I would love your Sourcer config Chkcpu 😀

Chkcpu wrote on 2024-10-18, 09:02:

Hi dm-, […]
Show full quote

Hi dm-,

Thanks for joining the patched BIOS test team!
Nice to hear the K6-III+ is able to run on this P5SD-A board with the help of apple_rom’s BIOS patcher tool.

I will send you a PM shortly with the BIOS I patched, and some test instructions.

Cheers, Jan

Hello Jan, here is some results.
Can't attach more than 1 photo in PM so posting here.

1. CPU has been detected correctly. 83x6=500Mhz
2. HDD 80gb has been detected correctly, although, modes was set wrong (LRG) and UDMA6 (on ss7?!?). This is probably not because of you mod, but the way how original bios detecting drives.
Set mode LBA manually in bios and then the drive work fine.
3. Write allocation may be enabled by k6init, but not a setk6 util. anyway, it's working.

Meanwhile I have a Soyo SY-5XA, can I beta test the bios you made ? 😁

dm- wrote on 2024-10-19, 05:37:

Hello Jan, here is some results. Can't attach more than 1 photo in PM so posting here. […]
Show full quote

Chkcpu wrote on 2024-10-18, 09:02:

Hi dm-, […]
Show full quote

Hi dm-,

Thanks for joining the patched BIOS test team!
Nice to hear the K6-III+ is able to run on this P5SD-A board with the help of apple_rom’s BIOS patcher tool.

I will send you a PM shortly with the BIOS I patched, and some test instructions.

Cheers, Jan

Hello Jan, here is some results.
Can't attach more than 1 photo in PM so posting here.

The attachment photo_5231124790931221253_y.jpg is no longer available

The attachment photo_5231124790931221252_y.jpg is no longer available

The attachment photo_5231124790931221254_y.jpg is no longer available

1. CPU has been detected correctly. 83x6=500Mhz
2. HDD 80gb has been detected correctly, although, modes was set wrong (LRG) and UDMA6 (on ss7?!?). This is probably not because of you mod, but the way how original bios detecting drives.
Set mode LBA manually in bios and then the drive work fine.
3. Write allocation may be enabled by k6init, but not a setk6 util. anyway, it's working.

Hello dm-,

Thanks for testing the P5SD-A patch J.2 BIOS.
With your K6-III+ CPU, the patches for CPU detection, CPU speed, and K6 Write-Allocation work well indeed. 😀

The LARGE (LGR) mode on the 80GB SATA SSD is a strange one. The BIOS doesn’t detect LBA support of this drive and uses the next translation mode it supports. This Large mode however works only up to 1GB on the Award BIOS and is meant for old IDE drives without LBA support.
I’ve seen this LGR mode issue once before, on a Seagate Momentus 40GB SATAIII HDD from 2010, and thought this was a firmware fault. But now I think that on these modern drives, it is probably a depreciation of the fields returned by the “identify device” command that this 1999 BIOS looks at for LBA support.
Luckily, the Award BIOS allows to set the Drive mode to LBA manually, so this is easily corrected.

The UDMA 6 indication is indeed a BIOS bug on 1998/1999 Award BIOSes, know as the Win98 UDMA bug. This bug causes the BIOS to report the maximum transfer speed of the drive, instead of the actual UDMA mode on this UDMA 2 chipset.
Win98(SE) considers this high UDMA mode as invalid and doesn’t allow any DMA mode when enabled in Device Manager. This causes Windows to fallback to the slower POI mode 4, to protect your data.

When I made this patched BIOS 2 years ago, I didn’t have a fix for this bug, but I do now!
I will make a new patch J.3 BIOS with this UDMA bug fixed and send it over when ready.

Then I see a curious distortion in the PCI device listing on the BIOS System configuration page.
Was this a one time fluke, or a constant double line there? If so, do you get the same with another, or PCI instead of AGP, videocard?

Greetings, Jan

Deksor wrote on 2024-10-19, 10:36:

Meanwhile I have a Soyo SY-5XA, can I beta test the bios you made ? 😁

Hi Deksor,

Yes, please!
But I'm away from home for the rest of the day and evening.
I will send you the patched SY-5XA BIOS by PM tomorrow. 😀

Cheers, Jan

I have this motherboard, is it possible improve 2.36 bios or it is already up to date? I think they tweak some settings to improve performance out of the via mvp3 default (lower performance settings), what i have not done. Just Memory interleave enabler for VIA chipsets is enabled @ Win9x OS.

https://www.philscomputerlab.com/aopen-ax59-pro.html
https://www.georgebreese.com/net/software/

rasz_pl wrote on 2024-10-19, 02:03:

I would love your Sourcer config Chkcpu 😀

Hi rasz_pl,

Here are the files I used for the disassembly by Sourcer of the Elitegroup P5SD-A Ver. 1.4a BIOS.

