zapbuzz wrote on 2021-07-23, 15:30:

being so old I would settle with 1mb sticks as their lack of 2mb density mean they would last much longer in the same package but best of luck with 2mb

I will keep that in mind, still have the original 1x8mb modules anyhow, thank you

Are your memory sticks 3-chip or 9-chip? Some motherboards *require* 9-chip SIMMs for timing purposes. I have an Everex Tempo 486/33C and an MX 305 386DX motherboard that are like this, they will absolutely not work well with 3-chip memory at all, it's 9-chip or nothing.
Many 3-chip SIMMs are sold as '1Mx9' or '4Mx9' but definitely do not work quite the same as true 9-chip SIMMs.

Eep386 wrote on 2021-07-23, 15:43:

Are your memory sticks 3-chip or 9-chip? Some motherboards *require* 9-chip SIMMs for timing purposes. I have an Everex Tempo 486/33C and an MX 305 386DX motherboard that are like this, they will absolutely not work well with 3-chip memory at all, it's 9-chip or nothing.
Many 3-chip SIMMs are sold as '1Mx9' or '4Mx9' but definitely do not work the same as true 9-chip SIMMs.

Photos of old 1x8mb and new 4x4mb modules (lower one)

Try getting some 9-chip SIMMs instead, and see if your system doesn't work with those better.

Eep386 wrote on 2021-07-23, 15:49:

Try getting some 9-chip SIMMs instead, and see if your system doesn't work with those better.

I'm trying to narrow down the issues to the smallest scope possible but from what I can tell it's going to be hard to eliminate the risks of buying stuff blindly. If I go for it I will try at least to find silver connector modules too just in case. The original modules I showed here have parity, right? The ones with the Siemens' chips

Yeah, they're both parity SIMMs. Odd number of chips = parity, even number = non-parity.
A few non-parity SIMMs cheat by including a parity generator chip, but thankfully those are hard to come by these days. Only way to tell is to look up the part # of the chip in question, if it's a RAM chip then it's true parity; on the other hand if it's something like an FPGA (Field Programmable Gate Array) or something of the like, then it's a parity generator.
The SIMMs you have do not look like they have parity generators, just a third chip for true parity in their case.

Eep386 wrote on 2021-07-23, 16:05:

Yeah, they're both parity SIMMs. Odd number of chips = parity, even number = non-parity.
A few non-parity SIMMs cheat by including a parity generator chip, but thankfully those are hard to come by these days. Only way to tell is to look up the part # of the chip in question, if it's a RAM chip then it's true parity; on the other hand if it's something like an FPGA (Field Programmable Gate Array) or something of the like, then it's a parity generator.
The SIMMs you have do not look like they have parity generators, just a third chip for true parity in their case.

Thanks, is 60, 70 or 80ns important for my particular case? FPM, EDO?

Maybe they are looser timings to meet 60ns specs, like cas 3, and need 1T refresh instead of 2T etc... so if you went into setup on the working SIMMs, slackened everything right off, or load failsafe defaults if nothing else available, then try them again they might work???

Since your memory table proclaims support for 16MB SIMMs, I wouldn't immediately think it would be a density or power problem, since they'd need large capacity chips on those.

Stick with FPM for 486s. EDO modules in 30-pin SIMM format do not exist, far as I know.
I always shoot for the fastest possible speed DRAM, I say strive for 60ns. Even if your system doesn't explicitly require it, it's nice to have a little extra wiggle room for timing.
Though, I think that if you're running 33MHz fsb, then your system will 'officially' require 60ns SIMMs for zero/low-waitstate operation. At 25MHz fsb you can use 70ns at those timings.

BitWrangler wrote on 2021-07-23, 16:13:

Maybe they are looser timings to meet 60ns specs, like cas 3, and need 1T refresh instead of 2T etc... so if you went into setup on the working SIMMs, slackened everything right off, or load failsafe defaults if nothing else available, then try them again they might work???

Since your memory table proclaims support for 16MB SIMMs, I wouldn't immediately think it would be a density or power problem, since they'd need large capacity chips on those.

Thanks again, gave it a shot, no luck

Eep386 wrote on 2021-07-23, 16:15:

Stick with FPM for 486s. EDO modules in 30-pin SIMM format do not exist, far as I know.
I always shoot for the fastest possible speed DRAM, I say strive for 60ns. Even if your system doesn't explicitly require it, it's nice to have a little extra wiggle room for timing.
Though, I think that if you're running 33MHz fsb, then your system will 'officially' require 60ns SIMMs for zero/low-waitstate operation. At 25MHz fsb you can use 70ns at those timings.

I'm inclined to take one more shot at this, really hope it works this time. 8mb RAM on a 486 is not the end of the world but I'd like to improve it as much as I can to a certain degree

Eep386 wrote on 2021-07-23, 16:15:

Stick with FPM for 486s. EDO modules in 30-pin SIMM format do not exist, far as I know.
I always shoot for the fastest possible speed DRAM, I say strive for 60ns. Even if your system doesn't explicitly require it, it's nice to have a little extra wiggle room for timing.
Though, I think that if you're running 33MHz fsb, then your system will 'officially' require 60ns SIMMs for zero/low-waitstate operation. At 25MHz fsb you can use 70ns at those timings.

