Scraphoarder wrote on 2024-05-07, 16:59:

Thanks and nice to see someone else also made progress on making theese boards usable. Maybe i will try to dig up one of mine that also have two unresponsive ram slots and flash your BIOS.

I don't think the issue with RAM slots could be fixed from the BIOS side. The RAM slot problem was more likely hardware. My BIOS files only added additional microcodes needed for proper support of 45nm CPUs as well as updated all of them to the latest available version.

The 008 version of BIOS apparently had a few additional options in the ACPI part that were not present in 009 version, though it was not too important to me. It's advised to use the same BIOS version as the one your board has, as flashing the other version requires a few extra options to "force" it.

I did not actually fix the ISA DMA problem on the BIOS side. I simply found out the cause of it and that can be easily corrected using programs that could manipulate PCI registers.

Its inability to fully utilize quad-core CPU was likely due to its ACPI tables declaring only two CPU entries compared to four in the ACPI tables from ASRock 865 boards. I just don't know which is the correct way to edit and rebuild ACPI tables for old AMI BIOSes...

It seems the 1.0s1.2 (865G with real AGP slot) version is becoming harder to find around my place... while 865GV, which 1.0s1.3 uses, is rather unfit for use with a discrete PCI video card, due to its onboard video cannot be disabled. The respective register controlling whether or not to enable the onboard video is read-only in 865GV, while read-write in 865G. As such, if using a discrete PCI video card, both the onboard and discrete video would be active in the OS which may lead to problems if you're not actively using the onboard video.

Was a little unclear what i wanted to do, but i was thinking to try tinkering with the board with bad ram slots if something went poop with flashing modded BIOS. I have two other thats ok. All my three boards are v1.0S1.2, but since i wasnt eager and able to fork out $275 for a russian BIOSmod with some software i put them all in storage waiting for better times.

I purchased one of these boards a month or so ago and am working on putting together a system with it.

The BIOS battery had already been replaced with a CR2032 holder.

All memory slots seem to work fine.

However, even with the posted BIOS with the added microcode, I am unable to get any QX CPU to work with it.
I have tried:
QX6700
QX6850
QX9650

All I get is a a single repeating beep over and over again.

I do have an X6800 on the way.

Is the FSB mod required for the 1066 FSB CPUs to work at all? Previous posts seems to make it look like it is not required.

cyclone3d wrote on 2024-06-10, 21:32:

I purchased one of these boards a month or so ago and am working on putting together a system with it. […]
Show full quote

I purchased one of these boards a month or so ago and am working on putting together a system with it.

The BIOS battery had already been replaced with a CR2032 holder.

All memory slots seem to work fine.

However, even with the posted BIOS with the added microcode, I am unable to get any QX CPU to work with it.
I have tried:
QX6700
QX6850
QX9650

All I get is a a single repeating beep over and over again.

I do have an X6800 on the way.

Is the FSB mod required for the 1066 FSB CPUs to work at all? Previous posts seems to make it look like it is not required.

I recommend against using Extreme edition CPUs including X6800. The BIOS' CPU ratio option is bugged. With an Extreme edition CPU, it will first try booting with a maximum ratio (something like 60x) and fail. You need to somehow force the BIOS to believe there's an overclock failure so it'll start in a failsafe mode at lowest possible ratio, then you can specify the ratio value you want.

Ideally you'd want 1066 FSB CPUs to work at 800 FSB. Getting 1333 FSB CPUs to work at 800 FSB may not be as easy as with 1066 FSB ones. The FSB mod in question was applied to a transistor related to a BSEL bit, not directly on the CPU, according to the Modlabs post I referred to when I did mine.

However, it seems BIOS still shows 1066 for an E7600 I tried. I wasn't able to conduct further tests to prove whether it really worked, as I kind of irreversibly borked the test board when I tried modifying its ACPI table to add a 3rd and 4th CPU entry (in hope this will enable proper quad-core CPU support). A key part is that with the FSB mod applied, I did not have too much trouble getting E7600 booted, while previously I had to set RAM down to 266MHz (lowest) in order to let it boot properly.

