VOGONS


First post, by NScaleTransitModels

User metadata
Rank Newbie
Rank
Newbie

Hello,
New member here but I've been messing around with old hardware for a loooong time, mostly 486s. I just finished cleaning/ upgrading an AST Advantage! Pro 486SX/25 and have been looking into 386 systems to diversity my collection and also as more of a challenge. I bought three new systems, and also brought out my old 486DLC clone. The first one I bought, a 386DX-25 clone, installed Windows 98 (yes, 98 😂) without any problems, but only supported 1mb SIMMs for a total of 8mb memory. My DLC machine also runs 98 fine with 16mb memory.

Since 8mb on a 386 clearly wasn't gonna be enough for me (lol), I bought two more 386 systems that support 4mb SIMMs (32mb total). However, I started getting strange memory errors as soon as I tried to install Windows (95 and 98) on them. Here's what went down:

1. 386DX-33 clone with a HOT-307 motherboard, OPTI 82C382 chipset, and 16mb (4x4mb 9-chip non-parity SIMMs): the Windows 98 setup ran fine until the first restart. After that, every time it tried to detect devices, it would crash with a blue screen of death. The BSoD is different every time, but is always a "fatal exception 0D" or "0E" with some memory address followed by a "VXD". I tried a different set of SIMMs (also 9-chip) and the install was able to finish, but upon restarting, would usually get a similar BSOD within seconds of loading the desktop. I've attached a picture of such a BSoD. Sometimes it would crash w/o errors and soft-reset. A couple times it didn't crash at all, and I could even load up my games. I also tried some brand new 3-chip SIMMs from Ebay, but it would freeze as soon as the mouse showed up.

2. PC Chips M326 motherboard, SARC chipset: I tried both a 386DX-40 and 486DLC-40. At first, it was with 8x 4mb 9-chip SIMMs (32mb), then 4x of those SIMMs (16mb). Same issues as above when installing 98. I then tried 95 with new 3-chip 4mb SIMMs, but no matter the combination, I would get "general protection fault" right after the "welcome to setup" screen. I then went back to 9-chip modules and the first phase of setup completed fine, but like with 98, would again blue screen at "setting up devices". I finally gave in and went back to 3-chip 1mb SIMMs, and the install finally completed (slowly).

At this point I feel like I've tried everything. I searched everywhere (including VOGONS) but nobody seems to be having issues across different memory modules (and especially not new ones). Or with 386s in particular. I'm pulling my hair out because I've never encountered anything like this on a 486 system. Oh, and I got my FX-3000 board today (combo 386/486) to try a newer chipset but it was DOA. 😕😤

Did I miss something obvious? Are 30-pin SIMMs just that unreliable? Did I just get unlucky and end up with 2 bad mobos? Or do neither of their chipsets have stable support for 4mb SIMMs? (Both support 32mb according to stason.org) Apologies for the long post but wanted to make sure I didn't leave out anything obvious.

Attachments

Builds:

  • ECS FX-3000; 386DX-40@50; ET4000AX, ISA 1mb
  • Acer VI9; 486DLC-40; Mach32, VLB 2mb
  • Chicony CH-471A; CX486s-40; Mach32, VLB 2mb
  • Gateway 2000 P5-60; Pentium-60@66; S3 928, PCI 3mb
  • DTK PKM-0033S; AM5x86-133@160

Reply 1 of 33, by Caluser2000

User metadata
Rank l33t
Rank
l33t

Bad ram.

There's a glitch in the matrix.
A founding member of the 286 appreciation society.
Apparently 32-bit is dead and nobody likes P4s.
Of course, as always, I'm open to correction...😉

Reply 2 of 33, by NScaleTransitModels

User metadata
Rank Newbie
Rank
Newbie
Caluser2000 wrote on 2020-06-19, 04:18:

Bad ram.

I tried the 9-chip SIMMs in another system (the 486DLC clone) and they work fine. In fact they were originally from that system. As for the 3-chip 4mb's, they are brand new from Ebay: https://www.ebay.com/itm/4MB-30pin-FAST-PAGE- … in/153940233188
so I feel like it's some kind of incompatibility with 3-chip design rather than defective sticks... I've heard of some members having trouble with 3-chip's in older 386 boards, but both of my boards are late-model 386 boards.

