3 Trio64, Diamond Monster 3D II, Sound Blaster CT2800, 32MB RAM
Satellite 220CS: Pentium 133, SVGA DSTN, Sound Blaster Pro, 64MB RAM
Contura 420CX: 486DX4 75, VGA TFT, Roland Serial MIDI, 16MB RAM

Reply 6 of 13, by neonumber6

User metadata
Rank Newbie
Rank
Newbie
Strahssis wrote on 2022-05-10, 17:51:

@neonumber6 you're welcome! I am not sure what Linux version you're going to use, but if it requires a minimum of 128MB of RAM, a 486DX2 66MHz is probably going to be a more joyful experience for you anyways. I hope that will fix the problem for you. 😀

Slackware 15 -- text only w/o X-Windows.
Small home router project.

Requires about 72-80 MB of RAM.

Reply 7 of 13, by mkarcher

User metadata
Rank l33t
Rank
l33t

I had a similar problem with a board that is limited to 32MB of RAM. It didn't wire up tag bits to support RAM addresses above 32MB. Unfortunately, "RAM relocation" is enabled by default, and if you put 8*4MB into it, RAM extends up to 32MB + 256KB. The L2 cache mixes up 0MB + xxx and 32MB + xxx, so the system goes unstable as soon as the top 256KB of RAM is used. As a workaround on that board, I could either set a non-cachable area at 32MB, or disable RAM relocation.

I found a photo of your board at Ultimate Retro. I noticed the board has 10 cache sockets of 28 pins, which is unusual for 486 board. That photo only has 9 sockets fitted with a chip. This is the usual configuration: 2 interleaved banks of 4 cache chips each, and a ninth one for the tag bits. And that might be the catch: The more RAM you have, the more tag bits you need to store. I'd guess your board needs the tenth cache chip socket fitted with a cache chip to not mix up addresses when you install more than 32MB RAM. If your board also has M10 empty, try putting a tenth cache chip there. It might very well fix your issue.

Reply 8 of 13, by neonumber6

User metadata
Rank Newbie
Rank
Newbie
mkarcher wrote on 2022-05-10, 18:05:

I had a similar problem with a board that is limited to 32MB of RAM. It didn't wire up tag bits to support RAM addresses above 32MB. Unfortunately, "RAM relocation" is enabled by default, and if you put 8*4MB into it, RAM extends up to 32MB + 256KB. The L2 cache mixes up 0MB + xxx and 32MB + xxx, so the system goes unstable as soon as the top 256KB of RAM is used. As a workaround on that board, I could either set a non-cachable area at 32MB, or disable RAM relocation.

Are "disable RAM relocation" and "non-cachable area" options set in BIOS setup?
If so right now the Dallas RTC chip (with an embedded battery) is dead preventing me from saving any settings.
I got myself a new chip but since the old one is soldered into the motherboard I'll have to learn some desoldering + resoldering skills (and using a DIN socket would be wiser).

Maybe it'll be worth it to replace that Dallas RTC chip first to allow properly setting those BIOS options to properly complete memory tests?

mkarcher wrote on 2022-05-10, 18:05:

I found a photo of your board at Ultimate Retro. I noticed the board has 10 cache sockets of 28 pins, which is unusual for 486 board. That photo only has 9 sockets fitted with a chip. This is the usual configuration: 2 interleaved banks of 4 cache chips each, and a ninth one for the tag bits. And that might be the catch: The more RAM you have, the more tag bits you need to store. I'd guess your board needs the tenth cache chip socket fitted with a cache chip to not mix up addresses when you install more than 32MB RAM. If your board also has M10 empty, try putting a tenth cache chip there. It might very well fix your issue.

Hmm... in my case I *already* have all 10 cache sockets filled up right now by 32Kx8 SRAM chips... that is what the stason manual requires to do for 256 KB cache.

