VOGONS


First post, by demiurge

User metadata
Rank Member
Rank
Member

Given that parity memory should work fine as non-parity:

dionb wrote on 2023-08-01, 23:52:

parity or ECC memory will generally happily run as non-parity/ECC, but you will not have the benefits of error checking (parity can identify single-bit errors, ECC double-bit errors) if you do. Controllers that don't support parity generally don't care if memory supports it or not. Conversely, some memory controllers not only support but also require parity/ECC memory.

I have a weird situation where only the first slot is working. I accidently bought parity EDO and now only bank 0 is works. Is it the parity issue or could it be the motherboard?

The motherboard is a ABit AB-PG5 and the memory is in the picture.

Is this something someone has seen before?

Reply 1 of 15, by dionb

User metadata
Rank l33t++
Rank
l33t++

Hang on... you're using a couple of terms interchangeably here that aren't synonymous.

This is a board for Pentium CPUs with a SiS 5511 chipset. That means that one (64b wide) bank generally consists of two (32b wide) 72p SIMMs that should be installed in identical pairs in adjacent slots. So if you say the first bank works, it implies that the first two slots work. But you then say that only the first slot works, which sounds contradictory.
In the case of the SiS 5511 the picture is complicated because the memory controller lets you run with a single 72p SIMM, albeit at half the normal bandwidth.

So to clarify:
- how many SIMMs of the type you photographed (32MB parity, with 18 4Mx4 chips) do you have?
- when you say "the first bank works", which SIMMs are in which slots?

In any event, so long as SIMMs are identical, the presence or absence of parity support on the SIMMs will not cause issues. If some SIMMs support parity and others do not, you should disable parity in BIOS. Then all will work the same. If you have parity enabled in BIOS but some SIMMs do not support parity, those SIMMs will not work so long as parity remains enabled in BIOS.

Reply 2 of 15, by demiurge

User metadata
Rank Member
Rank
Member
dionb wrote on 2024-02-25, 12:49:

Hang on... you're using a couple of terms interchangeably here that aren't synonymous.

Yeah the bank/slot are used interchangeably on the The Micro House Encyclopedia of Main Boards sheet. So I did it incorrectly. It is only the first slot working. One SIMM works fine and then all others don't. Each SIMM functions in the first slot just fine so it isn't a bad SIMM. The only configuration that works is with one SIMM in the slot 1.

dionb wrote on 2024-02-25, 12:49:

In any event, so long as SIMMs are identical, the presence or absence of parity support on the SIMMs will not cause issues. If some SIMMs support parity and others do not, you should disable parity in BIOS. Then all will work the same. If you have parity enabled in BIOS but some SIMMs do not support parity, those SIMMs will not work so long as parity remains enabled in BIOS.

I can't find any option with respect to parity in the BIOS. That is one of the problems. The Micro House Encyclopedia of Main Boards sheet only lists 8M x 32 and these are 8M x 36. I thought that was just the parity speaking.

I ordered some 8M x 32 and hopefully that will work. But really do I have a bad slot or something? I don't know.

Reply 3 of 15, by Paar

User metadata
Rank Member
Rank
Member

It's probably not a bad slot but damaged component/trace. Check pins on the SiS chipset, if some are not bent and are not touching.

Reply 4 of 15, by dionb

User metadata
Rank l33t++
Rank
l33t++
demiurge wrote on 2024-02-25, 15:59:

[...]

Yeah the bank/slot are used interchangeably on the The Micro House Encyclopedia of Main Boards sheet. So I did it incorrectly. It is only the first slot working. One SIMM works fine and then all others don't. Each SIMM functions in the first slot just fine so it isn't a bad SIMM. The only configuration that works is with one SIMM in the slot 1.

But are you just using a single SIMM or two identical ones?

If it's a single one, this behaviour is to be expected- from the datasheet:

Half populated bank is also supported for bank 0

So you can only use a single SIMM in the first bank, assumedly only in slot 1.

If it's two or more, then please indicate what SIMMs you are using in which combination.

Reply 5 of 15, by demiurge

User metadata
Rank Member
Rank
Member
dionb wrote on 2024-02-25, 20:27:
But are you just using a single SIMM or two identical ones? […]
Show full quote
demiurge wrote on 2024-02-25, 15:59:

[...]

Yeah the bank/slot are used interchangeably on the The Micro House Encyclopedia of Main Boards sheet. So I did it incorrectly. It is only the first slot working. One SIMM works fine and then all others don't. Each SIMM functions in the first slot just fine so it isn't a bad SIMM. The only configuration that works is with one SIMM in the slot 1.

But are you just using a single SIMM or two identical ones?

If it's a single one, this behaviour is to be expected- from the datasheet:

Half populated bank is also supported for bank 0

So you can only use a single SIMM in the first bank, assumedly only in slot 1.

If it's two or more, then please indicate what SIMMs you are using in which combination.

Thanks for your help and patience, but the following configurations still only produce the same memory outcome. I also tried the same memory in slots 1 and 3 instead of slots 1 and 2.

Reply 6 of 15, by Horun

User metadata
Rank l33t++
Rank
l33t++

I do not think the board likes those simms. I bet if you tried some 16MB 8 chip EDO matching they would work....fwiw: I do not have any EDO parity in all my bags of simms but do have some EDO ECC simms...
That may not be the issue and could be the board but I still would try diff simms.
added: Another interesting issue is the part number on sticker matches Micron for FPM ECC, not EDO. see: https://digchip.com/datasheets/parts/datashee … D836-6X-pdf.php

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. Stuff: https://archive.org/details/@horun

Reply 7 of 15, by dionb

User metadata
Rank l33t++
Rank
l33t++

What happens with those SIMMs in slots 3 and 4?

