In my experience these early SDRAM adopters are very picky about the sticks. A bit like 486 mobos and EDO memory vs FPM. Do you have any 8 chip SDRAMs? Do not limit yourself to older sticks, sometimes much newer ones will work without any issues while being mis-detected as lower capacity. For example I have a 133MHz 128MiB stick in a VIA Socket7 mobo, along K6-2 400MHz. That one is detected as 64MiB and passes all kinds of tests, including memtest86. Was cheap too.

majestyk wrote on 2024-07-22, 14:33:

This is a nice socket 7 mainboard with Intel´s 430VX chipset. […]
Show full quote

This is a nice socket 7 mainboard with Intel´s 430VX chipset.

The attachment AB-PR5_103a.JPG is no longer available

According to the manual is supports both EDO and SDRAM. In the 4 SIMM slots all FPM and EDO sticks are detected correctly.
When I populate the 2 168-pin SDRAM slots with two 64 MB EDO sticks, 128MB RAM (EDO) are detected and can be used.
As soon as I try any SDRAM, for example 32MB sticks with 16 chips, the system hangs at "C6 C1", just like when no RAM is present at all.
I also tried different BIOS versions.

Since 128 MB EDO work, I assume all the data- and address lines are present at the 168-pin slots. Are there any lines exclusively needed for SDRAM?

Did You set JP11 to 3,3V?

I wish I had this guy's luck - one of these DIMMs ought to work with my PCChips M520, but just try and find one. The M520 only has one RAM slot, so I doubt I'll ever be in a position to try SDRAM with that board.
Intel 430VX Triton II SIMM vs DIMM performance

VX boards are picky indeed, but in the past after some trial and error some sticks out of my stock would always work, sometimes the size was not detected correctly.

Jumper 11 is at the 3V position (2-3) all the time and there´s 3.3V at the Vcc pins of the SDRAM slots.

Speaking ot the M520 - I do have one on my desk at the moment and these sticks do work:

Both have 16 chips, the 32MB stick is detected correctly. the 256MB stick is detected as 64MB.

On the AB-PR5 there´s no reaction at all.

While I was thinking about signals that SDRAMs need and EDOs don´t, I came across the clock signals.
Pin 42 (clock 0) and pin 125 (clock 1) are connected to the clock generator. Pins 79 (clock 2) and 163 (clock 3) were "N/C" - for whatever reason.
All clock signals are needed by regular SDRAM sticks.

Next I had a look at the M520, where SDRAM is working and found two traces bridging pins 42 <-> 79 and 125 <-> 163.
So I added the two bridges at each SDRAM slot:

...and Bob´s your uncle:

For 128MB I used 2 x 256MB quite regular cheap CL3 sticks, just one quarter is being used by the system. And it´s a pity only 64MB are cacheable.

I think that´s a severe case of layout flaw.

majestyk wrote on 2024-07-23, 08:37:

Pin 42 (clock 0) and pin 125 (clock 1) are connected to the clock generator. Pins 79 (clock 2) and 163 (clock 3) were "N/C" - for whatever reason.
All clock signals are needed by regular SDRAM sticks.

Possibly low capacity SDRAMs might not be using the other clock signals, I would assume these are duplicated to make the routing on the memory module itself easier, and keep delays equal. Whoever routed, and then validated this mobo was using such sticks. They either didn't understand the datasheets or plain forgot to connect the other signals as well.

Again I have flashbacks of a 486 mobo I spent a lot of time on that has both 30-pin and 72-pin SIMM slots, and the RAS signals are shared between them, so only certain combinations of old and new sticks are possible. On 72-pin SIMMs there are 4 RAS signals, but the front side of the PCB uses only RAS0 and RAS2. The odd ones are for the back side, which is not even populated on most common 4M sticks. And in fact not all mobos properly support these odd signals, making dual-sided sticks unusable or working at half capacity.

Maybe this was done in a hurry or with special SDRAM sticks in mind.
Funny how this is a rare case of Abit failing while PC-Chips got it right 😉

Abit even released a follow up version "R2" of this board - EDO only.

majestyk wrote on 2024-07-23, 08:37:

While I was thinking about signals that SDRAMs need and EDOs don´t, I came across the clock signals. Pin 42 (clock 0) and pin 12 […]
Show full quote

While I was thinking about signals that SDRAMs need and EDOs don´t, I came across the clock signals.
Pin 42 (clock 0) and pin 125 (clock 1) are connected to the clock generator. Pins 79 (clock 2) and 163 (clock 3) were "N/C" - for whatever reason.
All clock signals are needed by regular SDRAM sticks.

