I like buying a scsi card with floppy. The dma of the scsi (if it has it) can make things faster 😀

Most 16-bit ISA IDE cards are just providing a pathway from the motherboard to the hard drive, so it's not that critical which one you use.

But you get better performance out of a VLB IDE controller.

XT-IDE cards give you good compatibility but someone will correct me I'm sure if I'm wrong when I say that the fact that it's 8-bit ISA might slow things down a little (not the XT-IDE BIOS per se, but the actual card itself).

I think there is no performance uplift from VLB IDE controller in DOS, since it's all 16-bit code (and 8086 compatible). It would require Windows or Linux to take advantage of it. XTIDE (the ROM, not the 8-bit CF carrier card for XTs) will also help when configured properly for 32-bit access but only if the mobo BIOS is decent enough to offer RAM shadowing.

Deunan wrote on 2024-07-04, 10:23:

I think there is no performance uplift from VLB IDE controller in DOS, since it's all 16-bit code (and 8086 compatible). It would require Windows or Linux to take advantage of it. XTIDE (the ROM, not the 8-bit CF carrier card for XTs) will also help when configured properly for 32-bit access but only if the mobo BIOS is decent enough to offer RAM shadowing.

VLB certainly is faster, even in DOS... And VLB is a 32-bit interface. There's a thread that compares speeds between different controllers.

Mostly I will use the controller/io card which I know the jumper settings for. I have a few that I am unable to configure, one I am using because the settings I received it with are close enough. You can plug it in and (if it boots) use MSD to check which resources are active to get started.

Sound cards can provide ide controller for additional devices. I have found better support for atapi devices like CD-ROM using sound card ide vs old style io card. Configuration can sometimes be a challenge.

I also agree that scsi is a nice option if you have the drives. Scsi also can provide a bios and larger drive support. I currently have 5 systems that predate onboard io (286,386,486) 2 use scsi 3 use ide for boot drive. One of the ide systems also has scsi for optical.

mockingbird wrote on 2024-07-04, 10:40:

VLB certainly is faster, even in DOS... And VLB is a 32-bit interface.

It's not about interface width, at least not much unless you can negotiate something faster than PIO 0 with the drive. This requires a ROM BIOS extension on the card, or a device driver loaded (or at least run) during boot. In which case XTIDE will do the same thing, even for 16-bit ISA controllers, and you get most of the speedup without VLB.

Now, on a Pentium system with IDE integrated into the chipset going from 16 to 32 bits does indeed improve things a lot. But that's because you will be wasting way less CPU cycles on PCI latency if you can combine reads and writes. And even here it still requires that the drive is not stuck at PIO 0, because that will be the main bottleneck.

If OP want's to have VLB IDE then sure, why not, but I'm just saying it's not stricly required to get more peformance out of a HDD on 486 VLB system. On the other hand the VLB cards can be very picky and unstable, I'd rather have a VLB SVGA and no other VLB cards, and a stable system, than trying to get two VLB cards to cooperate - which is sometimes next to impossible to achieve.

Deunan wrote on 2024-07-04, 13:39:

It's not about interface width, at least not much unless you can negotiate something faster than PIO 0 with the drive. This requires a ROM BIOS extension on the card, or a device driver loaded (or at least run) during boot. In which case XTIDE will do the same thing, even for 16-bit ISA controllers, and you get most of the speedup without VLB.

In theory, yes XT-IDE provides a "16-bit" option in the configuration utility. Practically speaking, there are a couple of issues:

• Even with 16-bit selected ('16-bit ISA/VLB/PCI' option in XTIDECFG), it's still slow
• This option will not work on an 8-bit XT-IDE controller

Now, on a Pentium system with IDE integrated into the chipset going from 16 to 32 bits does indeed improve things a lot. But that's because you will be wasting way less CPU cycles on PCI latency if you can combine reads and writes. And even here it still requires that the drive is not stuck at PIO 0, because that will be the main bottleneck.

Did you read that thread I referenced? VLB can certainly be very fast for IDE even on a 486... I have even used SATA and PCI on a 486, it was very fast.

If OP want's to have VLB IDE then sure, why not, but I'm just saying it's not stricly required to get more peformance out of a HDD on 486 VLB system. On the other hand the VLB cards can be very picky and unstable, I'd rather have a VLB SVGA and no other VLB cards, and a stable system, than trying to get two VLB cards to cooperate - which is sometimes next to impossible to achieve.

You do make a valid point here... I have had stability issues with SiS and the combination you speak of. I'm still testing this though. I switched from EDO to FPM memory and I set the VLB to 1WS but I've not concluded my testing yet. Yes, I agree, better to have VLB SVGA and settle for plain old 16-bit IDE if that's your only option.

Deunan wrote on 2024-07-04, 13:39:

It's not about interface width, at least not much unless you can negotiate something faster than PIO 0 with the drive. This requ […]
Show full quote

mockingbird wrote on 2024-07-04, 10:40:

VLB certainly is faster, even in DOS... And VLB is a 32-bit interface.

