VOGONS


Reply 20 of 37, by jakethompson1

User metadata
Rank Oldbie
Rank
Oldbie
Roman555 wrote on 2021-10-27, 19:01:

P.P.S. What will the Patcher do if the string *BBSS* is added into the BIOS manually ? 😀

The issue is, this BIOS is plain and uncompressed, i.e. it predates all these patching utilities. It isn't divided up into separate compressed sections with names like the later Award BIOSes are.
If the issue is the stray wrmsr instruction like I said, I suspect that using a hex editor to change the bytes at 0x19515 from 0x0f 0x30 to 0x90 0x90, and the checksum at 0x1ffff from 0x2a to 0x0b, and then reflash, that it would work. But, we haven't heard back from the OP about the post code or any further testing.

Reply 21 of 37, by Sphere478

User metadata
Rank l33t
Rank
l33t

Try talking to Jan

jan.jpg
Filename
jan.jpg
File size
6.15 KiB
Views
401 views
File license
Fair use/fair dealing exception

He has modded many bioses for k6 support

Last edited by Stiletto on 2021-10-29, 03:22. Edited 2 times in total.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 23 of 37, by Sphere478

User metadata
Rank l33t
Rank
l33t
majestyk wrote on 2021-10-28, 06:24:

I wonder if Jan is still active in this field.
He has some BIOSes as "work in progress" for a long time now.

I talked to him a few months back.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 24 of 37, by rmay635703

User metadata
Rank Oldbie
Rank
Oldbie
majestyk wrote on 2021-10-28, 06:24:

I wonder if Jan is still active in this field.
He has some BIOSes as "work in progress" for a long time now.

Likely not his fault.
That’s the fault of folks like me getting a bios for a special motherboard then not responding for a decade because said board developed difficult to repair issues and the bios test went by the wayside

Reply 25 of 37, by Repo Man11

User metadata
Rank Oldbie
Rank
Oldbie
majestyk wrote on 2021-10-28, 06:24:

I wonder if Jan is still active in this field.
He has some BIOSes as "work in progress" for a long time now.

He recently created an account here, though I can't recall his handle at the moment.

"Everyone is ignorant, only on different subjects." - Will Rogers

Reply 27 of 37, by zyga64

User metadata
Rank Member
Rank
Member

@Sphere478
I think Jan's email shouldn't be published here in non-obfuscated form.

1) VLSI SCAMP /286@20 /4MB /CL-GD5422 /CMI8330
2) i420EX /486DX33 /16MB /TGUI9440 /YMF718+GUS
3) i440BX /P!!!750 /256MB /MX440 /SBLive!+Vibra16s

Reply 30 of 37, by Chkcpu

User metadata
Rank Member
Rank
Member

Okay, you are calling on me to join this thread and I’m happy to share my views on this Biostar socket 5 BIOS issue.

Yes, I recently joined this fine community and find you all very knowledgeable in a wide range of retro PC subjects. I hope I can contribute with my specific knowledge of the Award v4.50/v4.51 BIOSes.

I started patching BIOSes for better CPU support in 1997 when my 486 rig wouldn’t run an Am5x86-133, and I’ve mainly patched socket 3 and socket 7 Award BIOSes since.
I’ve only once tried to patch a socket 5 Award BIOS. This was for the Edom MP039 board, that runs on the same OPTi Python (546/822) chipset as the Biostar from the OP.

At the time, I found that patching this BIOS for proper AMD K6 support was not feasible. No MMX or AMD support, not even for the AMD K5 and therefore a lot of support functions for non-Intel CPUs where missing.

However, I find jakethompson1’s idea very appealing. Just remove any offending code for the K6-2, like the Write to MSR 2, so the AMD can at least boot. We have good utilities now that can fix any AMD CPU registers programming from DOS or Windows and optimize performance.

I will start analyzing the 08/31/94-OPTI-546-822-2A5UMB01-00 BIOS for the 1500PCT series and report back what I find.
@jheronimus, can you confirm this is the BIOS you are using? And please let us know at what POST code the BIOS hangs when using the K6-2. This will greatly help in finding the culprit. 😉

Kind regards,
Jan.

CPU Identification utility
The Unofficial K6-2+ / K6-III+ page

Reply 31 of 37, by jheronimus

User metadata
Rank Oldbie
Rank
Oldbie
Chkcpu wrote on 2021-10-28, 19:33:
I will start analyzing the 08/31/94-OPTI-546-822-2A5UMB01-00 BIOS for the 1500PCT series and report back what I find. @jheronimu […]
Show full quote

I will start analyzing the 08/31/94-OPTI-546-822-2A5UMB01-00 BIOS for the 1500PCT series and report back what I find.
@jheronimus, can you confirm this is the BIOS you are using? And please let us know at what POST code the BIOS hangs when using the K6-2. This will greatly help in finding the culprit. 😉

Kind regards,
Jan.