Next I had a look at the M520, where SDRAM is working and found two traces bridging pins 42 <-> 79 and 125 <-> 163.
So I added the two bridges at each SDRAM slot:

The attachment ab_pr5_sdram_bridges.jpg is no longer available

...and Bob´s your uncle:

The attachment ab_pr5_sdram_128mb.jpg is no longer available

The attachment ab_pr5_speeds.jpg is no longer available

For 128MB I used 2 x 256MB quite regular cheap CL3 sticks, just one quarter is being used by the system. And it´s a pity only 64MB are cacheable.

I think that´s a severe case of layout flaw.

Nice!!!!

I'm reminded of how some early VX chipset motherboards had a USB header that did not work and the manual told you that it wasn't functional.

majestyk wrote on 2024-07-23, 05:48:

VX boards are picky indeed, but in the past after some trial and error some sticks out of my stock would always work, sometimes […]
Show full quote

VX boards are picky indeed, but in the past after some trial and error some sticks out of my stock would always work, sometimes the size was not detected correctly.

Jumper 11 is at the 3V position (2-3) all the time and there´s 3.3V at the Vcc pins of the SDRAM slots.

Speaking ot the M520 - I do have one on my desk at the moment and these sticks do work:

The attachment vx_detected.JPG is no longer available

Both have 16 chips, the 32MB stick is detected correctly. the 256MB stick is detected as 64MB.

On the AB-PR5 there´s no reaction at all.

This post inspired me to try out the SDRAM slot on my M520. At the same time, my inner monologue warned me that it was probably a bad idea as the computer was working just fine with 96 megabytes of EDO SIMMs.

I grabbed my bag of SDRAM and set to work. Being a PCChips board, the PCB is thin and it has warped over the years. The SDRAM slot is at the very top of the board so it's poorly supported, and the board's warp is in the direction that made the SDRAM slot slightly concave. Add in oxidized contacts, and getting a stick of RAM full inserted and detected was a real job. I have it in a tiny AT case so it's crowded - even getting to the memory voltage jumpers was difficult.

After lots of trial and error I was able to find a 256 meg stick that was detected as 64 megs and it would reliably POST with it. But getting into Windows was another matter. No matter what I tried, Windows 98 was not stable and I was never able to complete a pass of 3D 2000 without it locking up or crashing. So I gave up and put the EDO SIMMs back in. But now it still wasn't stable? I decided to reinstall Win98, but I was having no luck with that either. I was having difficulties formatting the hard drive with Super FDISK and other weirdness. At one point I thought I had it, and was ready to reinstall 3D 2000. I have a Word doc that has the serial numbers for the versions of 3D Mark that I commonly use so I can copy and paste it. I made it on my main computer with Office 2007 and saved it so that it would be compatible with older OS and it has always worked fine with Win98. I tried to open it and I got an "Illegal Operation" error. I installed Word 97, and only then was I able to open it. I then tried to open it with another Win98 machine, and it too gave me the "Illegal Operation" error now! I put the thumb drive into my main computer and opened it (no problem) and saved it again, but it still wouldn't open in a Win98 computer? I found another thumb drive that had a copy of this doc, and that one worked fine in any computer!

No matter what I tried, I couldn't make the system stable. I checked it with Memtest, and that passed fine. I have the M520 set up with a K6-3+ 400 @ 450 using an Evergreen Spectra adapter, a 256 megabyte PCI FX5200 video card, and an SSD on a Sil3114 controller. It occurred to me that these are some power hungry components for a very old AT PSU. I had pulled the cover off the PSU and inspected it when I bought this computer and nothing looked amiss, but that hardly meant it was in the clear now. I tried swapping out the power hungry video card for a one megabyte Trident video card, and the machine got much further along installing Windows, so that strengthened my suspicion that the antique PSU was failing. I swapped in a known good AT PSU (Again a PIA with the tiny AT case) and tried again to reinstall Windows and this time everything was back to normal.

Being a glutton for punishment, I wondered if the reason the SDRAM wouldn't work was actually that the power supply was failing? So I tried it again, the result was the same, and I got to reinstall Win98 again.

Repo Man11 wrote on 2024-07-29, 19:41:

This post inspired me to try out the SDRAM slot on my M520. At the same time, my inner monologue warned me that it was probably […]
Show full quote

majestyk wrote on 2024-07-23, 05:48:

VX boards are picky indeed, but in the past after some trial and error some sticks out of my stock would always work, sometimes […]
Show full quote

VX boards are picky indeed, but in the past after some trial and error some sticks out of my stock would always work, sometimes the size was not detected correctly.

Jumper 11 is at the 3V position (2-3) all the time and there´s 3.3V at the Vcc pins of the SDRAM slots.

Speaking ot the M520 - I do have one on my desk at the moment and these sticks do work:

The attachment vx_detected.JPG is no longer available

Both have 16 chips, the 32MB stick is detected correctly. the 256MB stick is detected as 64MB.

On the AB-PR5 there´s no reaction at all.