It's not about interface width, at least not much unless you can negotiate something faster than PIO 0 with the drive. This requires a ROM BIOS extension on the card, or a device driver loaded (or at least run) during boot. In which case XTIDE will do the same thing, even for 16-bit ISA controllers, and you get most of the speedup without VLB.

Now, on a Pentium system with IDE integrated into the chipset going from 16 to 32 bits does indeed improve things a lot. But that's because you will be wasting way less CPU cycles on PCI latency if you can combine reads and writes. And even here it still requires that the drive is not stuck at PIO 0, because that will be the main bottleneck.

If OP want's to have VLB IDE then sure, why not, but I'm just saying it's not stricly required to get more peformance out of a HDD on 486 VLB system. On the other hand the VLB cards can be very picky and unstable, I'd rather have a VLB SVGA and no other VLB cards, and a stable system, than trying to get two VLB cards to cooperate - which is sometimes next to impossible to achieve.

There are a number of VLB cards that let you control the IDE speed with jumpers. They tend to be some of the earlier ones though, so they tend to be a bit limited. But they will be faster in DOS without an option rom or a driver.

Edit: My recommendation for VLB would be to get something with a promise ide 20630 controller. They lack native XtIDe universal bios (XUB) support for the faster speeds, but have the best drivers, best performance, and dma support. QDI vision QD658x0 and Promise 20230c would be my second choices because XUB knows how to make these contollers go as fast as the 20630 if paired with the right storage devices. No DMA support though.

For an ISA build, look for ecp or 16550 serial ports if that matters to you. There’s usually not a lot of difference between the different boards as far as IDE goes. They just pass the IDE device straight through to the ISA bus. Well, unless you can find one of these oddities:
https://theretroweb.com/expansioncards/s/siig … -isa-i-o-cn2424
https://theretroweb.com/expansioncards/s/gsi-inc-model-4c

Something that would have been a bigger deal back then compared to now is whether you want 16550 serial (for 56k external modem) or ECP/EPP parallel port (for zip drives, parallel port ethernet, etc.). Otherwise the ISA I/O cards are really a commodity part.

Alex_03 wrote on 2024-07-04, 00:37:

Are all IDE controllers bootable? should I just get an XT-IDE?

Thanks
Alex

Thats a little tricker. Yes, all should be bootable but…

Storage Devices <= 528MB shouldn’t give you any trouble. Larger storage devices likely require a BIOS upgrade, Option Rom like XUB (xtide universal bios can be put on an XTIde card or on an old network card or an ide card if they have a socket), or a drive overlay like EZ drive or Ontrack. They may also require a newer operating system like Win95b or newer.

I typed up a bunch of stuff here: https://www.vogonswiki.com/index.php/Storage

Feel free to give me feedback, while my writing skills are poor, I dont take offense easily so it balances out!

In general, avoid "Full AT height" ISA multi I/O cards with a lot of through hole components on. (Full AT height meaning extending over height of bracket a half inch) They tend to be long also. Those are quite early multi I/O cards, they tend to have problems in machines as bleeding edge as a 486, due to maybe not supporting 1.44 floppy drives, maybe not liking ATAPI refinements in later hard drives (i.e. do fine with 120MB and down, but the sizes that only came in the 90s, 420Mb say, not really for size but for more fully ATAPI standard, so 3 years or so newer, there might be problems/flakiness) They tend not to like an ISA bus faster than 8mhz, catch interference/noise and cause interference/noise to sound and network. And some issues occur where you're not sure if it's the CPU too fast for the card, or the HDD too fast for the card, unless using something from the dawn of ATA time, like a 40MB.

Anyway, leave those to folks who want to be ruthlessly period correct with their 286 and early 386 boards, and thus also need to use a lot of pre-1991-ish stuff with them to keep them happy, because a raging hot performance machine like a 486 doesn't like them.

Wow, I did not expect to get so much response in a short amount of time, thanks!

I actually do have two 8-bit isa SCSI cards and a couple of drives, but one is "bad" and the other I could not get it to be detected by the bios, I need to troubleshoot this at some point. I could easily go that route, but I would still not have any floppy controller.

I will shop around for I/O cards, VLB or otherwise, I am not too concerned with performance, I just like tinkering to see what I can do.

Thanks!

Alex_03 wrote on 2024-07-04, 15:52:

Wow, I did not expect to get so much response in a short amount of time, thanks! […]
Show full quote

Wow, I did not expect to get so much response in a short amount of time, thanks!

I actually do have two 8-bit isa SCSI cards and a couple of drives, but one is "bad" and the other I could not get it to be detected by the bios, I need to troubleshoot this at some point. I could easily go that route, but I would still not have any floppy controller.