Builds:

  • ECS FX-3000; 386DX-40@50; ET4000AX, ISA 1mb
  • Acer VI9; 486DLC-40; Mach32, VLB 2mb
  • Chicony CH-471A; CX486s-40; Mach32, VLB 2mb
  • Gateway 2000 P5-60; Pentium-60@66; S3 928, PCI 3mb
  • DTK PKM-0033S; AM5x86-133@160

Reply 3 of 33, by Caluser2000

User metadata
Rank l33t
Rank
l33t

So ram not suitable for your system=Bad ram.

There's a glitch in the matrix.
A founding member of the 286 appreciation society.
Apparently 32-bit is dead and nobody likes P4s.
Of course, as always, I'm open to correction...😉

Reply 4 of 33, by kalohimal

User metadata
Rank Member
Rank
Member

30 pins SIMMs have speed grades. If the board requires -70 RAMs and you put in -80 or slower, then it is possible that they won't be able to keep up. To ensure the RAMs are ok, I usually test and label all my RAMs using this program: memtest86+. Oh and there are 2 versions, memtest86 is the commercial one and memtest86+ is the free one.

Slow down your CPU with CPUSPD for DOS retro gaming.

Reply 5 of 33, by mkarcher

User metadata
Rank l33t
Rank
l33t
NScaleTransitModels wrote on 2020-06-19, 03:34:

Since 8mb on a 386 clearly wasn't gonna be enough for me (lol), I bought two more 386 systems that support 4mb SIMMs (32mb total). However, I started getting strange memory errors as soon as I tried to install Windows (95 and 98) on them. Here's what went down:

[snipped typical symptoms of memory problems]

Did I miss something obvious? Are 30-pin SIMMs just that unreliable? Did I just get unlucky and end up with 2 bad mobos? Or do neither of their chipsets have stable support for 4mb SIMMs? (Both support 32mb according to stason.org) Apologies for the long post but wanted to make sure I didn't leave out anything obvious.

4MB SIMMs are generally supported since late 286 chipsets (i.e. on many 386SX systems and virtually all 386DX/486 systems,). So *in theory* your 4MB modules should work in the mainboards. Furthermore, 9-chip modules and *proper* 3-chip modules should behave mostly the same. While there are significant differences in the behaviour of x1 chips (on 9-chip modules) and x4 chips (on 3-chip modules), these SIMM interface is designed in a way that only the common behaviour of x4 and x1 chips is used *for the 8 standard data lines*. The parity chip has separate "data in" and "data out" lines, and a conforming 30-pin SIMM *cannot* be built with a x4 chip containing the parity. So proper 3-chip modules use two x4 chips for the 8 data bits and a dedicated x1 chip for the parity bit. The modules you linked are proper modules with exactly that pinout. (But there is cheap crap built from 3 factory-rejected x4 chips to pick 9 out of the 12 available bits, they happen to "work" in many boards, but the x4 chips were rejected for a reason at the factory...)

There is one important difference between 1MB and 4MB modules though, and that is refresh. Every *row* of the module needs to be refreshed often enough, and the classic RAM requires the mainboard to tell the row number to the RAM chip. 4MB modules have twice as many rows as 1MB modules (in general), so the number of the row to be refreshed needs to have an extra bit. If the mainboard does not provide enough row number bits for refresh, your symptoms are explainable. There is something to the rescue, though. "Modern" RAM chips (and all 4MB SIMM chips are modern in that regard) have their own built-in refresh row counter, and if you don't use the classic refresh protocol ("RAS-only refresh"), but the newer one ("CAS before RAS refresh"), refresh works independent of the mainboard. If you happen to have an option like "CAS-before-RAS refresh" or "hidden refresh" in the BIOS setup, try enabling it. *If you have an option called slow refresh, do disable it*. Slow refresh is meant for low-power modules that don't need to be refreshed that often. I have seen at least one 386SX board that enabled slow refresh *by default*. Especially if there are many rows (4MB modules...), a reduced rate of refreshing rows (one at a time) can be problematic.

