I'm not sure but it could also be CMOS battery related (sorry, long time ago I had the same problem but I can't remember the exact solution).
I only remember I also had to fix a dead battery in the RTC-module, but like said, not sure if that also caused the ESCD message.
Maybe others here can confirm (or not).

Yoghoo wrote on 2024-01-30, 15:46:

appiah4 wrote on 2024-01-30, 15:39:

A UV EPROM is not something the board can update you need an EEPROM?

Probably true. But it's a little strange as it is the original BIOS EPROM. Seems funny to me that Biostar sold motherboards which couldn't save ESCD information. 😀

Its not original, picture with original sticker peeled off Biostar MB-8433UUD-A

rasz_pl wrote on 2024-01-31, 21:05:

Yoghoo wrote on 2024-01-30, 15:46:

appiah4 wrote on 2024-01-30, 15:39:

A UV EPROM is not something the board can update you need an EEPROM?

Probably true. But it's a little strange as it is the original BIOS EPROM. Seems funny to me that Biostar sold motherboards which couldn't save ESCD information. 😀

Its not original, picture with original sticker peeled off Biostar MB-8433UUD-A

See attachment for the bios sticker. It was undamaged before I removed the sticker to get the part number. As it's impossible to remove it without damaging the sticker I believe it is original. I know you can use heat to removing it more easily but even then you leave some marks which were not there before. Also note there are many revisions of this motherboard so parts can change between revisions (mine is rev 3.1). But all in all not very important as I just want to solve my problem. 😀

If I am lucky my new parts will arrive today so I can change the bios to a flashable bios which hopefully will solve the problem I am having.

All I can say is my board also has an EEPROM. That said, if you disable Update ESCD on net bootup in BIOS it will help your pain until the EEPROMs arrive.

I may remember incorrectly, but I sorta recall motherboard versions which were sold with a UV EPROM chip contained a BIOS revision which did not try to update ESCD information. Did you see Updating ESCD... with your BIOS before you flashed it? Or did this only happen after you flashed it?

I'm pretty sure you need an EEPROM to update ESCD information. I don't think the UV EPROMs normally contain a EEPROM section to maintain ESCD.

I just received some flashable EEPROMS (faster then I thought it would take). Replacing the EPROM with a 5v EEPROM solved the problem.

The "Updating ESCD" was there before the flashing. Hoped that flashing a newer BIOS version would solve it. But as we now know it didn't for obvious reasons. 😜

Good to know you fixed it. Now enjoy this glorious motherboard 😁

By the way, the Dallas RTC on mine from 2002 is still going but I probably ought to start looking around for a modern replacement. Does anyone know if Necroware's solution works with this board?

appiah4 wrote on 2024-02-02, 05:46:

Good to know you fixed it. Now enjoy this glorious motherboard :D

By the way, the Dallas RTC on mine from 2002 is still going but I probably ought to start looking around for a modern replacement. Does anyone know if Necroware's solution works with this board?

I have a unit from a guy who designed a RTC replacement long before necroware, but I forget his user name. He sent me a prototype that I tested for over a year in the MB-8433UUD and it worked perfectly. Do the smaller batteries on these modern remake RTC modules last as long as on Dallas RTC modules?

I have not bothered for order or make any of the modern homebrew RTC modules because the motification for dead Dallas/ODIN RTC modules is so simple that this is the approach I take. I've only run into one situation in which adding a CR2032 atop a Dallas RTC was too tall to fit under an expansion card. For that system, I use the modern RTC solutions.

appiah4 wrote on 2024-02-02, 05:46:

Good to know you fixed it. Now enjoy this glorious motherboard 😁

By the way, the Dallas RTC on mine from 2002 is still going but I probably ought to start looking around for a modern replacement. Does anyone know if Necroware's solution works with this board?

Thanks. Will certainly enjoy it. 😁

Have a Necroware solution on it's way. So can tell in a couple of days if it will work or not. There's a modified Dallas chip in mine atm but it's blocking an ISA and PCI slot a bit so that's why I ordered one.