I will shop around for I/O cards, VLB or otherwise, I am not too concerned with performance, I just like tinkering to see what I can do.

Thanks!

SCSI devices need BIOS extensions to boot, yes.

IDE drives don’t need bios extensions because they have Integrated Drive Electronics (IDE) that can act like an MFM drive attached to a WD1003 controller, and PCs have had support for that hardware in the system BIOS since the IBM AT / 286 days. If you go back earlier than a 286, like a 5160 XT or 5150 PC, then you need BIOS extensions in order to boot from an IDE drive. XT-IDE universal bios is a popular choice these days.

douglar wrote on 2024-07-04, 16:53:

IDE drives don’t need bios extensions because they have Integrated Drive Electronics (IDE) that can act like an MFM drive attached to a WD1003 controller, and PCs have had support for that hardware in the system BIOS since the IBM AT / 286 days. If you go back earlier than a 286, like a 5160 XT or 5150 PC, then you need BIOS extensions in order to boot from an IDE drive. XT-IDE universal bios is a popular choice these days.

There is an exception to this rule -- PCI IDE adapters in a 486 PCI platform need their own BIOS extension to boot -- unless you use the integrated PCI IDE controller. Why this is, I'm not certain, as Award 4.51 is fully capable of detecting an HDD connected to a VLB card, but not PCI.

mockingbird wrote on 2024-07-04, 21:44:

douglar wrote on 2024-07-04, 16:53:

IDE drives don’t need bios extensions because they have Integrated Drive Electronics (IDE) that can act like an MFM drive attached to a WD1003 controller, and PCs have had support for that hardware in the system BIOS since the IBM AT / 286 days. If you go back earlier than a 286, like a 5160 XT or 5150 PC, then you need BIOS extensions in order to boot from an IDE drive. XT-IDE universal bios is a popular choice these days.

There is an exception to this rule -- PCI IDE adapters in a 486 PCI platform need their own BIOS extension to boot -- unless you use the integrated PCI IDE controller. Why this is, I'm not certain, as Award 4.51 is fully capable of detecting an HDD connected to a VLB card, but not PCI.

Does that include early PCI boards with no onboard IDE, and using a PCI IDE controller with a "paddle board" to grab IRQ 14, etc. from a nearby ISA slot?

jakethompson1 wrote on 2024-07-04, 21:50:

Does that include early PCI boards with no onboard IDE, and using a PCI IDE controller with a "paddle board" to grab IRQ 14, etc. from a nearby ISA slot?

You got me -- I have no clue. But it seems like you're on to something.

mockingbird wrote on 2024-07-04, 21:59:

jakethompson1 wrote on 2024-07-04, 21:50:

Does that include early PCI boards with no onboard IDE, and using a PCI IDE controller with a "paddle board" to grab IRQ 14, etc. from a nearby ISA slot?

You got me -- I have no clue. But it seems like you're on to something.

Having read some of the PCI docs before, I think it has to do with "native mode" where the IDE controller waits to be assigned resources by the chipset, versus "compatibility mode" where it defaults to ISA-compatible settings.

Most addon IDE cards would be in "native mode" except those early multi-IO cards that perhaps would be in compatibility mode just like onboard PCI IDE.

jakethompson1 wrote on 2024-07-04, 22:02:

mockingbird wrote on 2024-07-04, 21:59:

jakethompson1 wrote on 2024-07-04, 21:50:

Does that include early PCI boards with no onboard IDE, and using a PCI IDE controller with a "paddle board" to grab IRQ 14, etc. from a nearby ISA slot?

You got me -- I have no clue. But it seems like you're on to something.

Having read some of the PCI docs before, I think it has to do with "native mode" where the IDE controller waits to be assigned resources by the chipset, versus "compatibility mode" where it defaults to ISA-compatible settings.

Most addon IDE cards would be in "native mode" except those early multi-IO cards that perhaps would be in compatibility mode just like onboard PCI IDE.

Could just be my hardware, but I was not able to get a 486 PCI motherboard to boot from a PCI IDE card, BIOS or no BIOS.

I have these Motherboards:
https://theretroweb.com/motherboards/s/anigma-bat4ip3e
https://theretroweb.com/motherboards/s/fic-486-vip
https://theretroweb.com/motherboards/s/pcchip … m919-ver-3.3b-f

I have a random assortment of promise and via cards that are ATA-33 and newer

I suspect that if I found a really early PCI IDE card, it might work, but, you know, why bother if it isn’t going to be faster than the integrated IDE ? (or in the case of the FIC board, VLB, since the bridged PCI design is noticeably slow)

douglar wrote on 2024-07-05, 18:57:

Could just be my hardware, but I was not able to get a 486 PCI motherboard to boot from a PCI IDE card, BIOS or no BIOS.
<snip>

It's your hardware... I was able to successfully use a relatively late Promise PCI card and it worked fine. The motherboard was a Shuttle HOT-433.