Reply 6 of 33, by mkarcher

User metadata
Rank l33t
Rank
l33t
kalohimal wrote on 2020-06-19, 07:05:

30 pins SIMMs have speed grades. If the board requires -70 RAMs and you put in -80 or slower, then it is possible that they won't be able to keep up. To ensure the RAMs are ok, I usually test and label all my RAMs using this program: memtest86+. Oh and there are 2 versions, memtest86 is the commercial one and memtest86+ is the free one.

And if you want to run memtest on a 386 board, make sure to get a version that does not try to measure clock frequency or memory throughput using the time-stamp counter, i.e. a very old version. memtest86 (without the plus) is worse in that regard than memtest86+, but IIRC either of those tools need to be old enough. Sadly, I don't recall what version of memtest86+ works fine on 486 (and 386) processors, but my intution says that you should start trying with a version around 2003 (or earlier). The time-stamp counter has been introduced with the Pentium Processor.

Reply 7 of 33, by konc

User metadata
Rank l33t
Rank
l33t

You didn't mention memory timing in BIOS so I'll add this parameter too: run memtest86+ and try safer/lowest values (=add wait states) for both DRAM and SRAM until it's stable

mkarcher wrote on 2020-06-19, 10:42:

Sadly, I don't recall what version of memtest86+ works fine on 486 (and 386) processors, but my intution says that you should start trying with a version around 2003 (or earlier).

The last version of memtest86+ that runs on a 386 is 4.10 (2010) and there is no version prior 2.0 (2008) to download

Last edited by Stiletto on 2020-06-19, 16:04. Edited 1 time in total.

Reply 8 of 33, by kalohimal

User metadata
Rank Member
Rank
Member
konc wrote on 2020-06-19, 11:11:

You didn't mention memory timing in BIOS so I'll add this parameter too: run memtest86+ and try safer/lowest values (=add wait states) for both DRAM and SRAM until it's stable

The last version of memtest86+ that runs on a 386 is 4.10 (2010) and there is no version prior 2.0 (2008) to download

Oh yes, these SIMMs do not have SPD.

Slow down your CPU with CPUSPD for DOS retro gaming.

Reply 9 of 33, by NScaleTransitModels

User metadata
Rank Newbie
Rank
Newbie

Thanks for all the excellent responses.
I am currently testing with 4x 9-chip modules in the HOT-307 system. If they work I'll try the 3-chip again.
1. I double checked the SIMM speeds. All four of them are marked 70ns. The 3-chip are all 60ns... should be good for consumer grade boards, right?
2. I ran memtest86+ (can confirm 4.10 is the last version that works on 386's) overnight and completed 6 passes with 0 errors.
3. Sounds like row refresh could be the issue. The BIOS on this machine has options for both "hidden refresh" and "slow refresh", both of which defaulted to disabled. (The CMOS battery leaked so I removed it and cleaned w/ vinegar) I'll enable "hidden refresh" and report back. 🙂
4. The BIOS also defaulted to 0 W/S. I'll see what happens if it's switched to 1 W/S.

Builds:

  • ECS FX-3000; 386DX-40@50; ET4000AX, ISA 1mb
  • Acer VI9; 486DLC-40; Mach32, VLB 2mb
  • Chicony CH-471A; CX486s-40; Mach32, VLB 2mb
  • Gateway 2000 P5-60; Pentium-60@66; S3 928, PCI 3mb
  • DTK PKM-0033S; AM5x86-133@160

Reply 10 of 33, by mkarcher

User metadata
Rank l33t
Rank
l33t
NScaleTransitModels wrote on 2020-06-19, 20:08:

2. I ran memtest86+ (can confirm 4.10 is the last version that works on 386's) overnight and completed 6 passes with 0 errors.
3. Sounds like row refresh could be the issue. The BIOS on this machine has options for both "hidden refresh" and "slow refresh", both of which defaulted to disabled. (The CMOS battery leaked so I

RAM problems in real world use with no problems in memtest86+ can indeed be a symptom of refresh problems. Memtest seems to access all the rows often enough that they get refreshed just from being accessed. In real-world use, some rows might be left untouched for seconds or even minutes (normal refresh timing is refreshing all rows multiple times a second) and lose their contents.