Again, the board's ACPI table had only two CPU entries so only dual-core CPUs can work as expected. Similar 865 motherboards that do have proper BIOS support for Kentsfield (65nm C2Q) have 4 CPU entries.

With correct microcodes added 45nm Pentium Dual-Core E5000 family (FSB 800) will work best. 65nm CPUs would occasionally drop bus speed to 100MHz, while I don't see this happening often on 45nm ones. The only major caveat with 45nm CPUs is that ACPI shutdown no longer works. The system would simply freeze once it finishes shutdown, and you need to manually power it off by holding the power button for at least 5s.

Ok, that clears things up a bit. I bet I can get quads working. If I brick the BIOS somehow I can pull the chip and reflash it.

I also want to try to recreate the Russian fixes so ISA DMA works without having to have a program to set it.

cyclone3d wrote on 2024-06-11, 03:22:

Ok, that clears things up a bit. I bet I can get quads working. If I brick the BIOS somehow I can pull the chip and reflash it.

I also want to try to recreate the Russian fixes so ISA DMA works without having to have a program to set it.

Just be careful. I think you should backup a complete copy of your BIOS ROM (using an external programmer) before you try to make changes.

AFUDOS won't save everything. It's possible some stuffs in NVRAM may be needed in case you bricked and requires recovery.

I never did a full ROM backup, and now the board doesn't even boot anymore after I replaced the BIOS chip with another one flashed a known working AFUDOS backup -- it just gives out a long beep followed by 6 short ones.

As for ACPI table editing -- IASL probably isn't a good idea. I made my modifications with IASL and while there were no errors when recompiling and re-decompiled results did not look too out of place for me, it borked the BIOS when I tried it (the BIOS can boot, but will hang at a POST code supposedly about building ACPI tables). For old AMI BIOSes like these you probably need to look for Microsoft's ACPI tools to do the job.

On the other hand, if you have knowledge with BIOS editing perhaps you can try peeking into the behaviors of the DMA reserving toggles in the PCI/PnP section, and see if you can somehow alter their behavior, like setting "Reserved" to make the particular DMA channel route to PC/PCI instead of LPC (according to the datasheet).

I had looked at IASL and came to the conclusion that it would not be the right tool to use. From what was coming up, it looks like it is specifically meant for hackintoshes and the instructions say that it automatically adds the changes when using the tools to get the AHCI tables which is not what we want.

Edit: were you using the iASL from Intel or one of the ones posted around different hackintosh sites?

cyclone3d wrote on 2024-06-11, 14:38:

I had looked at IASL and came to the conclusion that it would not be the right tool to use. From what was coming up, it looks like it is specifically meant for hackintoshes and the instructions say that it automatically adds the changes when using the tools to get the AHCI tables which is not what we want.

Edit: were you using the iASL from Intel or one of the ones posted around different hackintosh sites?

I did my edit from Arch Linux. IASL is part of the "acpica" package in "extra".

Just I don't know which is the correct way to replace the ACPI table in old AMI BIOSes...

The x6800 CPUs I ordered arrived.

A simple clear CMOS via jumper got it to boot.

Yes, thee CPU ratio initially came up at 60.

Weirdness observed:
1. When changing the CPU voltage from Default to a manual setting, the board automatically changes the CPU voltage based on the ratio set with it limiting the lowest voltage available based on the set ratio.
2. Even at default voltage, this CPU is POSTing at up to a 31x multiplier. Anything past that and the board gives an error and sets the multiplier to 6. Setting CPU voltage does not change this.
3. RAM speed must be set at 266 or Auto for it to POST with the X6800.

Something is definitely screwy with this BIOS. Next step is to set up an OS, probably Windows 98 so I can see what the FSB is actually running at.

Want to get some more things figured out before I try to get qauds or anything else possibly fixed with the BIOS.

Got this board.

it has a bios ver 00C , so i'll post it here.
btw why there is no page for this board on theretroweb?

does anyone have a bios from working rev 02 board?

sadly, my board refuses to accept agp 8x cards. agp 4x cards working, detecting, but has some resources conflicts in windows.

dm- wrote on 2024-08-08, 03:58:

does anyone have a bios from working rev 02 board?

