After trying lots of motherboards, I finally found an old Socket 7 with a 430TX chipset that works fine with the DVI output (MB550).

My hypothesis so far is that AMI BIOSes are perfectly compatible with this card's DVI output, while Award bioses are mostly problematic. Sometimes they POST and work except in the aforementioned situations/programs/games, and other motherboards don't even POST with this card.

I still don't know why the 5BT (2.0?) is not compatible with 2Mbit bioses.

tauro wrote on 2023-08-10, 10:14:

After trying lots of motherboards, I finally found an old Socket 7 with a 430TX chipset that works fine with the DVI output (MB550).

My hypothesis so far is that AMI BIOSes are perfectly compatible with this card's DVI output, while Award bioses are mostly problematic. Sometimes they POST and work except in the aforementioned situations/programs/games, and other motherboards don't even POST with this card.

I still don't know why the 5BT (2.0?) is not compatible with 2Mbit bioses.

According to Soyo this is the board revision mark...

...meaning its revision 2.0j

As far as the 2Mbit bioses go, are you programming externally as the files from theretroweb are intended for different bios chips - 2sa5 for an SST / Intel part and 2a5 for a Macronix MXIC part...not sure how this might impact on the issues you're seeing

That's very nice, what you did there zooming in. Thanks for that subtlety. So you're sure this is a 2.0 board.

PC Hoarder Patrol wrote on 2023-08-10, 15:42:

As far as the 2Mbit bioses go, are you programming externally as the files from theretroweb are intended for different bios chips - 2sa5 for an SST / Intel part and 2a5 for a Macronix MXIC part...not sure how this might impact on the issues you're seeing

Oh yeah, I'm using a TL866II+. I used MXIC EEPROMS since I don't have Intel ones. The board works fine with 128K bioses, so maybe this revision was designed with those in mind and 256K ones are incompatible?

Does anybody else have a 2.0 revision J to confirm?

tauro wrote on 2023-08-10, 21:37:

That's very nice, what you did there zooming in. Thanks for that subtlety. So you're sure this is a 2.0 board. […]
Show full quote

That's very nice, what you did there zooming in. Thanks for that subtlety. So you're sure this is a 2.0 board.

PC Hoarder Patrol wrote on 2023-08-10, 15:42:

As far as the 2Mbit bioses go, are you programming externally as the files from theretroweb are intended for different bios chips - 2sa5 for an SST / Intel part and 2a5 for a Macronix MXIC part...not sure how this might impact on the issues you're seeing

Oh yeah, I'm using a TL866II+. I used MXIC EEPROMS since I don't have Intel ones. The board works fine with 128K bioses, so maybe this revision was designed with those in mind and 256K ones are incompatible?

Does anybody else have a 2.0 revision J to confirm?

In a word, no (not sure at all!)

If I'm reading the old Soyo SY-5BT BIOS archive page correctly, it seems that the only determining factor between 1.x & 2.x is the pin count on JP30, so as your board pic shows 12 that makes it rev 2 and the revision letter thing comes from this FAQ link on the same page...

...so it seemed a safe assumption to say 2.0j (also mirrors the manual version)

However, there are a few wrinkles which complicate your situation...

...it appears that both rev 1.x and 2.x boards could come with either the 1Mb or 2Mb chips, so they state the correct file size is best determined by the flash app (not much help if the original chip is missing though)

...not all boards with the 'j' have the 12-pin JP30 header - here's one with 10, so rev 1.x???

...none of which helps you much further, sorry! (this is the archive page I'm quoting above https://web.archive.org/web/20000620230415/ht … ok=&faq=1&cpu=1 )

Last point, though can't see it helping - there's a 2Mb BIOS version 2a3 (attached), which theretroweb doesn't carry, that you might want to try anyway.

PC Hoarder Patrol wrote on 2023-08-11, 13:45:

tauro wrote on 2023-08-10, 18:37: That's very nice, what you did there zooming in. Thanks for that subtlety. So you're sure […]
Show full quote

tauro wrote on 2023-08-10, 18:37:
That's very nice, what you did there zooming in. Thanks for that subtlety. So you're sure this is a 2.0 board. […]
Show full quote

In a word, no (not sure at all!)

To my surprise your bios worked! 👏

Should I keep using the 1B7 (2/4/99) or the 2A3 (7/28/97)? Performance is identical with a Pentium 133 0.35µm, SY022. What's the benefit of choosing one over the other?

Both still have compatibility problems when using the Radeon's DVI out. The only ones I've had success with are the AMI bioses (different motherboards).