I got these symptoms with the crap modules I mentioned already that seem to be made of factory-rejected memory chips. In a slow-down-the-refresh test, these crap modules failed slightly earlier than good modules (still way over the typical refresh timing). This test confirmed that they *should* keep their data just idling with slowed-down refresh, so I guess the problems might be caused by rowhammer-like access patterns. One of the many things I want to play with when I get around to. I don't think this is the issue with your modules, though, because I don't think that all your 4MB simms are crap modules, and because they work in other systems.

Reply 11 of 33, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

These is the very reason I carefully buy memory modules that is manufactured by OEM brands like IBM (72 pin simms), Micron, Samsung etc.

The generic stuff can take a hike, Had issues with these all the time, even OEM brands (rarely). But one unusual stood out to me in my mind: Back in the day, Hyundai chips was failing in large number that I had to rebuild many 30 pin SIMMS with tested chips and test each chip individually on a DRAM tester by pushing them slightly to catch weaker ones and some just failed quickly even at 100ns, these was 80ns chips. Back then DRAM were expensive even 256K x 1 (1990 or so).

Cheers,

Great Northern aka Canada.

Reply 12 of 33, by NScaleTransitModels

User metadata
Rank Newbie
Rank
Newbie

So after quite a bit of testing:
1. 9-chip with "hidden refresh" enabled and 1 W/S (read and write): On the first attempt, I managed to get to the desktop, but it glitched and froze as soon as the welcome screen came up:

20200619_134057.jpg
Filename
20200619_134057.jpg
File size
1.08 MiB
Views
1139 views
File license
Public domain

The second attempt gave me a "windows protection error":

20200619_162902.jpg
Filename
20200619_162902.jpg
File size
1.52 MiB
Views
1139 views
File license
Public domain

Not stable.

2. 9-chip with "hidden refresh" enabled and 2 W/S: the first 2 attempts were fine, and I was able to do basic tasks and load Doom. However, the third attempt resulted in the welcome dialog crashing with a white box error, followed by a fatal exception:

20200619_163148.jpg
Filename
20200619_163148.jpg
File size
1.26 MiB
Views
1139 views
File license
Public domain
20200619_163202.jpg
Filename
20200619_163202.jpg
File size
890.61 KiB
Views
1139 views
File license
Public domain

The fourth attempt went straight to a fatal exception:

20200619_163435.jpg
Filename
20200619_163435.jpg
File size
1.81 MiB
Views
1139 views
File license
Public domain

Also not stable.

3. New 3-chip with hidden refresh enabled and 1 W/S: four attempts, each boot I was able to load Windows and Doom with no errors. Also was able to bring up the start menu and system properties with no problems. Appears to be stable.

So it appears the 9-chip SIMMs could be toast. It's kinda strange that they work on my other system (Biostar MB-1340AEQ), but that board could just have a higher tolerance. And plus I'd have to check the memory timing and refresh settings on that one. Anyways, the 3-chips seems to work, so I'm moving onto the PC Chips M326 with a Windows 95 install...

Builds:

  • ECS FX-3000; 386DX-40@50; ET4000AX, ISA 1mb
  • Acer VI9; 486DLC-40; Mach32, VLB 2mb
  • Chicony CH-471A; CX486s-40; Mach32, VLB 2mb
  • Gateway 2000 P5-60; Pentium-60@66; S3 928, PCI 3mb
  • DTK PKM-0033S; AM5x86-133@160

Reply 13 of 33, by NScaleTransitModels

User metadata
Rank Newbie
Rank
Newbie
pentiumspeed wrote on 2020-06-19, 23:25:

These is the very reason I carefully buy memory modules that is manufactured by OEM brands like IBM (72 pin simms), Micron, Samsung etc.

The generic stuff can take a hike, Had issues with these all the time, even OEM brands (rarely). But one unusual stood out to me in my mind: Back in the day, Hyundai chips was failing in large number that I had to rebuild many 30 pin SIMMS with tested chips and test each chip individually on a DRAM tester by pushing them slightly to catch weaker ones and some just failed quickly even at 100ns, these was 80ns chips. Back then DRAM were expensive even 256K x 1 (1990 or so).