Also, have you tried both SIMMs solo in slot 1? Or tested both on another board? If not the second SIMM could just be dead...

Reply 8 of 15, by demiurge

User metadata
Rank Member
Rank
Member
Horun wrote on 2024-02-26, 04:42:

I do not think the board likes those simms. I bet if you tried some 16MB 8 chip EDO matching they would work....fwiw: I do not have any EDO parity in all my bags of simms but do have some EDO ECC simms...
That may not be the issue and could be the board but I still would try diff simms.
added: Another interesting issue is the part number on sticker matches Micron for FPM ECC, not EDO. see: https://digchip.com/datasheets/parts/datashee … D836-6X-pdf.php

I never thought to look this up, I supposedly bought EDO parity but never thought I would get FPM ECC.

Yeah I have ordered different SIMMs to try. I don't have any others. I do have one stick of 4M x 32, and it doesn't work with the 8M x 32 but the manual specifically says this configuration isn't supported.

dionb wrote on 2024-02-26, 08:16:

What happens with those SIMMs in slots 3 and 4?

Also, have you tried both SIMMs solo in slot 1? Or tested both on another board? If not the second SIMM could just be dead...

I haven't tried all 4! ways that are possible but I have tried a lot of them. There are many permutations that I was just swapping around. I have tried all 4 SIMMs as the only SIMM in socket 1 and they work on their own.

Reply 9 of 15, by dionb

User metadata
Rank l33t++
Rank
l33t++
demiurge wrote on 2024-02-27, 00:10:

[...]

I haven't tried all 4! ways that are possible but I have tried a lot of them. There are many permutations that I was just swapping around. I have tried all 4 SIMMs as the only SIMM in socket 1 and they work on their own.

OK, then we are sure the SIMMs all work, and do so in combination with this board, at least in solo/half-bank configuration. That also means that they can be considered interchangeable in terms of configuration, so you don't have 4! permutations, just three:

1) one SIMM in slot 1 - the test only possible because of oddball feature of this chipset. We know now that works with all the SIMMs.
2) two identical SIMMs in bank 0, i.e. slots 1 and 2. That appears not to work.
3) two identical SIMMs in bank 1, i.e. slots 3 and 4. That's the test that should indicate whether the problem is just a broken slot 2, or a broader compatibility issue (or, black swan, damage to both slot 2 and either 3 and/or 4 as well)

Reply 10 of 15, by demiurge

User metadata
Rank Member
Rank
Member
dionb wrote on 2024-02-27, 10:50:

3) two identical SIMMs in bank 1, i.e. slots 3 and 4. That's the test that should indicate whether the problem is just a broken slot 2, or a broader compatibility issue (or, black swan, damage to both slot 2 and either 3 and/or 4 as well)

It does not work either. Looks to be a compatibility issue.

Reply 11 of 15, by Horun

User metadata
Rank l33t++
Rank
l33t++

Yep, think your simms are specific for a server type board (checked and appear they are EDO plus ECC/Parity) and will not work as expected in a standard board.
Remember getting some dimms like that once, did not work proper in standard boards. Sorta along the lines of those simms with presence detect...in reverse..

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. Stuff: https://archive.org/details/@horun

Reply 12 of 15, by demiurge

User metadata
Rank Member
Rank
Member

Well the different SIMMs from China didn't work. They are clearly not ECC. Same problem of only first slot working.

I'm going to update the BIOS and hope it is a bug limiting it to 32 MB somehow.

Reply 13 of 15, by hasat

User metadata
Rank Newbie
Rank
Newbie

Also check, if you have correct BIOS for your board. I had a similar problem with 386 board (two banks, 8 RAM slots, only 4 worked) and problem was only in wrong BIOS. Board normally worked with this wrong BIOS, but only 4 RAM slots were functional. Some previous owner used wrong BIOS.

My thread here on Vogons about this Half of the memory slots not working - 386 board

Reply 14 of 15, by demiurge

User metadata
Rank Member
Rank
Member
hasat wrote on 2024-03-06, 21:23:

Also check, if you have correct BIOS for your board. I had a similar problem with 386 board (two banks, 8 RAM slots, only 4 worked) and problem was only in wrong BIOS. Board normally worked with this wrong BIOS, but only 4 RAM slots were functional. Some previous owner used wrong BIOS.

My thread here on Vogons about this Half of the memory slots not working - 386 board

I just flashed with the last available BIOS from Abit. It was a pain because AWDFLASH.EXE is crap. Use uniflash instead.

So, no, the BIOS that was in EEPROM was stock and the last official BIOS doesn't fix it.

Reply 15 of 15, by demiurge

User metadata
Rank Member
Rank
Member

Okay this is weird, like the machine is limited to 32 MB but there are no jumpers or settings.

One 4 MB in slot 1 and one 32 MB in slot 3 is 32 MB on boot.
One 32 MB in slot 1 and one 4 MB in slot 3 is 32 MB on boot.

This means neither slot is defective and the machine sees both banks.

I did find one bad trace going to a slot but when I connected the trace on boot, it didn't change anything.