Also increased the level 2 cache from 256kb to 512kb yesterday. Didn't see much of a speed increase from 15ns to 10ns but need to do some more testing. Anybody knows a good test program (besides cachechk) to test it?

feipoa wrote on 2024-02-02, 07:59:

appiah4 wrote on 2024-02-02, 05:46:

Good to know you fixed it. Now enjoy this glorious motherboard 😁

By the way, the Dallas RTC on mine from 2002 is still going but I probably ought to start looking around for a modern replacement. Does anyone know if Necroware's solution works with this board?

I have a unit from a guy who designed a RTC replacement long before necroware, but I forget his user name. He sent me a prototype that I tested for over a year in the MB-8433UUD and it worked perfectly. Do the smaller batteries on these modern remake RTC modules last as long as on Dallas RTC modules?

I have not bothered for order or make any of the modern homebrew RTC modules because the motification for dead Dallas/ODIN RTC modules is so simple that this is the approach I take. I've only run into one situation in which adding a CR2032 atop a Dallas RTC was too tall to fit under an expansion card. For that system, I use the modern RTC solutions.

I've considered using a modded RTC myself but the one on this board is in a rather annoying spot. I have socketed mine, so if I also add a battery holder on top it's going to make using the ISA slot below rather difficult..

Another thing I recently noticed is that the board also seems to have provision for a battery holder under the PS/2 port. I am not keen on desoldering that to actually mount one (it would probably require more components in the charging circuitry as well regardless) but I have noticed an unpopulated header at JP42. This is suspiciously close to where the Dallas RTC is located, sits next to some unpopulated diodes, and I am guessing it is BAT_EXT header. I wonder if that could be used to provide external +3V hassle free...

EDIT2: There is, actually, yet another such unpopulated header at JP4 to the left of the RTC battery that sits suspiciously close to unpopulated diodes, but this one is 5 pin and even more curious..

I think JP42 is for use with a traditional 386-style external lithium ion battery, but not when a Dallas RTC is installed. I have seen people run wires from a dead/modded-RTC out to the location of where the CR2032 coin holder would be installed. This gets around issues whereby the modded RTC is too tall, but then you have some ugly wires to look at.

I don't know what JP4 is for. Sometimes these newer boards had an IR header

Yoghoo wrote on 2024-02-02, 08:02:

Also increased the level 2 cache from 256kb to 512kb yesterday. Didn't see much of a speed increase from 15ns to 10ns but need to do some more testing. Anybody knows a good test program (besides cachechk) to test it?

My experience is that running these tests usually indicates bad L2 cache timings quite fast:

• Running Quake (e.g. the demo run in DOSBENCH)
• Running Doom (again, DOSBENCH provides an easy way)
• Booting Windows 2000
• Run Stunts 4D Sports car driving.

If all these tasks run without problems, the L2 timings are likely conservative enough for stable system operation. Faster cache RAM will not make your system run faster just by installing. The system only gets faster when you enable faster cache timings in the CMOS setup. A read burst of 2-1-1-1 and 0WS cache writes is the fastest possible setting. At FSB33, you should easily obtain that speed using 15ns RAM. At FSB40, timing is marginal. The upgrade to 512KB might make performance worse, because you lose bank interleaving, which makes the "-1-1-1" part harder to obtain. Furthermore, be very wary about 1024k x 8 DIP chips claiming to have 10ns timings. There are a lot of chips marked this way, but as far as the community knows, e.g. ISSI never produced their IS6C1024N chip at 10ns. They only packaged 10ns chips in SOJ. Any chip claimning to be an ISSI 128k x 8 chip at 10ns has a fake marking. Don't expect these "10ns" chips to deliver better performance than your current 15ns already do. When you order SOJ chips from chip resellers shipping from China, you can also get chips with fake speed markings. I got bitten by that: A made a PCB adapter to install 1MB cache into this board using SOJ chips I got (IIRC I ordered them at AliExpress) which didn't support what 10ns was supposed to support. I then checked response times using a suffiently fast scope and found out they behave like 15ns chips, so it wasn't surprinsing I couldn't get reliable 2-1-1-1 at 40MHz.

mkarcher wrote on 2024-02-03, 12:57:

My experience is that running these tests usually indicates bad L2 cache timings quite fast: […]
Show full quote

Yoghoo wrote on 2024-02-02, 08:02:

Also increased the level 2 cache from 256kb to 512kb yesterday. Didn't see much of a speed increase from 15ns to 10ns but need to do some more testing. Anybody knows a good test program (besides cachechk) to test it?

