128MB EDO 72 pin modules were kinda rare even back in the day, its not surprising they need active cooling as they would have originally been server ram or in machines with active airflow. As for the voltage you could measure the ram input voltage at the fet, measure it with both 5v and 3.3v modules and see if the board changes the voltage or if it remains constant. It does sound like the board is giving 5v regardless of what memory is there which wouldn't be unusual for the age of the motherboard and chipset which is only expecting 5v Simms and doesn't have the capacity to autodetect.

The question then is is it worth running the memory out of spec even if the board lets you at which point youll have to decide if adding active cooling to the memory is worth the trouble.

For my part I would run it with a fan, so long as it passes mem test with the fan then it should be fine, pretty sure you wont be using the system heavily so the risk to the modules would be pretty low, heck if there was room I would add thin copper shims to the modules to help with heat dissipation.

For example. the earlier HX chipset supports FPM, EDO, 3.3V and 5V - according to the datasheet:

Never mind what someone writes on "chip.de".

My 512mb of 60ns 128x4 runs cold. It sounds like yours are being over volted. Remove them before damage

majestyk wrote on 2022-11-06, 12:56:

For example. the earlier HX chipset supports FPM, EDO, 3.3V and 5V - according to the datasheet:

hx_ram.JPG

Never mind what someone writes on "chip.de".

The question is did ASUS use the require circuirty to allow memory voltage autodetect by the chipset or did they just reuse an earlier version of the board and just update the chipset but cut a few corners, there were a few different models of this board with different chipsets.

PC mainboards generally run 30-pin SIMMs and 72-pin SIMMs at 5V supply voltage. This is definitely too high for the 3.3V Micron chips. They are specified at 4.6V "absolute maximum", which technically means that once you install them into a 5V mainboard, they might be permanently damaged (although I don't expect they are). There are 72-pin SIMM modules that use Z diodes to drop around 1.7V of the supply voltage so the chips get 3.3V even if the slot is connected to +5V. If the chips are "5V compatible", they accept 5V at the data pins even if they just get 3.3V supply voltage. They will still not provide 5V for logic high, though. The Micron chips you mentioned are not "5V compatible". At 3.3V supply voltage, the signal voltage should not exceed 3.6V (recommended condition) and must never exceed 4.6V. Exceeding 3.6V may cause erratic behaviour. Exceeding 4.6V may cause permanent damage.

It seems the 82439HX is itself a 5V tolerant 3.3V device on its memory interface. It outputs a high level high enough to be detected as high by both 5V TTL logic and 3.3V CMOS logic, and it doesn't break of 5V RAM sends 5V into the data pins. So you don't need to worry about signal voltage exceeding the recommended operating conditions of your 3.3V memory chips.

At the time the 430TX came around, DIMM sockets started to appear on the mainboards, because the new SDRAM only exists on DIMMs. SDRAMs run at 3.3V power voltage. There also are uncommon EDO DIMMs, if I remember correctly there are both 3.3V and 5V variants. Boards at that time often had a jumper to select DRAM supply voltage, which needed to be set to 3.3V for SDRAMs and (usually) 5V for EDO RAM. So while the HX chipset already supports interfacing with 3.3V RAM (no challenge on that fact), most HX-based boards do not support supplying 3.3V to the PS/2 SIMM sockets.

Please post a photo of your RAM modules. Possibly there are Zener diodes to drop the supply voltage. In that case, it would be OK to run them at 5V supply voltage.

TrashPanda wrote on 2022-11-06, 13:08:

The question is did ASUS use the require circuirty to allow memory voltage autodetect by the chipset or did they just reuse an earlier version of the board and just update the chipset but cut a few corners, there were a few different models of this board with different chipsets.

Agreed - of course the mainboard / BIOS must be prepared to provide the appropriate supply voltages. I only wanted to amend the statemant that the 430TX chipset was the first to support 3.3V RAM. Here´s the timeline:

I currently can't upload something. The EDOs don't seem to have a Zener diode, only one Tantalum and some ceramic capacitors (when I identified everything correct). Photos will come later.

