LSS10999 wrote on 2026-02-10, 10:00:

AFAICT doing BSEL mod on the CPU will not work with this board. It needs to be done on the motherboard by adding a wire to alter the BSEL value passed to the NB so it'll configure the FSB to 800. I've put that in the OP.

Yeah this is what I was referring to. Just a simple wire. The X6800 has a 11x multi which ends up being 2.2Ghz @ 200Mhz. I'm hoping that it can get to 16x or over.

Here's a benchmark from @agent_x007 and his X6800/HD 3850. Different mobo/chipset and DDR2 ram though.

But you can see slightly better CPU performance at only 3.3ghz and much better GPU performance compared to the benchmark I posted previously. I'm guessing the extra GPU performance could be due to the 4MB L2 cache. Or maybe it could be due to DDR2, or the mobo chipset. Who knows.

Either way, it's compelling me to try an X6800. I would really like to have proper shutdown.

I just did an inspection across V008, V009 and V00C regarding PCI IRQ Routing table using AMIBCP v3.13.

It seems V008 had two entries related to PCI Bus 01h which is probably the AGP section. V009 had one, while V00C had none. This probably explains why AGP slot is kind of broken with V00C.

However, it appears that I could not add new entries into the table directly from AMIBCP. Only modify existing, or removing it using "Cut" in the right click menu.

EDIT: Here's the particular PCI IRQ Routing line in question. The columns are comma-separated.

1PCI Bus,Dev.#,IntA Reg,IntA Bitmap,IntB Reg,IntB Bitmap,IntC Reg,IntC Bitmap,IntD Reg,IntD Bitmap,Phys.Slot
201,00,60,DCF8,00,0000,00,0000,00,0000,01

The PCI IRQ Routing table of V00C is otherwise identical to that of V009, just that the line above regarding PCI Bus 01h is absent. On the other hand, on all BIOS versions, there's a line at the end of the routing table that's filled with all zeros. Since I never really modified PCI IRQ Routing table for an AMI BIOS I'm not really sure if this all-zero line can be used for the purpose of adding back the missing entry so the AGP slot could work.

I'm not sure which AMIBCP version works safest with our board's BIOS, that is, ensuring a good chance the edited BIOS can boot without major issues.

LSS10999 wrote on 2026-02-10, 13:14:

I just did an inspection across V008, V009 and V00C regarding PCI IRQ Routing table using AMIBCP v3.13. […]
Show full quote

I just did an inspection across V008, V009 and V00C regarding PCI IRQ Routing table using AMIBCP v3.13.

It seems V008 had two entries related to PCI Bus 01h which is probably the AGP section. V009 had one, while V00C had none. This probably explains why AGP slot is kind of broken with V00C.

However, it appears that I could not add new entries into the table directly from AMIBCP. Only modify existing, or removing it using "Cut" in the right click menu.

EDIT: Here's the particular PCI IRQ Routing line in question. The columns are comma-separated.

Copy code to clipboard

1PCI Bus,Dev.#,IntA Reg,IntA Bitmap,IntB Reg,IntB Bitmap,IntC Reg,IntC Bitmap,IntD Reg,IntD Bitmap,Phys.Slot
201,00,60,DCF8,00,0000,00,0000,00,0000,01

The PCI IRQ Routing table of V00C is otherwise identical to that of V009, just that the line above regarding PCI Bus 01h is absent. On the other hand, on all BIOS versions, there's a line at the end of the routing table that's filled with all zeros. Since I never really modified PCI IRQ Routing table for an AMI BIOS I'm not really sure if this all-zero line can be used for the purpose of adding back the missing entry so the AGP slot could work.

I'm not sure which AMIBCP version works safest with our board's BIOS, that is, ensuring a good chance the edited BIOS can boot without major issues.

I have no experience with BIOS modifications so keep that in mind.

I asked ChatGPT (Which should be taken with a grain of salt), it seems to indicate that you can utilize the last line (all zeros) to add the missing AGP mapping entry, but it may not work if the BIOS expects a terminator at the end of the routing table.

Another option would be to replace the first PCI slot (Phys slot #2) entry with the missing AGP entry. This would be the ideal scenario if your AGP GPU is a two slot card, which effectively renders the first PCI slot as defunct. However the HD3850 is a single slot card so the first PCI slot is still accessible.

Perhaps the better option would be to replace the last PCI slot (Phys slot #6) with the missing AGP entry. This slot is not really usable if you're utilizing two ISA cards, which is going to be true in my case.

If someone with a programmer is willing to test, I would try replacing the all zeros entry first. If that doesn't work, try replacing the PCI slots instead.