My experience is that running these tests usually indicates bad L2 cache timings quite fast:

• Running Quake (e.g. the demo run in DOSBENCH)
• Running Doom (again, DOSBENCH provides an easy way)
• Booting Windows 2000
• Run Stunts 4D Sports car driving.

If all these tasks run without problems, the L2 timings are likely conservative enough for stable system operation. Faster cache RAM will not make your system run faster just by installing. The system only gets faster when you enable faster cache timings in the CMOS setup. A read burst of 2-1-1-1 and 0WS cache writes is the fastest possible setting. At FSB33, you should easily obtain that speed using 15ns RAM. At FSB40, timing is marginal. The upgrade to 512KB might make performance worse, because you lose bank interleaving, which makes the "-1-1-1" part harder to obtain. Furthermore, be very wary about 1024k x 8 DIP chips claiming to have 10ns timings. There are a lot of chips marked this way, but as far as the community knows, e.g. ISSI never produced their IS6C1024N chip at 10ns. They only packaged 10ns chips in SOJ. Any chip claimning to be an ISSI 128k x 8 chip at 10ns has a fake marking. Don't expect these "10ns" chips to deliver better performance than your current 15ns already do. When you order SOJ chips from chip resellers shipping from China, you can also get chips with fake speed markings. I got bitten by that: A made a PCB adapter to install 1MB cache into this board using SOJ chips I got (IIRC I ordered them at AliExpress) which didn't support what 10ns was supposed to support. I then checked response times using a suffiently fast scope and found out they behave like 15ns chips, so it wasn't surprinsing I couldn't get reliable 2-1-1-1 at 40MHz.

Did you decide to shelf the 8433UUD cache interposer project? There were some 8 ns TSOP and SOJ modules from ISSI. I show a photo of the top/bottoms here: Re: Lets make new M919 Cache sticks? As they are from China, I don't know if they are authentic or not. The ISSI TSOPs work on the M919 1024K cache module.

Regarding 2-1-1-1 at 40 MHz. I recall pshipkov was able to obtain this setting by doing a lot of swapping around of DIP SRAM modules. See below:

pshipkov wrote on 2021-09-28, 03:06:

This last time it took few hours, but was doing other things as well. Order matters. Swapping chip in slot 0 with chip in slot 1 […]
Show full quote

feipoa wrote on 2021-09-27, 07:44:

How long did it take you to find the right combination of SRAM modules to work at 1024K, 160 MHz, and 2-1-1-1 timings? And does that particular order of modules matter? I assume it does. Unfortunately, I don't have a lot of different brands of 128kx8 modules to try out like you. I have dozens of the Chinese fakes and only 4 pieces of original 128kx8 modules from the 90's.

This last time it took few hours, but was doing other things as well.
Order matters. Swapping chip in slot 0 with chip in slot 1 can be the difference between fully stable and flaky.
I don't have that many 128kx8 chips. Once i got into the retro optimize and overclock thing, i had to scrub some of the motherboards i have here from their chips.
The 10ns "fake" ones are good in my opinion.

feipoa wrote on 2021-09-27, 07:44:

mkarcher wrote on 2024-02-03, 12:57:

My experience is that running these tests usually indicates bad L2 cache timings quite fast: […]
Show full quote

Yoghoo wrote on 2024-02-02, 08:02:

Also increased the level 2 cache from 256kb to 512kb yesterday. Didn't see much of a speed increase from 15ns to 10ns but need to do some more testing. Anybody knows a good test program (besides cachechk) to test it?