Hi, Jan! Great to see you here.

Yes, I'm using that BIOS.

The POST code is just C0.

Reply 32 of 37, by Stiletto

User metadata
Rank l33t++
Rank
l33t++
zyga64 wrote on 2021-10-28, 18:12:

@Sphere478
I think Jan's email shouldn't be published here in non-obfuscated form.

majestyk wrote on 2021-10-28, 18:34:

It´s at the bottom of Jan´s website, so I think at least it´s not his private email address.

Well... I fixed it anyways. 😀

"I see a little silhouette-o of a man, Scaramouche, Scaramouche, will you
do the Fandango!" - Queen

Stiletto

Reply 33 of 37, by jheronimus

User metadata
Rank Oldbie
Rank
Oldbie
jakethompson1 wrote on 2021-10-28, 00:45:

The issue is, this BIOS is plain and uncompressed, i.e. it predates all these patching utilities. It isn't divided up into separate compressed sections with names like the later Award BIOSes are.
If the issue is the stray wrmsr instruction like I said, I suspect that using a hex editor to change the bytes at 0x19515 from 0x0f 0x30 to 0x90 0x90, and the checksum at 0x1ffff from 0x2a to 0x0b, and then reflash, that it would work. But, we haven't heard back from the OP about the post code or any further testing.

So this actually worked:

SS7VLB.png
Filename
SS7VLB.png
File size
5.89 KiB
Views
378 views
File license
Public domain

I guess I am now a proud owner of a Super Socket 5 VLB system? Ran Quake 1.08 timedemo to test for stability and got 45.1FPS, just insane. K6INIT gives some error messages (and Quake crashes), going to investigate further.

Reply 34 of 37, by Chkcpu

User metadata
Rank Member
Rank
Member

Okay jheronimus, this is great news!
Patching out the WRMSR did the trick and jakethompson1’s analysis and fix were spot on. 😀

I didn’t complete my BIOS analysis yet, but so far this write to MSR 2 is the only offending instruction for the K6-2 I’ve found. So I think you are good to go with this patch.

In later socket 7 Award BIOSes, I still see this write to MSR 2 for the Pentium (MMX), but then there is extra code to bypass this WRMSR when the CPU is non-Intel.

Another curious thing about this early uncompressed Award socket 5 BIOS is that it is just a 486 BIOS with some Pentium and chipset support added. So the 486 detection routines are still present and the CPUID instruction that jakethompson1 found is only used to distinguish the i486DX2 from the i486DX2 WB Enhanced! 😉

@jheronimus, can you tell us what the BIOS boot screens now shows as CPU model and speed when booting the K6-2/400? Just curious.

Jan.

Last edited by Chkcpu on 2021-10-30, 15:09. Edited 1 time in total.

CPU Identification utility
The Unofficial K6-2+ / K6-III+ page

Reply 35 of 37, by maxtherabbit

User metadata
Rank l33t
Rank
l33t
jheronimus wrote on 2021-10-30, 02:49:
So this actually worked: […]
Show full quote
jakethompson1 wrote on 2021-10-28, 00:45:

The issue is, this BIOS is plain and uncompressed, i.e. it predates all these patching utilities. It isn't divided up into separate compressed sections with names like the later Award BIOSes are.
If the issue is the stray wrmsr instruction like I said, I suspect that using a hex editor to change the bytes at 0x19515 from 0x0f 0x30 to 0x90 0x90, and the checksum at 0x1ffff from 0x2a to 0x0b, and then reflash, that it would work. But, we haven't heard back from the OP about the post code or any further testing.

So this actually worked:

SS7VLB.png

I guess I am now a proud owner of a Super Socket 5 VLB system? Ran Quake 1.08 timedemo to test for stability and got 45.1FPS, just insane. K6INIT gives some error messages (and Quake crashes), going to investigate further.

Cool!

Reply 36 of 37, by jheronimus

User metadata
Rank Oldbie
Rank
Oldbie
Chkcpu wrote on 2021-10-30, 14:11:

@jheronimus, can you tell us what the BIOS boot screens now shows as CPU model and speed when booting the K6-2/400? Just curious.

Jan.

My AMD was actually configured for 266 originally, and the BIOS screen said Pentium 100. But when I switched to 400 it now says Pentium 16 😀

Reply 37 of 37, by Sphere478

User metadata
Rank l33t
Rank
l33t

Woohoo! Thanks Jan!!

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)