ChkCPU, I just tried the 2.21R bios on an m912 v1.7. It boots without any problems.
Is there anything I should look for ?

also, does anyone know what type/model/brand the original eeprom on the board is ? aside from AD,10-1 9451 on the top and 4C065-4 on the bottom, there is no clear indication of what it exactly is.

weedeewee wrote on 2022-01-14, 21:08:

ChkCPU, I just tried the 2.21R bios on an m912 v1.7. It boots without any problems.
Is there anything I should look for ?

also, does anyone know what type/model/brand the original eeprom on the board is ? aside from AD,10-1 9451 on the top and 4C065-4 on the bottom, there is no clear indication of what it exactly is.

weedeewee, thanks for the quick feedback. This is good to know, before I put in time to patch this BIOS for the Am5x86-133. 😉
Do you have real L2 cache and is it indicated correctly by this 2.21r BIOS?
And which CPU do you run on this M912 v1.7 now?

Jan

ChkCPU, it's currently running a 486DX2-66, Bios mentions the 256K cache on the boot summary

ChkCPU, either my board is dying or something is funky with the bios or both or... who knows but...

the 2.21r bios says in the bios summary 256k Cache.
pccheck mentions no L2 cache
system analyzer says 32k L2
cachechk says 32K

I switch back to the original bios
bios summary now says 32k cache
cachechk still says 32k cache.

😒

so I switch back to the 2.21r bios
bios summary still says 256k cache
and now the system no longer boots from hard drive.
WTF

edit:
switching back to the original bios, the system still boots from hdd, yet reports 32k cache
switching to the 2.21r bios, summary reports 256k cache, system boots from hdd and cachechk reports 32k cache
No idea what caused the system to fail boot earlier. I'm guessing some cmos setting due to switching between the bios version.
Now to figure out if the cache on my board has failed or some flaky contact.

weedeewee wrote on 2022-01-14, 22:53:

ChkCPU, either my board is dying or something is funky with the bios or both or... who knows but... […]
Show full quote

ChkCPU, either my board is dying or something is funky with the bios or both or... who knows but...

the 2.21r bios says in the bios summary 256k Cache.
pccheck mentions no L2 cache
system analyzer says 32k L2
cachechk says 32K

I switch back to the original bios
bios summary now says 32k cache
cachechk still says 32k cache.

😒

so I switch back to the 2.21r bios
bios summary still says 256k cache
and now the system no longer boots from hard drive.
WTF

edit:
switching back to the original bios, the system still boots from hdd, yet reports 32k cache
switching to the 2.21r bios, summary reports 256k cache, system boots from hdd and cachechk reports 32k cache
No idea what caused the system to fail boot earlier. I'm guessing some cmos setting due to switching between the bios version.
Now to figure out if the cache on my board has failed or some flaky contact.

I assume all the settings are jumpered correctly. Just to rule it out...
EDIT: if you use a CF it can switch to LBA instead of Large and not boot, but I didn't have that with my board (doesn't mean it's the same for everyone).

weedeewee wrote on 2022-01-14, 21:08:

ChkCPU, I just tried the 2.21R bios on an m912 v1.7. It boots without any problems.
Is there anything I should look for ?

also, does anyone know what type/model/brand the original eeprom on the board is ? aside from AD,10-1 9451 on the top and 4C065-4 on the bottom, there is no clear indication of what it exactly is.

My original chip was write once, markings were missing IIRC.
I bought some W27C512-45Z. Used many to test different BIOS versions.

Nexxen wrote on 2022-01-15, 08:47:

I assume all the settings are jumpered correctly. Just to rule it out...
EDIT: if you use a CF it can switch to LBA instead of Large and not boot, but I didn't have that with my board (doesn't mean it's the same for everyone).

To my knowledge, none of the jumpers have ever been changed on that board. I'll verify anyway, also will pull out all the cache chips and reseat'm.
and I'm using user sintech's ATA/ATAPI emulator, works great.

Nexxen wrote on 2022-01-15, 08:50:

My original chip was write once, markings were missing IIRC.
I bought some W27C512-45Z. Used many to test different BIOS versions.

