Maybe BIOS settings (boot order)?

Welcome to the forum btw!😃

Hi,

well the only setting in BIOS is to disable the SCSI BOOT ROM.
But if I do that, the board won't boot from my main SCSI drive, and it also won't boot from IDE. It is just sitting there doing nothing 🙁

The 486SP3G manual says:

So booting off an SCSI disk with an IDE drive installed does not seem to be possible.

One possible work-around would be to put a boot loader on a cheap, low-capacity memory card in an IDE to SD/CompactFlash adapter, which then properly redirects the boot sequence to the SCSI hard drive. Something like GAG ought to work.

http://gag.sourceforge.net/

Hi,
@pyrogx thank you for finding the crucial piece of information I overlooked in the manual!
Well then I will try that tomorrow afternoon and get back with the results.
What was confusing to me though is, when I disabled the NCR SCSI BIOS in the BIOS, it didn't boot at all.
It just got stuck before enumerating the SCSI devices, not even ctrl+alt+del would work, had to reset the whole machine every time.

Also I would have to sort out the dead battery in the dallas clock chip.
There are some pins and silkscreen next to it which look like a omitted battery socket.
Does anyone have experience which such situations?
I think I still have a cr2032 socket kicking around. I'll pop it in and see what happens.
It is already pretty annoying to remember all the bios settings and set them every time.

*edit*
Now I tried the following:
Booting from IDE Master drive -> System freezes after counting RAM
Booting from IDE with a drive in block mode -> System freezes after counting RAM
Booting from IDE with NCR Firmware in BIOS disabled -> System freezes after counting RAM
Booting from IDE with NCR Firmware in BIOS disabled and SCSI Disable Jumper set -> System freezes after counting RAM
Booting from IDE with NCR Firmware in BIOS disabled and SCSI Disable Jumper set and SCSI completely unplugged -> System freezes after counting RAM

Damn, what an odyssey 😁
@pyrogx
could you kindly point me to the manual you found this in?
I downloaded the manual from the asus website (surprisingly still online)
I could only found a part where it stated that IDE, SCSI and FDD controllers can be used at the same time

*edit2*
ok more info:
I removed all ISA and PCI cards except the videocard, and also tried to switch parallelport ECP jumpers between DMA1 and DMA3
Also Swapped IDE cables
Also used different IDE drives known to work and boot in other systems (fun thing on one of them is a copy of xenix installed 😀
AND
I noticed a specific symptom
When the machine freezes, either the LED on the A or B floppy drive stays on, as well as the HDD LED (for IDE)
no reaction on the keyboard whatsoever
not even ctrl+alt+del...
was that still a hardware interrupt on 486 systems?
if so, why would the cpu lock up then?

I think I found the full manual for that board somewhere here on VOGONS. The manual on the Asus website is incomplete. This Re: Build 486's And They Will Come! Suggestions please! should be the correct one.
I didn't have any issues with the onboard IDE controller on my 486SP3G, but I also didn't have to replace the clock chip yet (still ticking after >25 years!).
So, the system only boots from SCSI, right? And you have to disable the IDE controller completely? Does it boot from IDE if you use an external (ISA) IDE/FDD controller?

audiocrush wrote on 2022-01-05, 08:29:

Also I would have to sort out the dead battery in the dallas clock chip. There are some pins and silkscreen next to it which loo […]
Show full quote

Also I would have to sort out the dead battery in the dallas clock chip.
There are some pins and silkscreen next to it which look like a omitted battery socket.
Does anyone have experience which such situations?
I think I still have a cr2032 socket kicking around. I'll pop it in and see what happens.
It is already pretty annoying to remember all the bios settings and set them every time.

You can just populate the battery socket and put a battery into it. A RTC chip that is designed to use an external battery will immediately pick it up and work perfectly. The Dallas chips with integrated battery don't have the pins that connect to the battery, though. So the flat Dallas chip you already have can't make use of the battery, unless you add the missing pins (and preferably disconnect the internal battery at that time).

A word of warning: The BIOS revision on the PCI/I-486SP3G works mostly with a depleted BIOS battery, but it stubbornly refuses to tell DOS about the floppy drive type as stored in CMOS, because "the battery is flat", even if the settings are valid. You can sense the battery status of a Dallas chip (and the BIOS does exactly that) with integrated battery even if it gets +5V from the mainboard. This will cause strange issues with floppy drive access in DOS, see Re: Strange Floppy Disk behaviour 486 and the subsequent post in that thread.

@pyrogx
oh cool.
Yes that is correct, but it is sufficient just not to connect any IDE devices and it works perfectly fine.
I did not try a ISA IDE controller yet since I fear poor performance, and I only have VLB IDE controllers here so they wouldn't work anyways on that board.
But if I get hold of one I'll try.