Hirsch wrote on 2022-11-06, 18:38:

I currently can't upload something. The EDOs don't seem to have a Zener diode, only one Tantalum and some ceramic capacitors (when I identified everything correct). Photos will come later.

Yeah, the hoster of Vogons has some issues, attachments are temporarily disabled to keep the rest of the Forum alive. I searched the Web for photos of modules with dropper diodes, and I found These modules with a voltage regulator (likely a LM1117 variant) included, a design I have not yet seen. I did not find pictures of memory modules with dropper diodes, so I can't link to pictures. I have a pair of modules that contains one or two mini-MELF diodes next to the RAM chips.

majestyk wrote on 2022-11-06, 14:59:

TrashPanda wrote on 2022-11-06, 13:08:

The question is did ASUS use the require circuirty to allow memory voltage autodetect by the chipset or did they just reuse an earlier version of the board and just update the chipset but cut a few corners, there were a few different models of this board with different chipsets.

Agreed - of course the mainboard / BIOS must be prepared to provide the appropriate supply voltages. I only wanted to amend the statemant that the 430TX chipset was the first to support 3.3V RAM. Here´s the timeline:

i430xx.JPG

My tyan s1564 a 430hx supports 3.3v (has jumpers for it)

Sphere478 wrote on 2022-11-07, 03:02:

majestyk wrote on 2022-11-06, 14:59:

TrashPanda wrote on 2022-11-06, 13:08:

The question is did ASUS use the require circuirty to allow memory voltage autodetect by the chipset or did they just reuse an earlier version of the board and just update the chipset but cut a few corners, there were a few different models of this board with different chipsets.

Agreed - of course the mainboard / BIOS must be prepared to provide the appropriate supply voltages. I only wanted to amend the statemant that the 430TX chipset was the first to support 3.3V RAM. Here´s the timeline:

i430xx.JPG

My tyan s1564 a 430hx supports 3.3v (has jumpers for it)

Well if we are being fair Tyan boards were considerably better quality than your average ASUS consumer grade board and you paid for that extra quality.

Hirsch wrote on 2022-11-06, 18:38:

I currently can't upload something. The EDOs don't seem to have a Zener diode, only one Tantalum and some ceramic capacitors (when I identified everything correct). Photos will come later.

You can upload on Imgur and post the pictures here via IMG tag.

Like I told you before, Memtest86+ errors often get misinterpreted.
There are 2 types of errors. Type 2 is fixable, Type 1 isn't.

Type 1:
They will appear quickly and always at the same location. Sometimes you'll get only one error per pass. Sometimes you'll get thousands during one pass (like when I tested the DDR3 from your retired LGA1133 system 😁).
You always need to wait for two passes. The first pass skips half the tests. I had sticks who only produced errors on the 2nd and consecutive passes.
These sticks are definitely broken. Not completely dead yet, but also not really alive.

Type 2:
They're phantom errors and will appear on random locations. They will often only surface after a few hours of testing. If you get an error after 2nd pass, it's a Type 2 error.
RAM isn't ment to run at 100% workload for 24/7. If you're stresstesting RAM during summer, you might encounter errors during the day, while the RAM works flawless at night. When hardware gets hot, it becomes less efficient and requires more voltage to achieve the same performance. When they need more voltage than they get, they will start to produce small errors.
You can solve this by running them at a lower frequency. Or you can improve the efficiency, by improving the airflow. Or you can bump the voltage a tiny bit to compensate to the loss of efficiency due to high temperatures. That's the madman approach. 😁

From what I've read here it seems that you have unnecessarily and unintentionally overvolted the sticks by a lot. Find a way to run them at 3.3V and then perform another 24h hour test. With such expensive sticks you wanna get sure that they're ok.

Im sorry, but its wrong/not true.

RaiderOfLostVoodoo wrote on 2022-11-07, 06:52:

There are 2 types of errors. Type 2 is fixable, Type 1 isn't.
Type 2:
They're phantom errors

All errors are quite real. Maybe you meant to divide them into obvious hardware defects vs marginal operation.

