Yes.

Software RAID that is. ZFS requires a simple HBA, I'd highly recomend that.

Ok. Is there any way to do it with Windows 9X or 2000 Server? I know about how you have to set the IRQ for each HBA differently. I believe I also need to change the Interupt, correct me if I am wrong.

I'm not 100% sure, but I think that Windows 2000 can define RAIDs using the Disk Manager. As far as I remember, you could RAID volumes but only if the disks were defined as Dynamic Disk.

Thata capability was not in Windows 9x, so you would need some RAID software for that systems.

I wanted to try multiple ISA I/O controllers on a motherboard that already has an integrated controller. If I get the IRQ settings changed on these cards, would it be possible? I am building an IDE fileserver, and the board I am using has a couple of ISA slots on it.

I didn't even know stuff like this was possible using ISA IDE cards! 😲
Awesome! 😁

Yeah, well this was supposed to be on that Supermicro board which I can't find the drivers for. It has an ISA slot.

HardwareExtreme wrote:

Yeah, well this was supposed to be on that Supermicro board which I can't find the drivers for. It has an ISA slot.

Can you get into Windows and get an Internet connection? On Win2K or better you should be able to run one of those stupid driver finder tools like Driver Booster if it's really tough. Or let us take a crack at identifying the missing device drivers 😀

Stiletto wrote:

HardwareExtreme wrote:

Yeah, well this was supposed to be on that Supermicro board which I can't find the drivers for. It has an ISA slot.

Can you get into Windows and get an Internet connection? On Win2K or better you should be able to run one of those stupid driver finder tools like Driver Booster if it's really tough. Or let us take a crack at identifying the missing device drivers 😀

It's this one Supermicro 370DLE Drivers

The strange thing is that drivers for a similarly chipsetted motherboard made by ASUS were also "missing"

hum... you do know that ISA is very slow, right? Yeah you could use a software RAID, but it will be slower than a single decent IDE HDD on a lousy PCChips Socket 7 motherboard.
If you really want to use RAID and cannot use the onboard controller, disable it and get something PCI/PCI-X. Avoid ISA, as it is useless for that on such computer.

On my Supermicro Board, it has an integrated IDE controller. However, I will have like 6 drives hooked up at a time. And, to getting to the internet in Windows 2000, not happening. Windows 2000 won't even boot at all.

I'm not 100% sure about the IRQ, but depending on operating system support, devices should be able to share an IRQ without any problems (when the IRQ is triggered, the drivers have to know it is shared and remember to check which device is responsible for the interrupt.)

The problem is that IDE also has I/O port addresses. These cannot be shared at all, so if you put two IDE controllers in with the same I/O address then neither one will work. This is exactly the same as serial ports - they can share IRQs, but it's the I/O address that defines which COM port it is.

For IDE, the primary controller uses I/O addresses 0x1F0 to 0x1F7, while the secondary IDE controller uses 0x170 to 0x177.

There seems to be some sort of consensus that a third IDE controller can sit at 0x1E8 - 0x1EF, and a fourth at 0x168 - 0x16F, but this is a much later development so there is limited (if any) BIOS support. No BIOS support means you won't be able to boot off any drive on those controllers, unless of course the IDE controller card has a BIOS that is aware of the third and fourth controller addresses. I believe XTIDE is, or at least can have the I/O addresses it responds to changed.

The main difficulty is of course finding an IDE controller that can have its I/O addresses changed. Most of the ones I have only have a single IDE connector and it can be either set to primary or disabled completely - you can't even set it to be a secondary controller, let alone tertiary or quaternary. Possibly controllers designed for RAID setups might have jumpers to change their addresses, but that's what you will be looking for if you want more than two IDE controllers in the same system.

The other possibility is that your motherboard could let you move the onboard controllers to positions 3 and 4, so add-in cards can provide controllers 1 and 2, but I think this is rare. You might think a server board would support this, but usually people who wanted to do this would switch to SCSI where you can put 7 or 15 devices on a single controller depending on SCSI version, so even server boards probably only support two IDE controllers.

As for using ISA cards, remember that a 16-bit ISA bus running at 8MHz can theoretically transfer a maximum of 16MB/sec across *all* cards. Usually it will be significantly less than this. A single cheap SATA SSD from China with an IDE converter, with an onboard IDE controller capable of UDMA2 will max out at 33MB/sec - double the speed of anything connected to ISA, with no RAID involved, no noise, and a fraction of the power draw. So fair enough if you are doing it just to see whether it can be done, but if you want any sort of decent performance for what the machine is capable of, an ISA RAID array is definitely not the way to go!