That's the exact same chip I have stuck in there right now with the 2.21R bios on it.

I'm going to verify the v1.4 board I have though that one does have some floppy read & battery drain problems. 😒

weedeewee wrote on 2022-01-15, 09:28:

To my knowledge, none of the jumpers have ever been changed on that board. I'll verify anyway, also will pull out all the cache […]
Show full quote

Nexxen wrote on 2022-01-15, 08:47:

I assume all the settings are jumpered correctly. Just to rule it out...
EDIT: if you use a CF it can switch to LBA instead of Large and not boot, but I didn't have that with my board (doesn't mean it's the same for everyone).

To my knowledge, none of the jumpers have ever been changed on that board. I'll verify anyway, also will pull out all the cache chips and reseat'm.
and I'm using user sintech's ATA/ATAPI emulator, works great.

Nexxen wrote on 2022-01-15, 08:50:

My original chip was write once, markings were missing IIRC.
I bought some W27C512-45Z. Used many to test different BIOS versions.

That's the exact same chip I have stuck in there right now with the 2.21R bios on it.

I'm going to verify the v1.4 board I have though that one does have some floppy read & battery drain problems. 😒

I don't remember if you already did, can you post a pic of the board?

Nexxen wrote on 2022-01-15, 09:40:

weedeewee wrote on 2022-01-15, 09:28:

To my knowledge, none of the jumpers have ever been changed on that board. I'll verify anyway, also will pull out all the cache […]
Show full quote

Nexxen wrote on 2022-01-15, 08:47:

I assume all the settings are jumpered correctly. Just to rule it out...
EDIT: if you use a CF it can switch to LBA instead of Large and not boot, but I didn't have that with my board (doesn't mean it's the same for everyone).

To my knowledge, none of the jumpers have ever been changed on that board. I'll verify anyway, also will pull out all the cache chips and reseat'm.
and I'm using user sintech's ATA/ATAPI emulator, works great.

Nexxen wrote on 2022-01-15, 08:50:

My original chip was write once, markings were missing IIRC.
I bought some W27C512-45Z. Used many to test different BIOS versions.

That's the exact same chip I have stuck in there right now with the 2.21R bios on it.

I'm going to verify the v1.4 board I have though that one does have some floppy read & battery drain problems. 😒

I don't remember if you already did, can you post a pic of the board?

the m912? doubtful.
Which one, the v1.4 or the v1.7 ?

weedeewee wrote on 2022-01-15, 09:48:

Nexxen wrote on 2022-01-15, 09:40:

weedeewee wrote on 2022-01-15, 09:28:

To my knowledge, none of the jumpers have ever been changed on that board. I'll verify anyway, also will pull out all the cache […]
Show full quote

To my knowledge, none of the jumpers have ever been changed on that board. I'll verify anyway, also will pull out all the cache chips and reseat'm.
and I'm using user sintech's ATA/ATAPI emulator, works great.

That's the exact same chip I have stuck in there right now with the 2.21R bios on it.

I'm going to verify the v1.4 board I have though that one does have some floppy read & battery drain problems. 😒

I don't remember if you already did, can you post a pic of the board?

the m912? doubtful.
Which one, the v1.4 or the v1.7 ?

Mainly the 1.7, but both to tackle the issue together may be better.
When I had broken traces in the L2 area, it would hang on post (after displaying the size and "starting ms-dos") or show as 0 (zero) L2 in cache check and other tools.
L2 chips are 64Kb each + tag?
Strange behaviour to just show 32kb.

Nexxen, Chkcpu,

There's something fishy about the 2.21r bios. In the bios summary screen on bootup it always mentions 256k cache.

Now it also seems like the 1.4 board has either fake cache or problems with the cache, and the 1.7 board also has issues.
Sigh. 😒

edit: all speedsys results were with Bios 'Setup defaults' loaded.