RaiderOfLostVoodoo wrote on 2022-11-07, 06:52:

RAM isn't ment to run at 100% workload for 24/7.

Ram is indeed meant to run 24/7, as are CPUs and other chips.

RaiderOfLostVoodoo wrote on 2022-11-07, 06:52:

When hardware gets hot, it becomes less efficient and requires more voltage to achieve the same performance. When they need more voltage than they get, they will start to produce small errors.

The main parameter affected by higher temperature is faster discharge rate requiring faster refresh rate, but thats at the extreme end of allowable operating temperature spectrum, which for FPM/EDO chips was at 70-85C. Ram doesnt "becomes less efficient and requires more voltage to achieve the same performance". The trick of bumping supply voltage to make certain components more stable (mostly when overclocking) works by influencing signal integrity. Extra voltage is meant to push signals carrying data (be it transmitting, switching transistors, or charge stored in cells) that extra step above the noise, giving it higher margin of error against interference/inherent noise.

Running 3.3V ram at 5V can indeed results in errors because chips will be working outside their designed operational parameters. Temperature, as 5V might cause them to roast, isnt the only reason. All voltage tresholds and timings will be wrong. Higher voltage might also lead to accelerated electomigration if silicon wasnt designed for it.

RaiderOfLostVoodoo wrote on 2022-11-07, 06:52:

You can upload on Imgur and post the pictures here via IMG tag.

I don't like this because Imgur might change the URL or delete the photos in the future. Somebody reading this thread in the future might miss the photos then. Uploading photos fortunatelly works again, so here they are:

I think that the sticks survived, because the test with active cooling ran without errors and I didn't use them any more after that. I draw the conclusion from the discussion here that desktop/consumer boards always put +5V to the SIMM sockets (as mentioned by mkarcher), while server boards like Sphere478's Tyan S1564 might offer 3.3V / 5V selectable by jumper.

This means: The only way to run 3.3V only EDO SIMMs in a desktop board like the Asus P/I-P55T2P4 is a hardware mod on the mainboard. Do you agree or does anyone see another option (except from modding the EDO SIMMs)?

Agreed. No voltage droppers in the modules, and a label clearly indicating 3.3V operating voltage. That module is not compatible with most desktop boards. As I said, on boards with SDRAM and EDO support, with DIMM and PS/2 RAM slots, you might find RAM voltage jumpers even on desktop boards. This is one to two years later than 430HX boards You would still need to verify that the voltage jumpers affect the PS/2 slots as well as the DIMM slot(s).

You are right that you would need to modify either the board to supply 3.3V or modify the SIMMs to accept 5V to make them compatible.

In theory the following mod should do the job:

• desolder SIMM sockets
• insulate VCC pads 10, 30 and 59 of mainboard with Kapton tape
• cut VCC pins 10, 30 and 59 from the desoldered sockets
• solder thin enamelled copper wire to the VCC pins of the sockets
• resolder sockets with enamelled copper wire running out at the side of the sockets
• place somewhere a jumper connecting the enamelled copper wire selectable to 5V or 3.3V

Maybe I'll try this one day but not now, since it sounds lile lots of work for me.

Hirsch wrote on 2022-11-07, 20:41:

I don't like this because Imgur might change the URL or delete the photos in the future. Somebody reading this thread in the fut […]
Show full quote

RaiderOfLostVoodoo wrote on 2022-11-07, 06:52:

You can upload on Imgur and post the pictures here via IMG tag.

I don't like this because Imgur might change the URL or delete the photos in the future. Somebody reading this thread in the future might miss the photos then. Uploading photos fortunatelly works again, so here they are:
20221106_183806.jpg
20221106_183817.jpg

I think that the sticks survived, because the test with active cooling ran without errors and I didn't use them any more after that. I draw the conclusion from the discussion here that desktop/consumer boards always put +5V to the SIMM sockets (as mentioned by mkarcher), while server boards like Sphere478's Tyan S1564 might offer 3.3V / 5V selectable by jumper.