Cheers,

I'm not entirely sure what brand my 9-chip SIMMs are... TI = Texas Instruments? Which I thought was a good brand? Could just be too old 🤣... Here's a picture:

20200619_163654.jpg
Filename
20200619_163654.jpg
File size
1.34 MiB
Views
1138 views
File license
Public domain

Builds:

  • ECS FX-3000; 386DX-40@50; ET4000AX, ISA 1mb
  • Acer VI9; 486DLC-40; Mach32, VLB 2mb
  • Chicony CH-471A; CX486s-40; Mach32, VLB 2mb
  • Gateway 2000 P5-60; Pentium-60@66; S3 928, PCI 3mb
  • DTK PKM-0033S; AM5x86-133@160

Reply 14 of 33, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

Correct, that is generic memory even it used TI memory chips at 70ns. But no label on any of these but even that, seems decent quality and not your problem, could be your generic boards are the issue? What kind of chipset are you using on these 2 386DX motherboards?

Even that, your motherboards is not up to quality standards. Provide photos of both boards please.

Cheers,

Great Northern aka Canada.

Reply 15 of 33, by Horun

User metadata
Rank l33t++
Rank
l33t++
pentiumspeed wrote on 2020-06-20, 00:51:

Correct, that is generic memory even it used TI memory chips at 70ns. But no label on any of these but even that, seems decent quality and not your problem, could be your generic boards are the issue? What kind of chipset are you using on these 2 386DX motherboards?

Even that, your motherboards is not up to quality standards. Provide photos of both boards please.

Cheers,

Agree ! Those look like good 70nS SIMMS, should be compatible with most lower speed 386 and early 486. Your 386 boards may be very specific when it comes to 4Mb simms, some older chipset 386 were very quirky with them and required specific manufactured models and would not support them unless 60nS, specially at 40Mhz CPU speed. I did not look up your two listed boards but do know that AST were very picking on ram.

Hate posting a reply and then have to edit it because it made no sense 😁 First computer was an IBM 3270 workstation with CGA monitor. https://archive.org/details/@horun

Reply 16 of 33, by NScaleTransitModels

User metadata
Rank Newbie
Rank
Newbie
pentiumspeed wrote on 2020-06-20, 00:51:

Correct, that is generic memory even it used TI memory chips at 70ns. But no label on any of these but even that, seems decent quality and not your problem, could be your generic boards are the issue? What kind of chipset are you using on these 2 386DX motherboards?

Even that, your motherboards is not up to quality standards. Provide photos of both boards please.

Cheers,

Here are the two boards I was having problems with:
1. Shuttle HOT-307, OPTI 82C392 chipset, AM386DX-33

hot-307.jpg
Filename
hot-307.jpg
File size
412.17 KiB
Views
1120 views
File license
Public domain

https://stason.org/TULARC/pc/motherboards/S/S … 86-HOT-307.html

2. PC Chips M326, SARC RC4018A4 chipset (never heard of SARC before), i386DX-33 running at 40 mhz. I do not have a 40mhz-rated CPU at the moment, and the board will not boot w/ 33mhz jumper settings.

m326 ebay.jpg
Filename
m326 ebay.jpg
File size
513.21 KiB
Views
1120 views
File license
Public domain

https://stason.org/TULARC/pc/motherboards/U/U … -326-VER-2.html

Builds:

  • ECS FX-3000; 386DX-40@50; ET4000AX, ISA 1mb
  • Acer VI9; 486DLC-40; Mach32, VLB 2mb
  • Chicony CH-471A; CX486s-40; Mach32, VLB 2mb
  • Gateway 2000 P5-60; Pentium-60@66; S3 928, PCI 3mb
  • DTK PKM-0033S; AM5x86-133@160

Reply 17 of 33, by Horun

User metadata
Rank l33t++
Rank
l33t++

Those are pictures of your exact boards ? The PCChips board you cannot get 33Mhz to run has an issue, it should run at 33mhz. Also the chip near X2 does not look right on that board. Why is it silver ? Should be a black 8 pin clock DIP afaik. The Shuttle board is in a computer, you should remove it and "breadboard" test it. Added: I see 3 chip SIMMS on the one board, does the board run stable on those and can you take closeup pic of those SIMMS ?