By "rev 02" are you referring to 1.0s1.2? (the one with 865G and working AGP)

I once dumped a 865GV008 BIOS image (with microcode manually updated via MMTool so 45nm C2Ds can be used) from a now-bricked 1.0s1.2 board.

However, these BIOSes may be very different and require some additional AFUDOS parameters in order to be flashed.

Will be looking at your 865GV00C BIOS sample to compare with the version on my own board. Does DMA on ISA slot work out-of-box with your BIOS version? In my board's case the DMA was all set to LPC by default and has to be fiddled manually after boot.

dm- wrote on 2024-08-08, 03:58:

sadly, my board refuses to accept agp 8x cards. agp 4x cards working, detecting, but has some resources conflicts in windows.

For Win9x such resource conflicts may be due to bugs with ACPI. Not sure if lowering AGP Aperture size would help but you may consider disabling ACPI altogether.

On the other hand I'm afraid the board is plagued by all kinds of QC issues, so there may be other reasons behind its non-acceptance of AGP8x cards.

yes it is a 1.0s1.2
the north bridge chip was cracked and not working so i replaced with brand new 865G chip using pro soldering station.
this 0C bios has the same ISA DMA bug, so no surprises

i tried both you bios'es and it is the same, agp 8x does not work, stuck on code D3

i using WinXP

Recently I'm able to acquire another 1.0s1.2 board, which appears to be fully working (all RAM slots, AGP, onboard audio and such) as far as I've tested.

(EDIT: Sadly it also had flaws. Its PS/2 mouse port was dead, that OSes/drivers simply don't detect it. I'm afraid I'll have to rely on a PS/2-to-serial converter to connect it via serial, if it's working.)

During testing of the newly acquired board I found out the following:
- ACPI shutdown issue with 45nm CPUs appears to be specific to CPUs with 1067A* CPUID. When using a CPU with 10676 CPUID (e.g. E5200 SLAY7), the system can shut down fine, but it may drop down to 100MHz bus speed during runtime similar to E4700 (Allendale).
- FSB1066 CPUs (65nm or 45nm) always run at 100MHz bus speed and I've never seen such CPUs run at its native speed (266MHz) on this board. Even though I have to set DRAM frequency to 266MHz, MemTest86+ still reports DDR400 (200MHz) during testing.

* Neither 865GV008 nor 865GV009 had initial microcodes for CPUID 1067A. They only have initial microcodes for CPUID 10676.
** The aforementioned ACPI shutdown issue has been fixed in 865GV00C, see next post.

Some additional info per previous posts:
- I gave a look at the 865GV00C BIOS with AMIBCP and MMTool. It's lacking a few options compared to 865GV008 and 865GV009, but it appears to have initial microcodes for 1067A CPUID.
- As for the report on X6800. Considering FSB1066 CPUs always run at 100MHz, setting it to 31x would be 3100MHz, which isn't a high value if other components on the board are okay with it.

Also, it seems even with 1.0s1.2 (865G), onboard video controller can only be disabled by installing an AGP video card. When the AGP slot is empty, the board behaves the same way as 1.0s1.3 (865GV) -- onboard video remains present on the system but it will not claim VGA resources.

It should be noted that I cannot even modify the related bits via my PCICMD tool. Trying to change the bits have no effect. The documentation says the software needs to do a "clock synchronization" that I could not see a good explanation about it... but I suspect the IGDIS bit in question is in fact being controlled externally. That is, the bit is set to low when AGP slot is empty, and if an AGP card is installed, it will be pulled high (which disables onboard video). In this sense, 8x5GV is simply a variant of 8x5G that had related functions internally hardwired to onboard, without the said toggle for AGP.

For now I'm using a PCIe-PCI reverse adapter to connect a modern half-height video card with DP port to a PCI slot, but for best results, a PCIe-AGP adapter, even just a simple one using it as a 66MHz PCI slot, should be used instead, yet there's no such thing. The best match for such a connection is this method, PCIe video card -> PCIe-PCI reverse adapter -> RECNAS -> AGP slot, which has been tested working by the post author. However, it's not practical for installing inside a case due to their total height.