Just had another thought - shouldn't there be an entry for the onboard VGA? If so, replacing this with the missing AGP entry might be the way to go. The difficult part would be identifying it.

Are we certain the v009 doesn't have the IRQ routing issue? Personally I haven't tested it.

I found this AMIBIOS Rom user guide:
https://www.scribd.com/document/63494976/AMIB … ties-User-Guide

If you go to page 96 is has this note:

Which seems to indicate that you can edit the last entry, seeing as they are all zeros.

Also page 84 details which ROM version is compatible with v3.xx:

So it would appear that AMIBCP v3.13 is suitable, and that replacing the last PCI IRQ Routing entry could be possible. The question is, who has the cojones to try it out? Disclaimer: I have zero experience with BIOS modifications. I could be dead wrong here.

Shiftyy wrote on 2026-02-11, 01:36:

Just had another thought - shouldn't there be an entry for the onboard VGA? If so, replacing this with the missing AGP entry might be the way to go. The difficult part would be identifying it.

Are we certain the v009 doesn't have the IRQ routing issue? Personally I haven't tested it.

I couldn't remember whether it worked on V009. I only tested video cards on PCI slots like FirePro MV 2260, as well as PCIe FirePro W4100 over PCI-PCIe adapter. Actually, the V009 BIOS was from my 1.0s1.3 board which is 865GV. My 1.0s1.2 board had V008.

There were two changes between V008 and V009. The first line which was originally for PCI bus 01h was changed to 04h in V009, and the separate line originally for PCI bus 04h Device 40h in V008 got deleted.

Shiftyy wrote on 2026-02-11, 05:15:

I found this AMIBIOS Rom user guide: https://www.scribd.com/document/63494976/AMIB … ties-User-Guide […]
Show full quote

I found this AMIBIOS Rom user guide:
https://www.scribd.com/document/63494976/AMIB … ties-User-Guide

If you go to page 96 is has this note:

Which seems to indicate that you can edit the last entry, seeing as they are all zeros.

Also page 84 details which ROM version is compatible with v3.xx:

So it would appear that AMIBCP v3.13 is suitable, and that replacing the last PCI IRQ Routing entry could be possible. The question is, who has the cojones to try it out?

I don't think I would be trying this right now, as I'm not using any AGP video card on my 1.0s1.2 setup at the moment. My FirePro W4100 over PCI-PCIe adapter works very well in most use cases.

Though I do hope there's a usable solution to plug a PCIe video card into the AGP port, as that's the only way to actually disable onboard video. If the AGP port on 1.0s1.2 is not populated, it behaves the same as 1.0s1.3 where onboard video cannot be disabled even when a PCI video card is present.

The standoff arrived. I ordered 20mm, but received 18mm. Thanks Aliexpress.

I made it work:

I can now have the case upright with the side-panel on. Yay.

My Nvidia FX 5500 PCI card arrived today. The intention is to use the FX 5500 for DOS/98 gaming and the HD 3850 for XP gaming.

I then spent 10 hours troubleshooting GPU initialization priority. Setting it to "PCI\AGP" didn't work. Neither did "AGP\PCI" or "PCI\Int-VGA".

It would always prioritize the AGP card no matter what, except for "PCI\Int-VGA" which prioritized PCI but disabled AGP, as you would expect.

I spent considerable time in IDA, HxD and Sourcer, trying and manually patch the bios. Assembly code is not my forte so that wasn't fun.

After flashing the bios a few times, I suddenly noticed that the PCI card was showing an image. When you flash/update the bios, it will load the optimized default settings. The default setting for "Init. Graphic Adapter Priority" is "AGP\Int-VGA".

I was like "Huh? Shouldn't the PCI card be completely disabled???"

Well. Yeah. So, setting "Init. Graphic Adapter Priority" to "AGP\Int-VGA" will actually prioritize the PCI card, whilst keeping AGP enabled, the exact behaviour I am looking for. In other words, "AGP\Int-VGA" is actually "PCI\AGP". It is mislabeled.

I can confirm the same behaviour in both 008 and 00C revisions.

I hate whoever coded this bios. Sighs.

Shiftyy wrote on 2026-03-10, 14:47:

I hate whoever coded this bios. Sighs.

The BIOS is indeed really flawed, with many poorly-written or even outright nonfunctional options.

If you want to dig deeper, you can refer to BIOS Disassembly Ninjutsu Uncovered. Its AMI example used Soltek SL865PE which should be very similar to this board in terms of chipset as well as the BIOS codebase (AMIBIOS8).