P5SDA.IMG is a copy of the System BIOS component original.tmp from the 1.4a BIOS.
BIOS.DEF is the Sourcer definition file for this BIOS.
P5SDA.LST is the disassembly listing produced by Sourcer from the above 2 files.
CHANGES.TXT documents all the changes I made for the patch J.2 version of this BIOS.

Note that I did not fully resolve all code and data for this disassembly. I limited the entries in the definition file to the sections I needed to resolve for the K6plus and 128GB patches.

Let me know if you like to see other details.

Cheers, Jan

Chkcpu wrote on 2024-10-20, 17:01:

Here are the files I used for the disassembly by Sourcer of the Elitegroup P5SD-A Ver. 1.4a BIOS.

Thank you!

Chkcpu wrote on 2024-10-20, 17:01:

BIOS.DEF is the Sourcer definition file for this BIOS.

That looks very manual labor intensive. Am I correct in assuming you run the disassembly, make notes and annotations for BIOS.DEF, run disassembly again with those annotations taken into account, rinse repeat?
Im really unhappy with some of IDA quirks, but love the interactivity. Im yet to try Ghidra.

Yes, making a Sourcer definition file for a particular BIOS is very labor intensive and takes me usually 3 to 4 hours. Being a static disassembler, a lot of “rinse and repeat” cycles are indeed required to get a satisfactory result. Only a small part of BIOS.DEF is the same for every BIOS and can be used as a template.

Having used Sourcer for so long, it will take quite some effort to switch to a modern interactive disassembler. It will happen no doubt, I just have to invest the time! 😉

I will be interested to hear what you think of Ghidra, compared to IDA. Please send me a PM about Ghidra when you have some experience to share.

Greetings, Jan

Hi guys,

Had a thought that... I might be able to do on an individual basis to some extent, but you guys would be the experts on whether it's doable on a wider scale.
As you're aware, the various LBA limits require manual patching.
XTIDE can get around this.
XTIDE can, with some effort, be integrated into a BIOS as an oprom module.
Adding XT-IDE option ROM to Asus P3B-F BIOS [Thanks to DenizOezmen, it actually works!!!]

Do you think a generic module could be made for - at worst - each bios brand?
Do you think it might be possible to script like the Skylake mods were?

Ie, it would be nice to have it to the point where you guys are like:
"Ok, run this script, check that it opens in XX software (is x86box an option for a boot test?), flash, and it *should* work."
Instead of having to manually patch each one.

Thoughts?

myne wrote on 2024-10-21, 04:03:

Do you think a generic module could be made for - at worst - each bios brand?
Do you think it might be possible to script like the Skylake mods were?

There is a tool called "BIOS Patcher" made by rom.by community founder, apple_rom_.
English version of the README file: https://www.rom.by/articles/BP/index_english.htm
You could try versions 4.23 and 6.00 alpha15.
Here is a BIOS Patcher review by scorp: https://www.youtube.com/watch?v=VwZT40sRMzM

Yeah, it didn't detect the HDD limits on another guy here's bios, so it didn't work, unfortunately.
Also that site is fairly broken these days. alpha 15 didn't work the other day when I tried it.

I was wondering if it is possible to patch Shuttle HOT-566 (https://theretroweb.com/motherboards/s/shuttle-hot-566) for faster AMD K6 CPUs. I am running latest BIOS with K6-2 450 (with 75Mhz FSB) but the BIOS reports incorrect CPU frequency and I wonder whether it could be affecting other stuff, since sometimes I get an NMI. I tried different memory and it didnt help so I think it is not coming from memory, but maybe BIOS programs something wrong when it doesn't detect CPU frequency correctly. I see other shuttle BIOSes on the unofficial page but not HOT-566 ;(

0xDEADBEEF wrote on 2024-10-23, 00:26:

I was wondering if it is possible to patch Shuttle HOT-566 (https://theretroweb.com/motherboards/s/shuttle-hot-566) for faster AMD K6 CPUs. I am running latest BIOS with K6-2 450 (with 75Mhz FSB) but the BIOS reports incorrect CPU frequency and I wonder whether it could be affecting other stuff, since sometimes I get an NMI. I tried different memory and it didnt help so I think it is not coming from memory, but maybe BIOS programs something wrong when it doesn't detect CPU frequency correctly. I see other shuttle BIOSes on the unofficial page but not HOT-566 ;(

Curiously, I never had a HOT-566 BIOS patch request before, until now.
That’s why it’s not on my Unofficial K6plus webpage yet. 😉

Yes, I can patch the latest 566WAQ0D BIOS for better K6-2(+)/III(+) support and fix the 32GB IDE HDD limit bugs as well. However, I have several other BIOS projects to finish first so it may take a while…

Looking into this 566WAQ0D BIOS, I found that the x6 multiplier isn’t supported for any FSB. That’s why you get strange speed readings when running your K6-2 at 6x75. This may also cause the busspeed calculation in the BIOS to fail, resulting in incorrect programming of the chipset timing registers for memory and L2 cache. So this setting may indeed cause instabilities.