PC Hoarder Patrol wrote on 2023-08-11, 13:45:

...so it seemed a safe assumption to say 2.0j (also mirrors the manual version)

Exactly, the manual describes 2.0 as J

PC Hoarder Patrol wrote on 2023-08-11, 13:45:

...not all boards with the 'j' have the 12-pin JP30 header - here's one with 10, so rev 1.x???

So there also exists a 1.0 J 😮

Thanks for taking the time to find this BIOS! 😎
Where did you find it by the way? on the archived Soyo website? Is there a 2A4?

tauro wrote on 2023-08-13, 04:59:

To my surprise your bios worked! :clap: […]
Show full quote

PC Hoarder Patrol wrote on 2023-08-11, 13:45:

tauro wrote on 2023-08-10, 18:37: That's very nice, what you did there zooming in. Thanks for that subtlety. So you're sure […]
Show full quote

tauro wrote on 2023-08-10, 18:37:
That's very nice, what you did there zooming in. Thanks for that subtlety. So you're sure this is a 2.0 board. […]
Show full quote

In a word, no (not sure at all!)

To my surprise your bios worked! 👏

Should I keep using the 1B7 (2/4/99) or the 2A3 (7/28/97)? Performance is identical with a Pentium 133 0.35µm, SY022. What's the benefit of choosing one over the other?

Both still have compatibility problems when using the Radeon's DVI out. The only ones I've had success with are the AMI bioses (different motherboards).

PC Hoarder Patrol wrote on 2023-08-11, 13:45:

...so it seemed a safe assumption to say 2.0j (also mirrors the manual version)

Exactly, the manual describes 2.0 as J
manual.png

PC Hoarder Patrol wrote on 2023-08-11, 13:45:

...not all boards with the 'j' have the 12-pin JP30 header - here's one with 10, so rev 1.x???

So there also exists a 1.0 J 😮

Thanks for taking the time to find this BIOS! 😎
Where did you find it by the way? on the archived Soyo website? Is there a 2A4?

Glad it worked 😀 (of sorts!) - found it only on the German Soyo archive (no 2A4 🙁 ), though none of the 5BT pages there or on the archives of Soyo USA or Taiwan ever reference it so no changelog to work with (maybe an early 2Mb version or beta?)

There are a number of detailed 2.x settings sheets, though suppose that could be later board revision or manual based...

If it's no better or worse than 1B7 then choice is a toss-up really - maybe run then both for extended periods and see if any instabilities emerge.

Thanks for those "one sheet manuals". They're handy. Maybe those are manual revisions, and not different board revisions.

About the bioses, all the benchmarks I run (SysInfo, DrHardware, 3Dbench, Doom, NSSI, etc) give identical results, save for two exceptions.
NSSI: 1B7 = 94435 while 2A3 (as well as 1B2) = 104150.
And 1B7 is faster on IDE transfers (10819KB/s vs. 10414KB).

And that's all, I'll do more tests in the future.

For DOS usage, all BIOS are practically interchangeable. I think the difference may become apparent if you use Windows + certain problematic hardware.

tauro wrote on 2023-08-10, 10:14:

I still don't know why the 5BT (2.0?) is not compatible with 2Mbit bioses.

Hi Tauro, I have some info about the 5BT BIOSes.

The SY-5BT board is compatible with both 1Mbit and 2Mbit BIOSes, but each BIOS version only works on a limited number of EEPROM chips. So if you have boot problems or hangs, or a scrambled AWARDEPA logo like in your first message, the replacing EEPROM chip you used is not compatible with the BIOS.

Looking into the 2Mbit BIOSes, I noticed the extra space is used for additional SCSI option ROMs only. So if you don’t use a ROMless SCSI adaptor, you can use any 1Mbit or 2Mbit 5BT BIOS version.

Should I keep using the 1B7 (2/4/99) or the 2A3 (7/28/97)? Performance is identical with a Pentium 133 0.35µm, SY022. What's the benefit of choosing one over the other?
Both still have compatibility problems when using the Radeon's DVI out. The only ones I've had success with are the AMI bioses (different motherboards).

I believe the 2A3 BIOS works because it is so old it doesn’t need to write any small blocks of data into the EEPROM. Only the large ESCD block for storing PnP and Power Management data is used by this 2A3 BIOS, making it compatible with a wide range of EEPROM chips.
However, this 2A3 BIOS has several limitations:
- AMD CPU support is limited to the K5 and K6 model 6 (3.2V Vcore). The 2.2V K6 model 7, K6-2, and all later models are not supported. Also the Cyrix 6x86MX and IDT Winchip CPUs are not supported.
- HDD support via the on-board IDE channels is limited to 8GB.
- DMA and UDMA 33 on UDMA 66 and faster drives is not support under Win9x.