Nexxen wrote on 2022-01-15, 11:18:

[Mainly the 1.7, but both to tackle the issue together may be better. When I had broken traces in the L2 area, it would hang on […]
Show full quote

[Mainly the 1.7, but both to tackle the issue together may be better.
When I had broken traces in the L2 area, it would hang on post (after displaying the size and "starting ms-dos") or show as 0 (zero) L2 in cache check and other tools.
L2 chips are 64Kb each + tag?
Strange behaviour to just show 32kb.

agreed, indeed strange behaviour.

edit: removed the cache chips from the 1.4 board and there's some bodge work been done on the pcb under three of the sockets.
edit2 : the cache chips on the 1.4 board didn't get warm, setting them on the 1.7 board also reports no cache. assuming those are fake or dead. tag chip did work.

Quick reply, try 64 / 128 L2, swap chips. I had one not working and it was giving wrong results.

Nexxen wrote on 2022-01-15, 13:08:

Quick reply, try 64 / 128 L2, swap chips. I had one not working and it was giving wrong results.

? should I set the jumpers for 64 or 128 ?
or
use larger chips?

I swapped the chips around already. don't have any larger chips.
anyway, quick reply unclear 😀
currently lacking a desoldering iron to remove the sockets and fix the traces on the v1.4 board
visual inspection of v1.7 board doesn't show any immediatly visible damage to any traces.

edit: looking at the jumpersettings for the v1.7 board, only thing different seems to be JP36, which is currently open on mine, though is listed as factory default (closed) ...

https://www.ultimateretro.net/en/motherboards/573

weedeewee wrote on 2022-01-15, 13:19:

? should I set the jumpers for 64 or 128 ? or use larger chips? […]
Show full quote

Nexxen wrote on 2022-01-15, 13:08:

Quick reply, try 64 / 128 L2, swap chips. I had one not working and it was giving wrong results.

? should I set the jumpers for 64 or 128 ?
or
use larger chips?

I swapped the chips around already. don't have any larger chips.
anyway, quick reply unclear 😀
currently lacking a desoldering iron to remove the sockets and fix the traces on the v1.4 board
visual inspection of v1.7 board doesn't show any immediatly visible damage to any traces.

edit: looking at the jumpersettings for the v1.7 board, only thing different seems to be JP36, which is currently open on mine, though is listed as factory default (closed) ...

https://www.ultimateretro.net/en/motherboards/573

There is the manual, download Dx-6900. Set jumpers to 128kb.
Requires only 5 chips at a time.

After that you have the continuity test on the sockets but it takes forever and you need to know what to look for (I have a hand made drawing in case).

Nexxen wrote on 2022-01-15, 15:21:

https://www.ultimateretro.net/en/motherboards/573 […]
Show full quote

https://www.ultimateretro.net/en/motherboards/573

There is the manual, download Dx-6900. Set jumpers to 128kb.
Requires only 5 chips at a time.

After that you have the continuity test on the sockets but it takes forever and you need to know what to look for (I have a hand made drawing in case).

So, just set it to 128, gotcha.

I tried that just now with the chips I considered fakes or dead, and JP36 closed. and found at least one of the chips is broken. currently it's testing with 128k cache.

edit: of the 9 chips from the v1.4 board, two are not working.
to get the 256K cache working on the v1.7 board, JP36 needs to be closed, if JP36 is left open, only 32K cache will be seen.

edit2 : the 2.21R bios version always says 256K cache. Is this one of the fake cache mainboard bios's ?
The cache does work, it just reports the size always as 256K

edit3: add two speedsys results for good 256k cache, dx2-66 and dx4-100 and bios 2.21R

weedeewee,

Oh my, your L2 cache troubles revealed the DX-6900/M912 VER 2.21r BIOS to be a “fixed” BIOS for fake cache boards!!

I’ve been looking into the BIOS code for the past hour, comparing it with known good UMC498 BIOSes, and I believe I found the sneaky always 256KB Cache Memory patch! 😀
It is in the System Configurations display routines where they overwrote a call that fetches the measured L2 cache size and replaced it with a fixed value.
I can restore this “patch” to normal functionality.
Hopefully they were lazy and didn’t damage the cache detection code as well…

I will report back when I have fixed this fix. 😉

Jan

Chkcpu wrote on 2022-01-15, 21:03:

weedeewee, […]
Show full quote

weedeewee,

Oh my, your L2 cache troubles revealed the DX-6900/M912 VER 2.21r BIOS to be a “fixed” BIOS for fake cache boards!!