@mkarcher
Thanks for the tip.
I have had that exact problem that I needed to use driveparm.sys to tell dos it is a 1.44MB A drive and a 1.2MB B drive, otherwise it would use both as 360k only and refuse to read further sectors.
Soooo annoying to figure this out it took me 2 weeks.

Ok well then I guess it's time to work on that dallas chip
Wish me luck to not destroy that nice motherboard.

The IDE chip on that board is a NatSemi PC87332, which is ISA only anyway. So you won't loose any performance if you use a separate ISA IDE controller.

pyrogx wrote on 2022-01-06, 18:26:

The IDE chip on that board is a NatSemi PC87332, which is ISA only anyway. So you won't loose any performance if you use a separate ISA IDE controller.

Using a separate ISA IDE controller wouldn't change anything. The BIOS doesn't see any difference between the on-board IDE chip connected to the ISA bus, or a different IDE chip connected to the same ISA bus.

On the other hand, not setting up IDE drives in the on-board BIOS, but providing a different IDE BIOS implementation (like the XT IDE BIOS) on an extension ROM can be a solution to have IDE disks behind SCSI disks. If this approach works (I have no idea whether the on-board SCSI BIOS is initialized before or after ISA extension ROMs), it still doesn't matter whether the on-board IDE controller chip or a dedicated ISA card is used.

Hi,

ok I did it... after 2 hours of scraping I depotted the dallas chip, soldered a 2xAA holder on to it and bingo, bios saves settings now.
Unfortunately that didn't solve the IDE problem, still all the same symptoms.
I might later try to reflash the bios, just in case a bit flipped in there or something. Luckily the board supports flashing without taking the eeprom out.

@mkarcher
oh thats useful information.
Well the NCR SCSI chip is PCI
My dad found a 16-bit ISA IDE controller, I'll also try that later on
Unfortunately I don't have any PCI IDE controllers (yet)

I mean I have really big SCSI hard drives, so there wouldn't be any space issues...
But I'd really like to use my IDE HDD frames so that I can just exchange the boot disk depending on the OS I want to run, like DOS, Win 95, os2 or whatever I might want to try... just for convenience

By the way.. did anyone try to run the board on 8x64k Cache chips yet?
Because in the manual it says for 512k use 4x128k chips.. but I only have 64k chips 😒
Might I fry them if I try it anyways?

*edit*
I upgraded the BIOS from 0302 to 0306 and now it works like a charm!
Thank you a lot for your support!

By the way just if someone else has this problem:
I tried several bios files from different sources and also two different award bios flasher tools and they all didn't work
It only worked with the tools that the User Skyscraper in this thread thankfully provided on the forum: ASUS PCI-486SPG3 Questions

audiocrush wrote on 2022-01-07, 13:06:

Unfortunately I don't have any PCI IDE controllers (yet)

Don't expect PCI IDE controllers to be easy on early 486 PCI boards. I didn't try them on the PCI/I-486SP3G yet, but the experience wih UMC8881-based and SiS496-based boards is not good. I got a Highpoint-based controller work on the Shuttle HOT433 (UMC chipset), but didn't have any success with Promise or Silicon Image based controllers. Possibly PCI compatibility is better with Intel chipsets, though.

audiocrush wrote on 2022-01-07, 13:06:

I mean I have really big SCSI hard drives, so there wouldn't be any space issues...

You might want to look into my patch at https://github.com/karcherm/sdms_8g in case your disks are bigger than 8 gigabytes.

audiocrush wrote on 2022-01-07, 13:06:

By the way.. did anyone try to run the board on 8x64k Cache chips yet?
Because in the manual it says for 512k use 4x128k chips.. but I only have 64k chips 😒
Might I fry them if I try it anyways?

No danger of frying cache chips unless you insert them 180° turned around.. But the copy of the PCI/I-486SP3G I have at hand has only one bank equipped with 32-pin sockets, so you not physically insert 8 64KB chips. 64KB chips will not work correctly if you just leave the 4 excess pins hanging over the edge of the socket - they have a bigger case for a reason.

Try 3ware 2 port PATA PCI card. Perfect for DOS, but unknown using 95 and 98SE.

Cheers,

Hi @pentiumspeed
Thanks for your suggestion, but a bios upgrade already resolved all my problems.
It is just counting up the ram during post 4-5 times now, but no big deal for me.

@mkarcher
about the 64k chips
I didn't think of that... good point with the sockets 😁
I think it wouldn't make such a big difference anyways would it?

audiocrush wrote on 2022-01-10, 10:44:

It is just counting up the ram during post 4-5 times now, but no big deal for me.

(512 KB cache)
I think it wouldn't make such a big difference anyways would it?

If the BIOS counts the memory multiple times, it seems like "Advanced BIOS setup" / "Quick POST" is disabled.

The extra cache won't make a big difference in general, but there might be very specific tasks that don't fit 256k, but fit 512k and are dominated by memory performance, where a difference could be noticeable.

Oh thanks,
forgot about that Quick POST 😁