My experience is that running these tests usually indicates bad L2 cache timings quite fast:

• Running Quake (e.g. the demo run in DOSBENCH)
• Running Doom (again, DOSBENCH provides an easy way)
• Booting Windows 2000
• Run Stunts 4D Sports car driving.

If all these tasks run without problems, the L2 timings are likely conservative enough for stable system operation. Faster cache RAM will not make your system run faster just by installing. The system only gets faster when you enable faster cache timings in the CMOS setup. A read burst of 2-1-1-1 and 0WS cache writes is the fastest possible setting. At FSB33, you should easily obtain that speed using 15ns RAM. At FSB40, timing is marginal. The upgrade to 512KB might make performance worse, because you lose bank interleaving, which makes the "-1-1-1" part harder to obtain. Furthermore, be very wary about 1024k x 8 DIP chips claiming to have 10ns timings. There are a lot of chips marked this way, but as far as the community knows, e.g. ISSI never produced their IS6C1024N chip at 10ns. They only packaged 10ns chips in SOJ. Any chip claimning to be an ISSI 128k x 8 chip at 10ns has a fake marking. Don't expect these "10ns" chips to deliver better performance than your current 15ns already do. When you order SOJ chips from chip resellers shipping from China, you can also get chips with fake speed markings. I got bitten by that: A made a PCB adapter to install 1MB cache into this board using SOJ chips I got (IIRC I ordered them at AliExpress) which didn't support what 10ns was supposed to support. I then checked response times using a suffiently fast scope and found out they behave like 15ns chips, so it wasn't surprinsing I couldn't get reliable 2-1-1-1 at 40MHz.

Nice information. Bought some IS61C1024-10N from Ali and system is running stable at FSB33 with 2-1-1-1 and 0WS at 133Mhz.

But when overclocking to 160Mhz I got problems with Windows 95. All DOS programs, games and benchmarks were okay. But starting Windows 95 setup gave a black screen with some random colored characters. And starting an already installed Windows 95 gave HD corruption issues. I blamed it on the FSB-to-PCI multiplier settings. But changing that to 1:2/3 didn't solve it. Will probably try some other L2 Cache Wait States in the future.

Yoghoo, start out with the most conservative timings, e.g. 3-2-2-2, DRAM read = 2 ws, then slowing improve upon them. You first need to ensure the slowest settings result in a stable system before pressing the accelerator to the floor.

feipoa wrote on 2024-02-03, 13:41:

Yoghoo, start out with the most conservative timings, e.g. 3-2-2-2, DRAM read = 2 ws, then slowing improve upon them. You first need to ensure the slowest settings result in a stable system before pressing the accelerator to the floor.

Thanks. Will try that. But will take some time as I am busy with some other retro restorations atm. 😀

feipoa wrote on 2024-02-03, 13:21:

Did you decide to shelf the 8433UUD cache interposer project? There were some 8 ns TSOP and SOJ modules from ISSI. I show a photo of the top/bottoms here: Re: Lets make new M919 Cache sticks? As they are from China, I don't know if they are authentic or not. The ISSI TSOPs work on the M919 1024K cache module.

I recently found that Mouser CY7C1009D-VXI chips in stock. I might place a Mouser order some time, and in that case, I will order two sets of nine of these (hopefully authentic) chips, and assemble another interposer.

feipoa wrote on 2024-02-03, 13:21:

Regarding 2-1-1-1 at 40 MHz. I recall pshipkov was able to obtain this setting by doing a lot of swapping around of DIP SRAM modules. See below:

pshipkov wrote on 2021-09-28, 03:06:

This last time it took few hours, but was doing other things as well. Order matters. Swapping chip in slot 0 with chip in slot 1 […]
Show full quote

This last time it took few hours, but was doing other things as well.
Order matters. Swapping chip in slot 0 with chip in slot 1 can be the difference between fully stable and flaky.
I don't have that many 128kx8 chips. Once i got into the retro optimize and overclock thing, i had to scrub some of the motherboards i have here from their chips.
The 10ns "fake" ones are good in my opinion.