This means: The only way to run 3.3V only EDO SIMMs in a desktop board like the Asus P/I-P55T2P4 is a hardware mod on the mainboard. Do you agree or does anyone see another option (except from modding the EDO SIMMs)?

Perhaps a 3.3v simm interposer could be constructed?

I shal name it simmterposer

Sphere478 wrote on 2022-11-08, 02:21:

Hirsch wrote on 2022-11-07, 20:41:

I don't like this because Imgur might change the URL or delete the photos in the future. Somebody reading this thread in the fut […]
Show full quote

RaiderOfLostVoodoo wrote on 2022-11-07, 06:52:

You can upload on Imgur and post the pictures here via IMG tag.

I don't like this because Imgur might change the URL or delete the photos in the future. Somebody reading this thread in the future might miss the photos then. Uploading photos fortunatelly works again, so here they are:
20221106_183806.jpg
20221106_183817.jpg

I think that the sticks survived, because the test with active cooling ran without errors and I didn't use them any more after that. I draw the conclusion from the discussion here that desktop/consumer boards always put +5V to the SIMM sockets (as mentioned by mkarcher), while server boards like Sphere478's Tyan S1564 might offer 3.3V / 5V selectable by jumper.

This means: The only way to run 3.3V only EDO SIMMs in a desktop board like the Asus P/I-P55T2P4 is a hardware mod on the mainboard. Do you agree or does anyone see another option (except from modding the EDO SIMMs)?

Perhaps a 3.3v simm interposer could be constructed?

I shal name it simmterposer

Sounds Sus.

TrashPanda wrote on 2022-11-08, 03:00:

Sphere478 wrote on 2022-11-08, 02:21:

Hirsch wrote on 2022-11-07, 20:41:

I don't like this because Imgur might change the URL or delete the photos in the future. Somebody reading this thread in the fut […]
Show full quote

I don't like this because Imgur might change the URL or delete the photos in the future. Somebody reading this thread in the future might miss the photos then. Uploading photos fortunatelly works again, so here they are:
20221106_183806.jpg
20221106_183817.jpg

I think that the sticks survived, because the test with active cooling ran without errors and I didn't use them any more after that. I draw the conclusion from the discussion here that desktop/consumer boards always put +5V to the SIMM sockets (as mentioned by mkarcher), while server boards like Sphere478's Tyan S1564 might offer 3.3V / 5V selectable by jumper.

This means: The only way to run 3.3V only EDO SIMMs in a desktop board like the Asus P/I-P55T2P4 is a hardware mod on the mainboard. Do you agree or does anyone see another option (except from modding the EDO SIMMs)?

Perhaps a 3.3v simm interposer could be constructed?

I shal name it simmterposer

Sounds Sus.

SimmTerposer 5000?

But for real. It sounds like a good idea to make a interposer if there is a good supply of these simms now?

Someone drop a public domain 72 pin simm board and edge connector footprint and I can start playing with it in free time.

We could put muratas on them… adjustable voltage! :p …or just a boring old linear regulator. The linear used on Feipoa’s interposer that I made for him should do the trick. 5v to 3.3v (and adjustable)

Sphere478 wrote on 2022-11-08, 03:22:

SimmTerposer 5000? […]
Show full quote

TrashPanda wrote on 2022-11-08, 03:00:

Sphere478 wrote on 2022-11-08, 02:21:

Perhaps a 3.3v simm interposer could be constructed?

I shal name it simmterposer

Sounds Sus.

SimmTerposer 5000?

But for real. It sounds like a good idea to make a interposer if there is a good supply of these simms now?

Someone drop a public domain 72 pin simm board and edge connector footprint and I can start playing with it in free time.

We could put muratas on them… adjustable voltage! :p …or just a boring old linear regulator. The linear used on Feipoa’s interposer that I made for him should do the trick. 5v to 3.3v (and adjustable)

...sorry my social media amongus meme knowledge is leaking through.