Hate posting a reply and then have to edit it because it made no sense 😁 First computer was an IBM 3270 workstation with CGA monitor. https://archive.org/details/@horun

Reply 18 of 33, by NScaleTransitModels

User metadata
Rank Newbie
Rank
Newbie
Horun wrote on 2020-06-20, 01:23:

Agree ! Those look like good 70nS SIMMS, should be compatible with most lower speed 386 and early 486. Your 386 boards may be very specific when it comes to 4Mb simms, some older chipset 386 were very quirky with them and required specific manufactured models and would not support them unless 60nS, specially at 40Mhz CPU speed. I did not look up your two listed boards but do know that AST were very picking on ram.

Hmm... it could be that 70ns won't cut it for these two boards, even though I believe both are newer chipsets: the 82C392 even shows up on some 486 boards. Anyways, I have some new 9-chip SIMMs on the way, and those are rated at 60ns, so we'll see how they hold up.

Also an update on the M326 board: I swapped back in the 4mb 3-chip SIMMs (60ns), and I'm trying to reinstall 95. I checked the BIOS and there are way less options: none for "hidden refresh" or "CAS-before-RAS". Only W/S, which I set to 1 for now. (default was 2)
Last time I tried a 95 install with the same setup, I was getting a "general protection fault at module USER.EXE" error. I didn't get a shot, but it looked something like this:
1440457032.th.22699.png
Except it was right after the "preparing setup wizard" loading bar. This time, I'm having a different problem altogether. I'm not even sure if it has to do with the memory, or my graphics card (Cirrus GD5422), or the CPU running above spec. The setup completed without any crashes or freezes, but I'm getting strange graphical glitches upon reaching the desktop. Restarted and same deal. I even tried another graphics card, a GD5429, with similar graphics issues.

graphics glitch.png
Filename
graphics glitch.png
File size
517.82 KiB
Views
1108 views
File license
Public domain

I'm starting to think this board just has a bad chipset 😕

Builds:

  • ECS FX-3000; 386DX-40@50; ET4000AX, ISA 1mb
  • Acer VI9; 486DLC-40; Mach32, VLB 2mb
  • Chicony CH-471A; CX486s-40; Mach32, VLB 2mb
  • Gateway 2000 P5-60; Pentium-60@66; S3 928, PCI 3mb
  • DTK PKM-0033S; AM5x86-133@160

Reply 19 of 33, by NScaleTransitModels

User metadata
Rank Newbie
Rank
Newbie
Horun wrote on 2020-06-20, 02:05:

Those are pictures of your exact boards ? The PCChips board you cannot get 33Mhz to run has an issue, it should run at 33mhz. Also the chip near X2 does not look right on that board. Why is it silver ? Should be a black 8 pin clock DIP afaik. The Shuttle board is in a computer, you should remove it and "breadboard" test it. Added: I see 3 chip SIMMS on the one board, does the board run stable on those and can you take closeup pic of those SIMMS ?

Yes, the pictures are from Ebay but pulled from the exact listings I purchased from. For the PCChips board, the chip near X2 is a black clock-gen chip, but has a silver mini-heatsink on top of it. I'm not sure why it won't run a 33mhz, but I'll try the jumper settings again in a bit.

As for the Shuttle board, the picture shows the 3-chip 1mb SIMMs it originally came with, which ran stable from the beginning. Here's what the SIMMs look like:

1mb 3 chip.png
Filename
1mb 3 chip.png
File size
1.84 MiB
Views
1106 views
File license
Public domain

Currently, the Shuttle board is running stable with 3-chip 4mb SIMMs, with hidden refresh enabled and 1 W/S, as in one of my previous posts:

4mb 3 chip.png
Filename
4mb 3 chip.png
File size
1.86 MiB
Views
1106 views
File license
Public domain

But I'm still not entirely sure why it won't work with the 9-chip 4mb SIMMs.

Builds:

  • ECS FX-3000; 386DX-40@50; ET4000AX, ISA 1mb
  • Acer VI9; 486DLC-40; Mach32, VLB 2mb
  • Chicony CH-471A; CX486s-40; Mach32, VLB 2mb
  • Gateway 2000 P5-60; Pentium-60@66; S3 928, PCI 3mb
  • DTK PKM-0033S; AM5x86-133@160