Note that the other 5BT BIOS versions do support larger HDDs, but are still limited to 32GB except the 128GB patched versions.

The latest, and in my opinion best, 5BT BIOS is the 1Mbit 03/15/2000 Rel 2.2 BIOS upgrade by Unicorn Software. It is on TheRetroWeb.
This BIOS has full IDT, Intel, Cyrix, and AMD CPU support including the K6-2+/III+. It has several other improvements like boot from CDROM and support for HDDs up to 128GiB. And of course it has a fix for the Windows 98 UDMA bug. 😉
Being a more recent March 2000 BIOS, it should support a wider range of EEPROM chips as well.

Regards, Jan

Thank you very much Jan for all the detailed information about the bioses.

So apparently my 2Mbit chips are incompatible with most of them except for 2A3.

I tried the suggested 1Mbit BIOS by Unicore Software. When I enter the BIOS menu it freezes after a few seconds, so evidently there's something not quite right with it.

I wonder if the Win98 UDMA bug is fixed in 1B7.

It’s a pity that the Unicore BIOS didn’t work. 🙁

I’ve checked the 1B7 (02/04/99) BIOS and yes, it has a fix for the Win98 UDMA bug. Most 03/1999 and later Award BIOSes have this fix, but some as early as 11/1998 have it as well and the 1B7 is one of them. 😀

The CPU support in this 1B7 BIOS looks fine also: Pentium, Pentium MMX, K5, K6, K6-2, K6-2CXT, K6-3, Cyrix 686/686MX/MII, IDT Winchip, and Rise mP6 support. Only K6-2+/K6-III+ support is missing.

IDE HDD support is limited to 32GB, but there is a patched and tested version with 128GB support. This patched BIOS can be found at http://wims.rainbow-software.org/
Download the “Soyo SY-5BT 128KB.zip” file. It contains the same 1B7 BIOS but with a fix for the 32GB and both 64GB limit bugs.

Cheers, Jan

Chkcpu wrote on 2023-08-21, 14:06:

I’ve checked the 1B7 (02/04/99) BIOS and yes, it has a fix for the Win98 UDMA bug. Most 03/1999 and later Award BIOSes have this fix, but some as early as 11/1998 have it as well and the 1B7 is one of them. 😀

Thank you so much for taking the time to check that! You're my favorite BIOS hacker 👏

Chkcpu wrote on 2023-08-21, 14:06:

IDE HDD support is limited to 32GB, but there is a patched and tested version with 128GB support. This patched BIOS can be found at http://wims.rainbow-software.org/
Download the “Soyo SY-5BT 128KB.zip” file. It contains the same 1B7 BIOS but with a fix for the 32GB and both 64GB limit bugs.

This one worked fine! I'll be using it.

Chkcpu wrote on 2023-08-21, 14:06:

It’s a pity that the Unicore BIOS didn’t work. slightly frowning face

By chance, I discovered that an SDRAM module was the culprit. So... the Unicore BIOS also works! Although there's at least one downside.
Data transfer rate in DOS is about 36% slower, Dr. Hardware reports 6885KB/Sec. vs. 10819 KB/Sec. (1B7/1B7+). Also CPU performance is slightly lower in that same program 105535 vs. 112383 Hardstones.

Another thing to note is that BIOSes 1B2 and 2A3 yielded a better 16-bit integer performance in NSSI (104150 vs. 94435 Dhrystones with all the other BIOSes).

Other than that, performance is roughly the same with all the working BIOSes in DOS.

I also noticed that Memory Bandwidth in SpeedSys sometimes varies, it's usually 223 MB/s but sometimes it goes down to 200 MB/s after a soft reset, I don't know why.

For all I could test, 1B7+ is the most appropriate version, except in the instances you detailed above (romless SCSI, K6-2/3+).
Thanks for all your suggestions and for casting a light on this topic. 🔎

Speedsys benchmark with 1B7+ (from wims.rainbow-software.org)

For the Radeon 9250 DVI output I found a different board that does the job (M550). As I mentioned AMI BIOS is the key for compatibility.

Hi tauro,

Thanks for your detailed feedback and good to know you selected the 1B7+ BIOS for this project!

Also nice to hear the Unicore BIOS worked after all, albeit with lower performance. I don’t see any obvious differences in the CMOS Setup Default values between the 1B7/1B7+ and the Unicore BIOS , so the Unicore BIOS probably uses more conservative settings when programming the chipset registers during POST.
These register tables can be tuned as well, but this needs investigating. Let me know if you ever going to run a K6-2+/K6-III+ on this board and I will look into this. 😉

Cheers, Jan