You can try to lower the multiplier to x5.5 for a 412MHz CPU speed. This 5.5x75 setting is supported by the BIOS and should run stable if the instability was caused by the not supported x6 multiplier. In fact, all Multiplier/FSB combinations of x2.5, x3.0, x3.5, x4.0, x4.5, x5.0, x5.5 and FSB 50, 60, 66, 75, and 83MHz are supported, except for the x5.5/83 setting.

If you still experience an NMI, even at a 5.5x66 setting, then the BIOS is probably not the cause.
Do you use an ISA SoundBlaster soundcard? If so, remove this card and see if that fixes the NMI errors. Depending on the configuration of these SB cards, they can produce these NMI errors.

Greetings, Jan

Hah, I do have ISA soundblaster. SB AWE 64 Gold. And interestingly enough NMI often happens when running aweutil to initialize it.
I will switch to 5.5x75 and try 66MHz as well to see if NMI still happens.
Any idea if there is any fix/setting for sounblaster NMI issues? This soundcard is one of the main reasons I am building this computer to begin with (it has GUS + SB AWE64 for demos and games and modules).

Hmm it looks like I need to run aweutil /s or remove MFBEN jumper on the card. Apparently Midi emulation uses NMI. aweutil is supposed to handle them.

0xDEADBEEF wrote on 2024-10-23, 21:57:

Hmm it looks like I need to run aweutil /s or remove MFBEN jumper on the card. Apparently Midi emulation uses NMI. aweutil is supposed to handle them.

Indeed, AWEUTIL plays an important role here.

There are loads of message when you search for “aweutil” here on Vogons. An interesting example is Re: DOS32AWE - DOS4GW compatible DOS Extender with Sound Blaster AWEUTIL MIDI emulation support for Protected mode

Cheers, Jan

Hi! I'd like to report an anomaly with the modded 3.0 J.1 BIOS for the QDI Advance 3 P5MVP3/A3 motherboard at http://www.steunebrink.info/k6plus.htm . I have two different K6-2+ CPUs for testing and they act slightly different:

• AMD K6-2+ 400ATZ 1.6v

• AMD K6-2+ 475ACZ 2.0v

The difference is that the 1.6v version doesn't POST on cold boot. It just doesn't go anywhere and nothing on screen. But when I immediately restart the computer it acts normally. The next day, it's the same thing again. The 2.0v version doesn't seem to have this problem. It boots up normally. Jan Steunebrink notes that the 1.6v version should work with the 2.0v BIOS setting for CPU core voltage (it doesn't go any lower) and it does seem to work apart from this anomaly. It's like it needs a shot of current before it is willing to work.

Hi Cyanopsis,

Thanks for the report. This is a curious anomaly indeed!

As far as I know, all K6-2+/III+ CPUs were made with an 180nm process, opposed to the 250nm die structures used in the original K6-2/III.
So the lower 1.6V/1.7V/1.8V K6-2+/III+ models must be binned parts, meant for fanless / embedded applications. That’s why they will run fine at 2.0V, just like other K6-2+/III+ models, but will need a cooler with fan then.

Thinking about a possible explanation for the boot anomaly of the 1.6V K6-2+, maybe these binned parts are more susceptible to the higher start Vcore of this jumperless board.
I’ve been looking at how these jumperless socket 7 boards control the initial CPU voltage before the reset line goes low and the CPU starts executing code.
Without jumpers or dipswitches to set the voltage for the CPU, and with the wide range of required voltages for various S7 CPUs, you need the motherboard to provide a sufficient high voltage for 2.8V/2.9V Vcore CPUs to start reliably but not too high for 2.1V/2.2V models to get damaged.

Investigating this for the jumperless QDI Titanium B boards with SpeedEasy control, I found an initial 2.6V Vcore for Dual-voltage CPUs and 3.4V for single-voltage parts.
For dual voltage CPUs this 2.6V Vcore is sufficient to start higher Vcore models and not too high to damage low Vcore parts. After a few milliseconds the BIOS will take over and set the correct value for the installed CPU.

I will look into the SpeedEasy logic of the QDI Advance 3 patch J.1 BIOS to see how the Vcore control is done for a cold start. Maybe the Vcore is held high too long for the 1.6V K6-2+ to start properly.

I will report back when I have more.
Cheers, Jan

Great! I'm ready if you need someone to do benchmarks!

Another thing, not related to the 1.6v CPU but for the patched BIOS in general (maybe only for this board): entering BIOS with DEL needs precision and timing. If I press any button too quickly (not just DEL but F8 also), the computer restarts. It needs to be done just when the memory test starts. Could that be related to what you wrote above?