Here's a picture of the board (still with the 486 DX 33 MHz CPU) currently testing 8 x 4MB modules (32 MB) of RAM successfully with memtest-86 V3.0 (that is the maximum amount I'm currently able to test successfully, without being able to modify BIOS settings and with the current AMIBIOS 1.02 at least):
download/file.php?mode=view&id=137007

Attachments

Reply 9 of 13, by mkarcher

User metadata
Rank l33t
Rank
l33t

"Memory relocation" and "Non-cacheable area" are indeed options to be set in the Advanced Chipset Setup of most 486 mainboard BIOSes.

If your BIOS isn't extremely uncooperative (but maybe that AMI BIOS is), you should be able to save settings while the system is powered on. The Dallas chip doen't lose settings as long as +5V from the mainboard is present. OTOH, the Dallas chip does report whether the internal battery is flat even if +5V is present. An uncooperative BIOS will automatically revert to default settings if it sees the "battery flat" bit, even if the checksum is valid.

I checked the cache equipping instructions, and it indicates the tenth chip is needed for "proper" write-back cache support, to supply the bit indicating whether the cache contains updated data that needs to be written to RAM later. This bit is called "alter" or "dirty". So if the documentation is correct in that the second chip only supplies the "dirty" bit, it won't expand the cacheable area. At 256KB cache and the dirty bit stored in an extra chip, you should get 64MB cacheable area out-of-the-box.

The issue on your board might be caused by a broken trace, either a broken data trace to the tag RAM or a broken trace to the A11 trace to the RAM sockets. Can you take a photo of memtest failing on 64MB? You can make memtest pause on errors by pressing "enter" IIRC, so you can take a photo of the first error messages.

Reply 10 of 13, by neonumber6

User metadata
Rank Newbie
Rank
Newbie

Alright... finally had some time to try the following:

1) I've replaced the Intel 486 DX 33 MHz CPU with an AMD 486 DX2 66 MHz (and updated the relevant jumpers of course) :

20220513_142921.jpg

No change -- still getting errors when testing any amount of RAM over 32 MB -- so definitely not a CPU limitation in my understanding.

2) No BIOS update attempted yet (and it does appear that the board I have would currently be running the latest firmware anyway).

3) This is all the default advanced options that I can see:

20220513_142947.jpg

4) It does appear that I have an uncooperative BIOS -- if I try to change any settings the machine proceeds with a warm reboot and afterwards I get a low battery error which sends me back to the BIOS setup with all options set back to defaults....

20220513_143049.jpg

5) I've noticed that if I set a the "shadow" options to DISABLED and the "256 KB relocate" option to ENABLED I see some random characters on the main BIOS setup screen...

20220513_143300.jpg

6) Still if I try to pursue further WITHOUT saving BIOS settings (assuming they remain in effect while machine is on) I get some strange hang just before the boot process when "256 KB relocate" is ENABLED:

20220513_143134.jpg

(Will continue in a separate post so I can attach more files...)

Attachments

Reply 11 of 13, by neonumber6

User metadata
Rank Newbie
Rank
Newbie

7) If I disable all these "shadow" options as well as the "256 KB relocate" option the boot process works -- this allows me into MemTest86 3.0 which displays errors very early on:

20220513_143403.jpg
20220513_143541.jpg

This is with 64 MB NON-PARITY memory installed (as seen in the previous post).
Should I retry with 64 MB PARITY memory?

I'm starting to think that I might want to fix the Dallas RTC chip first and foremost to get the annoying bootup battery error message to go away and options to be saved...

Any thoughts?

Attachments

Reply 12 of 13, by mkarcher

User metadata
Rank l33t
Rank
l33t

The errors in memtest don't look like an addressing or cachability problem - they look like plain bad RAM. Are you sure the 16MB SIMMs are good? If they are good, you should check whether they have EDO chips on them. Most 486 boards of that time don't like EDO RAM at all, so maybe the RAM isn't bad but just incompatible with that board.

To find out whether the 16MB SIMMs are EDO (yes, there are EDO 30-pin SIMMs, although they never took of on PC mainboards), you could try to find a data sheet for the individual chips on the 16MB modules.

Reply 13 of 13, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

32MB is still lot for a 486 and for performance level, cannot make this any better with more ram due to CPU being bottlenecked.

Cheers,

Great Northern aka Canada.