basically any 29F00NT eprom will do , doesn't have to be from Macronix... we looked on eBay as there are some EON ones in the US but the shipping really kills the deal at 24$ + 13$ for the IC ...

The original is 50% ok, bank0 is r/w whereas bank1 is stuck in read mode and can't be altered - plus there is a chance it is corrupted.
27SF020 does boot the board but gets stuck in both BIOS and boot process, I think bank switching is different or something like that.

keropi wrote:

basically any 29F00NT eprom will do , doesn't have to be from Macronix... we looked on eBay as there are some EON ones in the US but the shipping really kills the deal at 24$ + 13$ for the IC ...

The original is 50% ok, bank0 is r/w whereas bank1 is stuck in read mode and can't be altered - plus there is a chance it is corrupted.
28SF020 does boot the board but gets stuck in both BIOS and boot process, I think bank switching is different or something like that.

Just a question,but can't you use anything else than MX? Like a Winbond,or a SST. Any EEPROM will do,as long as it fits the voltage (which I think is universal on those motherboards) and the size is the same.

Just get hold of,say,a Winbond W29C020,put the newest BIOS for the Soyo SY-6BA+ IV,boot onto DOS,and use Uniflash to flash the BIOS.

Uniflash saved my butt in a lot of cases btw.

oh yeah, I wrote that wrong, "any 29F002NT eprom will do" is the correct. Ofcourse it doesn't have to be Macronix that's why I pointed about the EON ones on eBay. It seems only Macronix , EON and AMD made them though.

I don't know about W29C020 if it will work , I did try with a 70ns 28SF020 that is also a 2mbit/8bit device like the original eprom , the board booted but froze while it complaint the BIOS settings are invalid. If you press DEL and get in the BIOS before this message it freezes after it loads the award bios screen and has a message at the bottom "determining cpu speed settings" or something like that - SOYO customization I guess because there are some extra menus on the screen. Ofcourse the CPU is OK.
I *think* 29F002NT devices do not react the same under certain conditions like the rest of the 2mbit/8bits/Jedec devices, much like the "xxxxPC" eproms for newer mobos. Bank1 on the flash does look like it's the promise controller BIOS to me, maybe some special trickery is done so both are accessed without conflicts or something... At any rate I don't have any W29C020 to test or else I would 😒

I am not using uniflash as there is no need, I have a couple of programmers that handle easily these eproms. 😎

keropi wrote:

oh yeah, I wrote that wrong, "any 9F002NT eprom will do" is the correct. Ofcourse it doesn't have to be Macronix that's why I po […]
Show full quote

oh yeah, I wrote that wrong, "any 9F002NT eprom will do" is the correct. Ofcourse it doesn't have to be Macronix that's why I pointed about the EON ones on eBay. It seems only Macronix , EON and AMD made them though.

I don't know about W29C020 if it will work , I did try with a 70ns 28SF020 that is also a 2mbit/8bit device like the original eprom , the board booted but froze while it complaint the BIOS settings are invalid. If you press DEL and get in the BIOS before this message it freezes after it loads the award bios screen and has a message at the bottom "determining cpu speed settings" or something like that - SOYO customization I guess because there are some extra menus on the screen. Ofcourse the CPU is OK.
I *think* 29F002NT devices do not react the same under certain conditions like the rest of the 2mbit/8bits/Jedec devices, much like the "xxxxPC" eproms for newer mobos. Bank1 on the flash does look like it's the promise controller BIOS to me, maybe some special trickery is done so both are accessed without conflicts or something... At any rate I don't have any W29C020 to test or else I would 😒

I am not using uniflash as there is no need, I have a couple of programmers that handle easily these eproms. 😎

You don't need to EXACTLY use a W29C020. You can use any 90ns EEPROM.

And 70ns? Now that's what I call ancient. But anyways,try a 90ns chip. Just so you know,most 29SF0x0 EEPROMs and W29C020-90 chips are 90ns. In fact,I still think any 440BX-onwards motherboard with a DIP32 socket has a 90ns chip.

Small note: I've checked the datasheet of F2bnp's EEPROM. It's a 90ns,so you need to use a 90ns chip. That's for sure.