LSS10999 wrote on 2026-03-11, 07:08:

Shiftyy wrote on 2026-03-10, 14:47:

I hate whoever coded this bios. Sighs.

The BIOS is indeed really flawed, with many poorly-written or even outright nonfunctional options.

If you want to dig deeper, you can refer to BIOS Disassembly Ninjutsu Uncovered. Its AMI example used Soltek SL865PE which should be very similar to this board in terms of chipset as well as the BIOS codebase (AMIBIOS8).

Thanks for the link. I'll read through it when I have time. I have some experience with decompilation/dissasembly, but none when it comes to BIOS mods. I am used to IDA converting instructions into obfuscated C code and then working on it, which I wasn't able to do with this.

Hey, I found this topic because I am very interested in "modern" motherboards with ISA slots. I have bought a couple of them in the last few months with i845 chipset but still have not installed them into complete systems. I am not an expert in hacking bios, etc., so I would have to rely on public sources to have it woking in a multi-OS environment for gaming purposes (from DOS to XP/Seven), but still I am very intrigued. Would it make sense to look for this model, too? Which version would be the best choice?
Thanks!

Currently dealing with another issue with this BIOS.

It does not initialize the floppy drive correctly on boot.

If I run the "TESTFDC.COM" utility included with ImageDisk 1.20, it returns "BIOS reports drive A: as Not installed."

However, if I change the boot order and boot into a floppy image first, it works.

I have no idea how to fix this, other than set the BIOS to boot from floppy by default and make a disk that will somehow execute my MBR.

So frustrating.

Any ideas?

Shiftyy wrote on 2026-03-14, 13:11:

Currently dealing with another issue with this BIOS. […]
Show full quote

Currently dealing with another issue with this BIOS.

It does not initialize the floppy drive correctly on boot.

If I run the "TESTFDC.COM" utility included with ImageDisk 1.20, it returns "BIOS reports drive A: as Not installed."

However, if I change the boot order and boot into a floppy image first, it works.

I have no idea how to fix this, other than set the BIOS to boot from floppy by default and make a disk that will somehow execute my MBR.

So frustrating.

Any ideas?

Which BIOS version? TBH I haven't seriously tested floppy drive on my system, either, as currently I don't have anything that absolutely required a floppy disk at the moment.

I think if FDD is enabled in BIOS but not initialized correctly, the BIOS would give out a warning. At least on older boards with AWARD BIOSes this was the case.

I'm not using actual floppy drive on my system. however. I'm using a Gotek modded with HxC firmware.

LSS10999 wrote on 2026-03-14, 14:59:

Which BIOS version? TBH I haven't seriously tested floppy drive on my system, either, as currently I don't have anything that ab […]
Show full quote

Shiftyy wrote on 2026-03-14, 13:11:

Currently dealing with another issue with this BIOS. […]
Show full quote

Currently dealing with another issue with this BIOS.

It does not initialize the floppy drive correctly on boot.

If I run the "TESTFDC.COM" utility included with ImageDisk 1.20, it returns "BIOS reports drive A: as Not installed."

However, if I change the boot order and boot into a floppy image first, it works.

I have no idea how to fix this, other than set the BIOS to boot from floppy by default and make a disk that will somehow execute my MBR.

So frustrating.

Any ideas?

Which BIOS version? TBH I haven't seriously tested floppy drive on my system, either, as currently I don't have anything that absolutely required a floppy disk at the moment.

I think if FDD is enabled in BIOS but not initialized correctly, the BIOS would give out a warning. At least on older boards with AWARD BIOSes this was the case.

I'm not using actual floppy drive on my system. however. I'm using a Gotek modded with HxC firmware.

008. I am using a Gotek with FlashFloppy. It's not a Gotek issue, I plugged in a real floppy drive and got the same problem.

Normally when the mobo is initializing a floppy drive during boot, you should see a green light indicating it is "seeking" whilst at the post screen, I'm fairly sure? Well I'm not seeing any seek activity on this mobo during post.

I ended up making a DOS 6.22 boot image for the Gotek, made it the default image when turning on the PC, and added grub.exe into the autoexec.bat file. This will launch Grub4DOS, which I was already using anyway. The entire process adds around 15-30 seconds to the boot time, but it fixes the issue.

You can also add the pcicmd command into the autoexec.bat and have working DMA for all operating systems.

I've discovered that ACPI is using IRQ2/9 and there's no way to disable it in the BIOS.

Does anyone know of a workaround for this issue?

Shiftyy wrote on 2026-04-30, 06:39:

I've discovered that ACPI is using IRQ2/9 and there's no way to disable it in the BIOS.