Malvineous wrote:

I'm not 100% sure about the IRQ, but depending on operating system support, devices should be able to share an IRQ without any p […]
Show full quote

I'm not 100% sure about the IRQ, but depending on operating system support, devices should be able to share an IRQ without any problems (when the IRQ is triggered, the drivers have to know it is shared and remember to check which device is responsible for the interrupt.)

The problem is that IDE also has I/O port addresses. These cannot be shared at all, so if you put two IDE controllers in with the same I/O address then neither one will work. This is exactly the same as serial ports - they can share IRQs, but it's the I/O address that defines which COM port it is.

For IDE, the primary controller uses I/O addresses 0x1F0 to 0x1F7, while the secondary IDE controller uses 0x170 to 0x177.

There seems to be some sort of consensus that a third IDE controller can sit at 0x1E8 - 0x1EF, and a fourth at 0x168 - 0x16F, but this is a much later development so there is limited (if any) BIOS support. No BIOS support means you won't be able to boot off any drive on those controllers, unless of course the IDE controller card has a BIOS that is aware of the third and fourth controller addresses. I believe XTIDE is, or at least can have the I/O addresses it responds to changed.

The main difficulty is of course finding an IDE controller that can have its I/O addresses changed. Most of the ones I have only have a single IDE connector and it can be either set to primary or disabled completely - you can't even set it to be a secondary controller, let alone tertiary or quaternary. Possibly controllers designed for RAID setups might have jumpers to change their addresses, but that's what you will be looking for if you want more than two IDE controllers in the same system.

The other possibility is that your motherboard could let you move the onboard controllers to positions 3 and 4, so add-in cards can provide controllers 1 and 2, but I think this is rare. You might think a server board would support this, but usually people who wanted to do this would switch to SCSI where you can put 7 or 15 devices on a single controller depending on SCSI version, so even server boards probably only support two IDE controllers.

As for using ISA cards, remember that a 16-bit ISA bus running at 8MHz can theoretically transfer a maximum of 16MB/sec across *all* cards. Usually it will be significantly less than this. A single cheap SATA SSD from China with an IDE converter, with an onboard IDE controller capable of UDMA2 will max out at 33MB/sec - double the speed of anything connected to ISA, with no RAID involved, no noise, and a fraction of the power draw. So fair enough if you are doing it just to see whether it can be done, but if you want any sort of decent performance for what the machine is capable of, an ISA RAID array is definitely not the way to go!

Would it in theory be possible to hack the BIOS's of these IDE boards so they can be used without "blocking" eachother?

The IDE boards generally don't have a BIOS, it's the system BIOS that you'd need to change. You can also use the XTIDE BIOS even on a system with IDE support in the BIOS. I used XTIDE BIOS on a 486 because the 486 BIOS used a nonstandard CHS/LBA translation method so I couldn't move a disk (in this case an SD card) between machines, so it's possible to work around limitations in the system BIOS.

But the "blocking" problem isn't software, it's hardware. When the system goes to talk to a device on port 0x1F0, it gets two responses at the same time (one from each card) so they arrive combined and garbled. One card might respond with the number 6, another card responds with the number 20, but these values get mixed as they travel over the ISA bus at the same time, so the CPU sees only one response with the value 22, which is not the value either card sent.

It's the same as asking a question and having two people answer at the same time. You can't clearly make out what either of them are saying.

Bog standard IDE controllers usually don't require special drivers. The onboard IDE should work fine on Windows 2000. I would double check the BIOS settings and make sure the controllers are enabled and detecting connected drives.

Another thing to try is the Alter UniATA drivers should the problem be with the built in IDE driver.

If I remember right ISA DMA transfers are a no-go if you are running Win95/98 with more than 16mb of ram. Well, yes, it can be done.. but not without an additional performance penalty. 🤣

AFAIK proper ISA DMA is even slower than ISA PIO. It is mostly used for floppy drives, sound cards and maybe printer ports (ECP/EPP).
Bus master DMA can be done, but to me that sounds more like a hack.
See http://www.vintage-computer.com/vcforum/showt … d-DMA-transfers for relevant discussion.