I’ve been looking into the BIOS code for the past hour, comparing it with known good UMC498 BIOSes, and I believe I found the sneaky always 256KB Cache Memory patch! 😀
It is in the System Configurations display routines where they overwrote a call that fetches the measured L2 cache size and replaced it with a fixed value.
I can restore this “patch” to normal functionality.
Hopefully they were lazy and didn’t damage the cache detection code as well…

I will report back when I have fixed this fix. 😉

Jan

Well, I was recently experimenting with an ata/atapi emulator on these boards, and was having a lot of annoying problems, one of them being the bios year problem and then you suggest that bios 😁
Which leads me to find and fix a few other problems, whilst at the same time causing me some other annoyances on the way.
Overall, I think it's been good.
the v1.4 board still has a few problems, 128k cache works, 256k not, floppy doesn't read (dma problem?) and rtc battery drains too quickly.
the v1.7 board now has 256k cache fully working (silly JP36) and bios support for two ide controllers and the year works.
one thing about the 2.21r bios, win95 detects a ps2 mouse port now, which isn't available on the m912 boards. (maybe possible to mod it ? )

just did a test on the 1.7 board with 2.21r bios by removing jp36 and running speedsys and 32k cache runs ok, as does 256k cache, so I guess detection is still ok.

Chkcpu wrote on 2022-01-15, 21:03:

weedeewee, […]
Show full quote

weedeewee,

Oh my, your L2 cache troubles revealed the DX-6900/M912 VER 2.21r BIOS to be a “fixed” BIOS for fake cache boards!!

I’ve been looking into the BIOS code for the past hour, comparing it with known good UMC498 BIOSes, and I believe I found the sneaky always 256KB Cache Memory patch! 😀
It is in the System Configurations display routines where they overwrote a call that fetches the measured L2 cache size and replaced it with a fixed value.
I can restore this “patch” to normal functionality.
Hopefully they were lazy and didn’t damage the cache detection code as well…

I will report back when I have fixed this fix. 😉

Jan

If you want to fix anything people here will love you forever 😀

weedeewee wrote on 2022-01-15, 21:18:

Well, I was recently experimenting with an ata/atapi emulator on these boards, and was having a lot of annoying problems, one of […]
Show full quote

Chkcpu wrote on 2022-01-15, 21:03:

weedeewee, […]
Show full quote

weedeewee,

Oh my, your L2 cache troubles revealed the DX-6900/M912 VER 2.21r BIOS to be a “fixed” BIOS for fake cache boards!!

I’ve been looking into the BIOS code for the past hour, comparing it with known good UMC498 BIOSes, and I believe I found the sneaky always 256KB Cache Memory patch! 😀
It is in the System Configurations display routines where they overwrote a call that fetches the measured L2 cache size and replaced it with a fixed value.
I can restore this “patch” to normal functionality.
Hopefully they were lazy and didn’t damage the cache detection code as well…

I will report back when I have fixed this fix. 😉

Jan

Well, I was recently experimenting with an ata/atapi emulator on these boards, and was having a lot of annoying problems, one of them being the bios year problem and then you suggest that bios 😁
Which leads me to find and fix a few other problems, whilst at the same time causing me some other annoyances on the way.
Overall, I think it's been good.
the v1.4 board still has a few problems, 128k cache works, 256k not, floppy doesn't read (dma problem?) and rtc battery drains too quickly.
the v1.7 board now has 256k cache fully working (silly JP36) and bios support for two ide controllers and the year works.
one thing about the 2.21r bios, win95 detects a ps2 mouse port now, which isn't available on the m912 boards. (maybe possible to mod it ? )

just did a test on the 1.7 board with 2.21r bios by removing jp36 and running speedsys and 32k cache runs ok, as does 256k cache, so I guess detection is still ok.

I was sure there was a wrong setting as the two boards were differently jumpered. Yes, the rigged bios for fake cache always reports 256K, and on my board the cache jumpers were hardwired on 256.
And sure that at least one non working cache chip.
After all the hours spent on it I learnt something 😀

Happy to see that something is moving on the bios side 😀