Weird that the back of the SIMM posted says EDO. I looked up the datasheet for MT4C4M4E8 and it does say EDO DRAM. What would 30-pin EDO actually have been used for though?

No clue. Most 486 boards will refuse to work with EDO DRAM. (Only 486 chipset I know that works with EDO, is a UMC PCI one, forgot the number.)
Perhaps they were used on those weird SIMM extender cards? Or maybe the EDO chips were simply added to bare SIMM PCBs long after the fact?

Edit: I noticed the datecode on those chips is 1998... that would indicate that the latter is the case.
At any rate, those EDO SIMMs will not work with this 486 system.

At least now I have a much stronger reason for the RAM not to work. Never thought that a simple RAM upgrade would be such a pain

After some time I managed to finish this build, with the correct 16MB Fast Page RAM and the DX2-66 update along with the new crystal needed for it. It has been a learning process for me, thank you everyone for your time and help, wouldn't have been possible without it.

I'm a bit intrigued regarding Doom and the new DX-2 66 / 16MB RAM, even in low detail it seems to me that it is playing worse than my DX-50 / 8MB RAM earlier config. Same Trident 8900CL in both cases, even though in DOS it doesn't matter that much I believe. Not that I'm counting on this 486 to play Doom regularly, but I noticed it anyway. Am I missing something here?

Check the ISA divider in CMOS Setup, if it was clock/5 for the DX50 it was running at 10Mhz, but now it will be a sluggish 6.6. clock/3 should work well, but you might have to go to clock/4 if NIC or soundcard act up (Mostly they're good to a bit over 12Mhz, but some glitch earlier, 8 bit cards might need it below 10 though)

BitWrangler wrote on 2021-10-19, 12:44:

Check the ISA divider in CMOS Setup, if it was clock/5 for the DX50 it was running at 10Mhz, but now it will be a sluggish 6.6. clock/3 should work well, but you might have to go to clock/4 if NIC or soundcard act up (Mostly they're good to a bit over 12Mhz, but some glitch earlier, 8 bit cards might need it below 10 though)

As soon as I can I will check it out, thank you. This is new territory for me, I'm kind of confused. Is this something that always needs to be done in order to upgrade processors from the era? I know most things in 486s and earlier need to be manually configured, but for this I thought that changing the crystal on the board would be enough. So "clock/3" is the optimistic setting, if not then "clock/4"

Yeah, some boards offer fixed frequencies also, so if it's set there by default then it's not that that was making it seem slower than the DX50. It's usually in Advanced Chipset Setup menu I think.

In an AMI BIOS it looks like this...

See where it says AT Clock Select and it's set CPUCLK/4

BitWrangler wrote on 2021-10-20, 01:46:

Yeah, some boards offer fixed frequencies also, so if it's set there by default then it's not that that was making it seem slower than the DX50. It's usually in Advanced Chipset Setup menu I think.

In an AMI BIOS it looks like this...

See where it says AT Clock Select and it's set CPUCLK/4

My Award BIOS seems more barebones than that AMI one, here's some images to show what I mean.

In the first boot to check this in the first panel it said "CPU clock...50mhz" even though I have the DX-2 66mhz installed, but in the next boot and after that it always said "CPU clock...66mhz". Doom still performs worse with this DX-2 66 comparing to the earlier DX-50 that came with the computer.

Does my BIOS have any option that I can tinker with to try and solve this issue? Or are there other solutions

Wow that is limited, I don't even see "load optimized defaults" option which might have set it more aggressively. It's either locked at 8mhz or motherboard is doing a best guess. However, what I would do here is write down any settings you need to remember* and do a full CMOS clear on the jumpers or by disconnecting battery to see if it does better setting up the DX2-66 fresh rather than with settings hanging over from when it had the DX50.

(* Though I guess those photos have all your settings)

BitWrangler wrote on 2021-10-21, 18:11:

Wow that is limited, I don't even see "load optimized defaults" option which might have set it more aggressively. It's either locked at 8mhz or motherboard is doing a best guess. However, what I would do here is write down any settings you need to remember* and do a full CMOS clear on the jumpers or by disconnecting battery to see if it does better setting up the DX2-66 fresh rather than with settings hanging over from when it had the DX50.

(* Though I guess those photos have all your settings)

Thanks, didn't quite make a difference. It's way too slow, even more so if I compare it to any "running Doom in a 486" video on Youtube.

If there are other options I'll try them out. Is it possible that the many solutions in the AMI BIOS have to be changed manually through jumpers on the board if like me you have the Award BIOS? Does my board's manual below have any option regarding changing ISA divider / clock like you said, or any other useful solution for this problem?

Edit: I also noticed another problem, and it may also be this: my graphics card is a Trident Quadtel TVGA8900CL 1MB 16-bit ISA, but it only says "512k" when booting..... Even though the modules are socketed I'm not sure how to safely remove and reseat them to check, I am missing a "safe" tool for it