LATE EDIT: I redid the soldering of the PS/2 port of that 1.0s1.2 board just in case, though it seems the issue with PS/2 was probably not the soldering, but rather the port was a bit loose that I need some extra effort to get the mouse connected somehow so the system can correctly detect it.

On the other hand, I replaced the ALC655 codec chip with a new one on the 1.0s1.3 board whose onboard audio was not working ("missing") regardless of its enabled status in BIOS, and now onboard audio can correctly show up after that... so, thankfully the issues on the boards I recently acquired are fortunately repairable...

A good news about 865GV00C BIOS -- ACPI Shutdown issue with 1067A CPUs has been fixed there, so this BIOS version should be recommended if you're using a 45nm Pentium (Dual-Core) E5x00 CPU.

- System shuts down immediately instead of hanging after pressing POWER button while in DOS, BIOS and some other environments.
- After issuing ACPI shutdown command, system no longer hangs and can be properly shut down.

Tested with E5800 (FSB800) and also E7600 (FSB1066, only runs at FSB400), and both 1.0s1.2 and 1.0s1.3.

Note that you need to do a more complete flashing when updating to it (by adding /B /C, which AFUDOS will do it for you if you press A to accept the recommendation).

PS: You may encounter some issues after flashing it from an older BIOS version. When I flashed my 1.0s1.2 board (865GV008) to the new version, the keyboard would stop responding shortly after entering BIOS via F1, though after pressing F2 to load defaults then reboot the system back to BIOS, it becomes working again.

PPS: I've added some info about how CPU FSB works on this board in the OP as a separate section.

dm- wrote on 2024-08-08, 07:13:

yes it is a 1.0s1.2 the north bridge chip was cracked and not working so i replaced with brand new 865G chip using pro solderin […]
Show full quote

yes it is a 1.0s1.2
the north bridge chip was cracked and not working so i replaced with brand new 865G chip using pro soldering station.
this 0C bios has the same ISA DMA bug, so no surprises

i tried both you bios'es and it is the same, agp 8x does not work, stuck on code D3

i using WinXP

Issue with agp slot was a broken pad with GNT# signal from AGP under the 82865G chipset, so i had to remove chipset and restore the pad.
after re-soldering chipset back to the board, it's back with fully working agp slot.

Hello! I'm interested in:

- Force the disk to IDE Slave. Some PCBs have jumper or footprint for such a toggle that you can just use it. In JM20330's case, if unsure, you can use a multimeter to inspect where you can access pin 33 (MSSEL). If this is not properly configured and the disk operates as Master, it will prevent installed CF card from working.
- "Cheat" 80-wire detection. In order to allow higher UDMA levels the pin 34 (PDIAG) must be grounded. A simple way to do is connect pin 30 and 34 with a wire. This is very important, as if 80-wire detection is not correctly configured, even the CF card will be forced to UDMA2. Additionally, 80-wire detection must be set to "Host" in BIOS to ensure everything works correctly.

Is anyone have photo how it's realized ? I didn't find on my motherboard jumper for master/slave CF and what we must to do with pin 33 ? Allso want to understand from which side is the counting pinouts?

dm- wrote on 2024-12-18, 05:25:

Issue with agp slot was a broken pad with GNT# signal from AGP under the 82865G chipset, so i had to remove chipset and restore the pad.
after re-soldering chipset back to the board, it's back with fully working agp slot.

So that really was a QC issue. On the other hand, from the screenshot it appears you have the modded BIOS. I wonder if that was also based on 865GV00C you previously uploaded.

Besides, I don't know if there were any newer official BIOS version existed for this board, though 865GV00C already brought in some major improvements, like proper ACPI shutdown for 1067A CPUs and the ability to select any hard disk / USB stick without having to change boot order.