I checked the timing of some of my "CY7C1009D" chips with the scope and didn't observe notable timing differences between those chips, and I also found that they generally do not comply to 10ns timing, so I wouldn't call the chips I have "good". This doesn't mean phipkov's chip are not good, or we might have a different oppinion about what is "good". Anyway, the next step is trying my PCB with original chips.

Yoghoo wrote on 2024-02-03, 13:30:

I blamed it on the FSB-to-PCI multiplier settings. But changing that to 1:2/3 didn't solve it.

My personal experience is that the 1:2/3 divider is sometimes causing extra issues. I had setups that worked at 40MHz 1:1, but failed at 1:2/3 if I remeber correctly. 1:2/3 generates an asymmetric clock signal, with it's high pulse as short as if you use the 1:1 settings. There may be situations in which 1:2/3 is useful, but if you have stability issues, I highly recommend to directly drop to 1:2 first, and only if that solves your issue, you might try whether 1:2/3 works in your setup.

Yoghoo wrote on 2024-02-03, 13:30:

mkarcher wrote on 2024-02-03, 12:57:

My experience is that running these tests usually indicates bad L2 cache timings quite fast: […]
Show full quote

Yoghoo wrote on 2024-02-02, 08:02:

Also increased the level 2 cache from 256kb to 512kb yesterday. Didn't see much of a speed increase from 15ns to 10ns but need to do some more testing. Anybody knows a good test program (besides cachechk) to test it?

My experience is that running these tests usually indicates bad L2 cache timings quite fast:

• Running Quake (e.g. the demo run in DOSBENCH)
• Running Doom (again, DOSBENCH provides an easy way)
• Booting Windows 2000
• Run Stunts 4D Sports car driving.

If all these tasks run without problems, the L2 timings are likely conservative enough for stable system operation. Faster cache RAM will not make your system run faster just by installing. The system only gets faster when you enable faster cache timings in the CMOS setup. A read burst of 2-1-1-1 and 0WS cache writes is the fastest possible setting. At FSB33, you should easily obtain that speed using 15ns RAM. At FSB40, timing is marginal. The upgrade to 512KB might make performance worse, because you lose bank interleaving, which makes the "-1-1-1" part harder to obtain. Furthermore, be very wary about 1024k x 8 DIP chips claiming to have 10ns timings. There are a lot of chips marked this way, but as far as the community knows, e.g. ISSI never produced their IS6C1024N chip at 10ns. They only packaged 10ns chips in SOJ. Any chip claimning to be an ISSI 128k x 8 chip at 10ns has a fake marking. Don't expect these "10ns" chips to deliver better performance than your current 15ns already do. When you order SOJ chips from chip resellers shipping from China, you can also get chips with fake speed markings. I got bitten by that: A made a PCB adapter to install 1MB cache into this board using SOJ chips I got (IIRC I ordered them at AliExpress) which didn't support what 10ns was supposed to support. I then checked response times using a suffiently fast scope and found out they behave like 15ns chips, so it wasn't surprinsing I couldn't get reliable 2-1-1-1 at 40MHz.

Nice information. Bought some IS61C1024-10N from Ali and system is running stable at FSB33 with 2-1-1-1 and 0WS at 133Mhz.

But when overclocking to 160Mhz I got problems with Windows 95. All DOS programs, games and benchmarks were okay. But starting Windows 95 setup gave a black screen with some random colored characters. And starting an already installed Windows 95 gave HD corruption issues. I blamed it on the FSB-to-PCI multiplier settings. But changing that to 1:2/3 didn't solve it. Will probably try some other L2 Cache Wait States in the future.

Most likely you have got 20ns chips. 15ns chips usually can do 2-1-1-1 @40MHz

CoffeeOne wrote on 2024-02-03, 18:19:

Most likely you have got 20ns chips. 15ns chips usually can do 2-1-1-1 @40MHz

... when you drive 2 banks.