What BIOS version do you have? There was a beta BIOS update to 1014.001 that specifically mentions Kingston 256MB SDRAM modules: https://www.asus.com/supportonly/TUSL2-C/HelpDesk_BIOS/

No idea what the actual cause would be. Looked up a photo of the RAM you listed, and the datasheet for the DRAM chips doesn't look at all odd.

The issue isn't single vs double sided, it's chip density. Because vendors felt that was too difficult for people to get their heads around, there was a lot of confusion spread around over sidedness.

Actually it's not so complicated. A DIMM is made up of a given number of chips, each with a certain size "density". Number of chips times density gives capacity of DIMM, or vice versa, divide DIMM capacity by number of chips for the density. The only complication is that DIMM sizes are quoted in (Mega)Bytes where chip densities are quoted in (Mega)bits, so you need to multiply/divide by 8 when moving between them, and be consistent in capitalization of the b/B or else you lose track very quickly.

So take those 256MB PC133 DIMMs. They are "single sided". That means they probably have eight chips. Each chip therefore is a 256/8*8=256Mb chip.

i440BX can handle max 128Mb chips, so unsurprisingly, this DIMM will work as 128MB on a BX board. The i815(PE) memory controller can however handle 256Mb, so it should show up as 256MB. If not, it could be a BIOS issue als snufkin mentions.

Note that there are other considerations where number of 'sides' of a DIMM do become relevant, particularly regarding how many DIMMs you can fit into a system. If you have multiple DIMMs installed at same time and you seem to be hitting odd detection issues, or if you are seeing different behaviour depending on which DIMM slot you use, these could come into play. However as far as I'm aware (and the manual states) the TUSL2-C has three identical 168p DIMM slots that willl accept any combination up to total 512MB (hard chipset limit, cynical ploy by Intel to push you towards i820 and Rambus if you needed more). By all means do try that single DIMM in different slots, just in case.

Thanks for the information dionb. Turns out I had one semi-faulty stick that was causing problems with the others. A combination of swapping DIMMs around and troubleshooting them on their own got everything working as indicated by the 815 specs. All DIMMs are operating at full capacity now.

I even was able to use 3 good sticks and get 768MB of RAM on the i815 Chipset. I didn't think that was supposed to work, but take a look at the screenshots.

Edit: Although it is detected and boots and seems to work, the system seems unstable with 3 sticks, locking up on Prime95's blended test even at stock frequencies.

Tried tweaking a few parameters like the "strong" memory drive setting in the bios, and the 3.3/3.4/3.5v RAM jumpers with no change.

dionb wrote on 2022-01-06, 15:06:

The issue isn't single vs double sided, it's chip density. Because vendors felt that was too difficult for people to get their heads around, there was a lot of confusion spread around over sidedness.

Actually it's not so complicated. A DIMM is made up of a given number of chips, each with a certain size "density". Number of chips times density gives capacity of DIMM, or vice versa, divide DIMM capacity by number of chips for the density.

Actually, it's not primarily about density, it's about the addressing. The 440BX supports 128 megabytes per rank. On 64-bit memory, this means 16M addresses containing 64 bits each. The chipset doesn't notice how many chips we use to implement the 16M x 64 organization. The 440BX will work as well with a module containing 4 16M x 16 chips (containing 256Mbit), as with modules containing 16 chips providing only 4 bits each. (those chips just have a "density" of 64 MBit). There is a limit on how many chips the chipset can drive at all, and I am unsure whether you may use 16M x 4 chips on all 8 ranks, as that would be 128 memory chips, which likely overloads the address lines.

The main point is: A dual-rank module with 16 128Mbit chips, each organized as 16M x 8 will work fine on a BX chipset, providing 256MB of memory. A single-rank module with 16 128MBit chips, each organized as 32M x 4 will not work properly. The important number is the depth (16M is OK, 32 is not) not the density (128MBit in both cases).

mkarcher wrote on 2022-01-09, 22:09:

[...]
Actually, it's not primarily about density, it's about the addressing. The 440BX supports 128 megabytes per rank. On 64-bit memory, this means 16M addresses containing 64 bits each. The chipset doesn't notice how many chips we use to implement the 16M x 64 organization. The 440BX will work as well with a module containing 4 16M x 16 chips (containing 256Mbit), as with modules containing 16 chips providing only 4 bits each. (those chips just have a "density" of 64 MBit). There is a limit on how many chips the chipset can drive at all, and I am unsure whether you may use 16M x 4 chips on all 8 ranks, as that would be 128 memory chips, which likely overloads the address lines.

It will not work on Intel memory controllers, they require x8 or x16 chips. DIMMs wtih 16 chips with x4 are not JEDEC-compliant but were sold as "Via-only" memory around 2001. Not sure what the benefit was to the vendors, probably offloading stock intended for registered DIMMs.

The main point is: A dual-rank module with 16 128Mbit chips, each organized as 16M x 8 will work fine on a BX chipset, providing 256MB of memory. A single-rank module with 16 128MBit chips, each organized as 32M x 4 will not work properly. The important number is the depth (16M is OK, 32 is not) not the density (128MBit in both cases).

Beg to differ, as if the depth were the sole determining factor, 16Mx16 chips would work at full capacity on a BX chipset, which they do not -a DIMM with 4 256Mb 16Mx16 chips will be detected as 64MB - as if it had been a 128Mb 8Mx16 chip - not as the full 128MB.