At present I'm using a FirePro W4100 with the help of a PCIe-PCI reverse bridge. So far it's working fine with some notes:
- On Linux, the default "radeon" driver doesn't work correctly. It would either glitch, freeze or simply no graphical output. Must add necessary boot parameters to make it use the newer "amdgpu" driver which works perfectly.
- The card doesn't appear to work on XP (Code 10). Don't know if it was just the card, or the reverse bridge setup. The card was never officially supported on XP, though W600, an older, full-height but otherwise identical chipset-wise card was supported. The FirePro V3900 card I previously used, however, worked on WinXP.
- The card does work fine with Win7. On 32-bit, the latest Radeon Pro Enterprise driver for this card (18.Q4) works okay. Some particular driver versions have issues, however. Those newer driver versions added Vulkan support though I haven't tested it at the moment.
- Don't know how Win7's SAT calculates graphical performance. Even when operating behind a PCIe-PCI reverse bridge, the tool reported a whopping 7.2 point for Aero and 3D Graphics, something that I feel very doubtful considering the very constrained bandwidth of PCI bus.

(NOTE: The onboard video controller remains present when not using the AGP slot, so I still have to disable Intel video drivers whenever necessary to avoid it messing with the startup process.)

As for the AGP slot (1.0s1.2), though I did make a reference about RECNAS (AGP-PCI adapter) in previous posts, I just realized something -- modern AGP slots are 1.5V instead of 3.3V. The adapter in question didn't appear to have any level shifting components so I'm afraid that's not going to work here. It would have to be a seriously designed AGP-PCIe adapter that takes all possible AGP voltages into consideration, which will likely require an external power input.

Mr.Hunt wrote on 2024-12-18, 10:11:

Hello! I'm interested in:

- Force the disk to IDE Slave. Some PCBs have jumper or footprint for such a toggle that you can just use it. In JM20330's case, if unsure, you can use a multimeter to inspect where you can access pin 33 (MSSEL). If this is not properly configured and the disk operates as Master, it will prevent installed CF card from working.
- "Cheat" 80-wire detection. In order to allow higher UDMA levels the pin 34 (PDIAG) must be grounded. A simple way to do is connect pin 30 and 34 with a wire. This is very important, as if 80-wire detection is not correctly configured, even the CF card will be forced to UDMA2. Additionally, 80-wire detection must be set to "Host" in BIOS to ensure everything works correctly.

Is anyone have photo how it's realized ? I didn't find on my motherboard jumper for master/slave CF and what we must to do with pin 33 ? Allso want to understand from which side is the counting pinouts?

There's no Master/Slave jumper for the onboard CF card slot. It's always Master. Also, if you're not connecting anything else to the secondary IDE channel (which uses a 44-pin connector instead of 40-pin) you don't have to worry as the CF slot by itself will not be constrained to UDMA2.

Also, the pin 33 I mentioned there was about the IDE/SATA bridge chip (JM20330). How to access that pin depends on your IDE/SATA adapter's PCB design and you'll need to use a multimeter if the PCB doesn't already expose it as a jumper. The pinouts for the bridge chip can be found in the datasheet, which can be found online.

There's no Master/Slave jumper for the onboard CF card slot. It's always Master. Also, if you're not connecting anything else to the secondary IDE channel (which uses a 44-pin connector instead of 40-pin) you don't have to worry as the CF slot by itself will not be constrained to UDMA2.

I already use ide44 hdd for 32Gb right now on this motherboard and I think about migrate loader to CF, so after use ide44 for DOS and Windows 9x and sata for Vista/7.
Buy the way, I can't start to use SSD on IDE44 sized 128Gb.

Mr.Hunt wrote on 2024-12-18, 11:50:

There's no Master/Slave jumper for the onboard CF card slot. It's always Master. Also, if you're not connecting anything else to the secondary IDE channel (which uses a 44-pin connector instead of 40-pin) you don't have to worry as the CF slot by itself will not be constrained to UDMA2.

I already use ide44 hdd for 32Gb right now on this motherboard and I think about migrate loader to CF, so after use ide44 for DOS and Windows 9x and sata for Vista/7.
Buy the way, I can't start to use SSD on IDE44 sized 128Gb.

So you're using an actual IDE (44-pin) SSD, not a SATA one behind a bridge adapter?

Sorry I'm not very familiar with those, as I've no idea how Master/Slave were handled on 2.5'' IDE (44-pin) disks.

I don't think this board's BIOS have any issue with disk size. With good quality bridge adapters even 2TB disks can work.