This post inspired me to try out the SDRAM slot on my M520. At the same time, my inner monologue warned me that it was probably a bad idea as the computer was working just fine with 96 megabytes of EDO SIMMs.

I grabbed my bag of SDRAM and set to work. Being a PCChips board, the PCB is thin and it has warped over the years. The SDRAM slot is at the very top of the board so it's poorly supported, and the board's warp is in the direction that made the SDRAM slot slightly concave. Add in oxidized contacts, and getting a stick of RAM full inserted and detected was a real job. I have it in a tiny AT case so it's crowded - even getting to the memory voltage jumpers was difficult.

After lots of trial and error I was able to find a 256 meg stick that was detected as 64 megs and it would reliably POST with it. But getting into Windows was another matter. No matter what I tried, Windows 98 was not stable and I was never able to complete a pass of 3D 2000 without it locking up or crashing. So I gave up and put the EDO SIMMs back in. But now it still wasn't stable? I decided to reinstall Win98, but I was having no luck with that either. I was having difficulties formatting the hard drive with Super FDISK and other weirdness. At one point I thought I had it, and was ready to reinstall 3D 2000. I have a Word doc that has the serial numbers for the versions of 3D Mark that I commonly use so I can copy and paste it. I made it on my main computer with Office 2007 and saved it so that it would be compatible with older OS and it has always worked fine with Win98. I tried to open it and I got an "Illegal Operation" error. I installed Word 97, and only then was I able to open it. I then tried to open it with another Win98 machine, and it too gave me the "Illegal Operation" error now! I put the thumb drive into my main computer and opened it (no problem) and saved it again, but it still wouldn't open in a Win98 computer? I found another thumb drive that had a copy of this doc, and that one worked fine in any computer!

No matter what I tried, I couldn't make the system stable. I checked it with Memtest, and that passed fine. I have the M520 set up with a K6-3+ 400 @ 450 using an Evergreen Spectra adapter, a 256 megabyte PCI FX5200 video card, and an SSD on a Sil3114 controller. It occurred to me that these are some power hungry components for a very old AT PSU. I had pulled the cover off the PSU and inspected it when I bought this computer and nothing looked amiss, but that hardly meant it was in the clear now. I tried swapping out the power hungry video card for a one megabyte Trident video card, and the machine got much further along installing Windows, so that strengthened my suspicion that the antique PSU was failing. I swapped in a known good AT PSU (Again a PIA with the tiny AT case) and tried again to reinstall Windows and this time everything was back to normal.

Being a glutton for punishment, I wondered if the reason the SDRAM wouldn't work was actually that the power supply was failing? So I tried it again, the result was the same, and I got to reinstall Win98 again.

Hahahaha, I did this kind of stupidity more times that i can even remember, hahahahaha. Finally when i remove the problem, some eediot thoughts are in my head, and i need to try again (presuming i had found the problem)! And i destroy everything again 😀 😀 😀 Don't be mad at Yourself, i think most of us did that a lot of times 😀
You really made me laugh, because i have remebered the same problems I have had before (and today still...)

Sockets with gold plated contacts like the 168-pin SDRAM or the 160-pin COAST slots seem to cause problems especially after years or decades of non-use. They often need to be cleaned thorougly both with water and IPA and you need to insert/remove a memory stick for like 50 times before they regain full contact.
Some vendors like PC-Chips make things worse by using super-cheapo components when it comes to CPU- and memory sockets (PCI as well).

Another weakness is the poorly designed 3.3V onboard VRM with two bipolar (BJT) transistors operating in parallel (!). This design flaw kills one of the transistors early in many cases. The programmable shunt also tends to fail on these mainbords.
Even with an additional VRM for Vcore or an interposer the SDRAM slot is powered with 3.3V from this crippled onboard VRM.
It´s no surprise when SDRAM modules don´t work correctly in this situation...

majestyk wrote on 2024-07-30, 07:38:

Sockets with gold plated contacts like the 168-pin SDRAM or the 160-pin COAST slots seem to cause problems especially after years or decades of non-use. They often need to be cleaned thorougly both with water and IPA and you need to insert/remove a memory stick for like 50 times before they regain full contact.
Some vendors like PC-Chips make things worse by using super-cheapo components when it comes to CPU- and memory sockets (PCI as well).

Interesting. Aren't gold plated contacts more resistant to rust? I have a couple of PCChips (and clones like Amptron, Elpina) motherboards and they seems to have goldplated contacts almost everywhere (even the DIN sockets for keyboard are goldplated) and they're all fine after 25+ years of retirement.

That´s true, the typical oxidation / corrosion /"rust" cannot happen to gold.
I suspect it could be remnants of flux, cleaning fluids, substances polluting the air like sulphides or just fine dust.

Having a card inserted all the time seems to prevent the process.