Does anyone know of a workaround for this issue?

I think you can put the ACPI SCI IRQ (which by default is wired to IRQ9) into APIC territory by enabling the "APIC ACPI SCI IRQ" option.

However, doing so will cause breakage in Win9x if ACPI is enabled there.

Though I think stuffs that rely on IRQ2/9 (such as intelligent MPU401) will work fine in ACPI-unaware environments since ACPI SCI IRQ will not be involved.

LSS10999 wrote on 2026-04-30, 10:42:

I think you can put the ACPI SCI IRQ (which by default is wired to IRQ9) into APIC territory by enabling the "APIC ACPI SCI IRQ" […]
Show full quote

Shiftyy wrote on 2026-04-30, 06:39:

I've discovered that ACPI is using IRQ2/9 and there's no way to disable it in the BIOS.

Does anyone know of a workaround for this issue?

I think you can put the ACPI SCI IRQ (which by default is wired to IRQ9) into APIC territory by enabling the "APIC ACPI SCI IRQ" option.

However, doing so will cause breakage in Win9x if ACPI is enabled there.

Though I think stuffs that rely on IRQ2/9 (such as intelligent MPU401) will work fine in ACPI-unaware environments since ACPI SCI IRQ will not be involved.

Huh.

So I moved the jumpers on my PCMIDI back to IRQ 2/9 to test out that setting.

Well, now it works perfectly, without needing to change anything in the BIOS.

I'm confused. I must have had a connection problem somewhere.

Thanks for the advice.

I have performed the motherboard BSEL mod and upgraded my CPU to the Extreme x6800 with success.

The x6800 booted into 16x mode at 200 FSB 3.2Ghz first try. I didn't have the 60x multiplier bug mentioned in the first post. I experienced zero issues. Shutdown is now working properly. Memory is running at 400mhz dual channel. The CPU no longer runs at 100mhz when I turn on the PC within 5 minutes after shutting it down. No problems at all.

If you can solder, I highly recommend going this route instead of the E5800. If you can't solder, you should buy yourself a soldering iron, practice a bit, and give it a go. It's probably one of the easiest solder jobs you can do. You only need kynar wire, rosin core solder wire and a soldering iron.

Here are some 3dmark 99 and 3dmark 2001 SE benchmarks comparing the E5800 @ 3.2ghz against the x6800 @ 3.2ghz. Later 3dmark benchmarks also show an improvement for the x6800 but not as drastic, because the GPU becomes the bottleneck.

E5800:

x6800:

You can see the extra cache on the x6800 does make a big difference.

I have also edited the stock v008 bios so that we can now disable C1E support and Intel SpeedStep. Turning these off will stop the CPU from throttling to low voltages and multipliers when under no load. I have attached the edited bios to this post for you to download. The toggle for C1E support can be found under Advanced > CPU Configuration (It was previously hidden), however the toggle for Intel SpeedStep is still hidden. Despite this, we can disable SpeedStep by going to the the bios "Exit" page and load the optimal settings. If you want to enable SpeedStep, you should instead load the failsafe settings.

I also found a program that allows you to adjust the memory timings within Windows XP. This is useful because the settings in the bios for adjusting memory timings seem to be broken or faulty. For example, if I try setting the CAS latency to 2, the board will refuse to post, even though my memory can handle it.

The program is called Tweaker865 and I have also attached it to this post.

You can use it to extract a bit more performance if you are using fast memory. My modules are rated for 2-3-2-6 and these are the settings I use:

These settings reduced my memory latency by a noticeable amount.

You can save a config file and use it when launching the program by using the following command:
tweak865.exe -file=config.txt
Which is great if you want to automatically launch it while booting Windows.

Shiftyy wrote on 2026-06-06, 16:55:

The x6800 booted into 16x mode at 200 FSB (800) 3.2Ghz first try. I didn't have the 60x multiplier bug mentioned in the first post. I experienced zero issues. Shutdown is now working properly. Memory is running at 400mhz dual channel. The CPU no longer runs at 100mhz when I turn on the PC within 5 minutes after shutting it down. No problems at all.

The 60x multiplier bug occurs if you have reset the CMOS and there were no previous settings stored for CPU ratio. In that case the board will try to boot at 60x, fails, then boot with a warning and low ratio setting allowing you to change in BIOS.

For a normal (locked) processor, that value is automatically set to their default multiplier and will be saved along with other settings.

If you previously used a E5800 then switched to X6800 then the BIOS will use the 16x from E5800 for the new processor, but will allow you to change the value since X6800 is unlocked.