BTW,a funny thing is,my GA-6BXC has a 120ns EEPROM. 🤣

I thought I remembered something from back in the BBC Micro days - the pin 1 problem (though that was with 16k's)
On the MX, a 5V programming version, pin 1 is NC (or not RESET on some others)
On the majority of 12V or higher programming EPROMS, pin 1 is VPP (programming voltage) while some will accept it NC on read, others require it at 5V for read, and may not work, or work unreliably if not.

For the 28SF020, I'd suggest bending pin 1 up so it misses the socket, and then wiring it to pin 32

@PcBytes
I can only assume you have understood the speed rating of (e)eproms in a wrong way.
The speed rating is in nanosecons (ns) and it is the time the memory device needs so it can spit out it's contents on the data pins once activated.
So a 70ns device needs 70ns and a 90ns device is slower and needs 20ns more to do the same job.
So more ns means a slower device. When replacing an (e)eprom you need devices with the same or LOWER ns rating so that you are sure data are within the time limits that the host machine/mobo/whatever needs.

@Matth79
ah the old tricks... will try that immediately - thanks for bringing that up!

edit:
turns out that the culprit was the ultra-crap-cheap ps2 keyboard I use for testing, the machine did not hung - the keyboard just stopped working. Using a good one and everything is OK. The onboard highpoint ide controller though complains about "unknown flash type" after it's boot screen.
Pin1 on the BIOS socket is being fed with 12v , I don't think I need to connect it to +5v since 12v is within the 27SF020's specs. Maybe if it is to kill it's ability to program or something?

The highpoint bios has no options regarding flash types, it does complain about "unknown flash type" but it seems settings are saved elsewhere so who cares 🤣

I now realize I wrote 28SF instead of 27SF 😵

Wow, turns out everything can rebel against you just to piss you off, even your own keyboard 😈 . Problem solved then, thanks for the answers and suggestions everybody.

Matth79 wrote:

I thought I remembered something from back in the BBC Micro days - the pin 1 problem (though that was with 16k's) On the MX, a 5 […]
Show full quote

I thought I remembered something from back in the BBC Micro days - the pin 1 problem (though that was with 16k's)
On the MX, a 5V programming version, pin 1 is NC (or not RESET on some others)
On the majority of 12V or higher programming EPROMS, pin 1 is VPP (programming voltage) while some will accept it NC on read, others require it at 5V for read, and may not work, or work unreliably if not.

For the 28SF020, I'd suggest bending pin 1 up so it misses the socket, and then wiring it to pin 32

It would be easier (and safer) for keropi to just program a 90ns chip and use that. The MX29F002NTP is a 90ns chip.

I think ATMEL chips might work too but I have to dig more into that to be sure.

@PcBytes
Please stop with the whole 90ns thing, I have explained why this is wrong.

What do you mean bank0 is ok and bank1 isn't? Looking at the datasheet there are no mentions of banks. Or do you mean there are two separate chips in a hi/lo configuration?

Looking at the datasheet it appears to be a standard JEDEC pinout. It matches the SST39SF020A used by the XTIDE controller and they seem easy to get (I bought five from RS Components for about AU$10 shipped.) But these are also the same pinout as the 28SF020 that you said didn't work, so any idea why that didn't work?

The write procedure does look a bit different so you probably won't be able to reflash non MX replacement chips in the original device, but if you flash them elsewhere first it looks like they should work fine so not sure why you had problems. You would still no doubt get the error about the unknown flash chip but since all chip types have different IDs the only way to avoid that would be to get exact parts which would probably be tricky.

Either that or reverse-engineer the BIOS to see what chip IDs it recognises as valid, which may give you some alternatives...

Or, of course, patch the replacement BIOS so that the ID of whatever chip you use is recognised as valid 😀

28SF020 worked fine, I explained it was the keyboard that was the problem . Ofcourse before even plugging it I checked the datasheets of both devices 😉

29F002NT is separated internally in segments, maybe bank0/1 wasn't 100% "correct" but what matters is that delete+write operations stop at 50% of it's flash memory and give an error. 0x000000 - 0x01ffff area is OK , 0x020000 - 0x03ffff is not , cannot be deleted and it's not certain that the data there is OK - I estimate they are a mix of original + failed update ones done on the pc with awdflash. My programmer fully supports the MX29F002NT eeprom so it's not the problem either.

I can't reverse the BIOS or change ID strings , 28SF020 will do for now and if problems arise from that then a replacement 29F002NT device would be the best course of action. 😀