IDE to Compact Flash as MS-DOS boot drive.

[Paddan1000]

I would only have used the flash drive for MS-DOS and not for Windows, to avoid damage from .
I've removed it from the motherboard now. It took a lot of force and a pair of pliers, so I'd rather not put it back. I got the error message at boot, right after the memory test, so I don't think it had anything to do with it not being formatted. The MS-DOS installation program gave me an error about there not being a hard disk at all, and not just the one about the invalid drive specification.
I will try a 16 GB Samsung SSD later and see if I can use it as a secondary storage drive. It might be too big, but I know the motherboard accepts hard disks of at least 6 GB as slave drives, as long as I don't try to boot from them.

[bestemor]

Anyone managed to actually make usable partitions(2+) and then format these things ?
(granted, I'm using a 'real' hdd, 2,5gb microdrive, but still via CF connector...)

Meaning, that any logical partitions exists and gets assigned a drive letter in DOS and is visible ?

I have partioned and formatted oh so many times, and looks just fine in winXP etc, but in DOS there's no such luck...

I finally! managed to boot of my disk (primary partition, made in XP), but the rest of the disk is useless/practically invisible.

Is there any way to make use of an extended partition ?

I've tried several versions of FDISK.
- The one in DOS 6.22 reads the remaining space correctly(1424mb), but then says that max partion is only about half of that value(709mb)....
And, even if I agree to this nonsense, at reboot nothing has happend.

The one in Win98 seems to work fine(via USB cable/adapter), giving me drive letter at reboot. AND correct size values(!).
But when I then transplant the disk to the DOS machine and boot, nothing has seemingly changed:
Only C:drive is working/visible, rest of the space is wasted (though Fdisk still shows me what I am missing, sorta).


I then gather it is absolutely impossible to have more than 1 partition, if booting a regular DOS environment ?

Only thing left to try now, is kill the disk completely, again...., and let DOS(6.22) do ALL the partitioning and formatting natively. If that doesn't work....

(or maybe I am using the wrong DOS version? DOS 7 would perhaps be better? though I can't fathom why, if so...)

[laxdragon]

Wouldn't it be better to run Windows from a regular hard drive and only use the CF for file storage? The swap file is going to destroy a CF drive if it gets read and written to a lot.

The way around that is to use a RAM drive for swap. ramdrive.sys maxes out at 32MB though. For me that was enough to keep Windows happy and stable enough. A 3rd party RAM drive could created larger ones if needed.

[goldeng]

Hi guys,

Searched a bit on the forum and found this thread, and I think that this is the right place to post my issue. I'd really appreciate some help:

I'm working on an episodic retro PC build project which most the items in it are new-old-stock items that I purchased of eBay. Part 1 with all the unboxing of the items and the assembly is already on YouTube if anyone is interested (https://www.youtube.com/watch?v=9wHApPiQJH4).
I'm currently working on Part 2 and encountered some issues booting into MS-DOS 6.22 from a CF card, which is connected to a bracket Syba CF to IDE adapter. I'm not sure whether I did something wrong or perhaps something is wrong with my CF cards.

The issue is that after successfully installing MS-DOS 6.22, I restart the system but the operating system just doesn't boot! It just hangs on the "Starting MS-DOS..." line until forever. See screenshot:

I did manage to successfully boot into MS-DOS 6.22 once from all my installation attempts, and even installed Norton Commander and made an image with GHOST, which I also managed to restore from in one of my tests. But after I turned the machine OFF and then ON again, the operating system just didn't want to boot anymore. Nor I could find the image file that I created with GHOST which I saved on another partition on the CF card - which was showed correctly at the restore attempt that actually worked before everything went bad.

Things I tried:
1. I checked that the MS-DOS 6.22 disks that I have actually work - I created img files from them (original disks) and installed the operating system in Virtual PC under my Windows 7 machine. Disks and images seem to be okay as the img creation and installation in the virtual machine finished without any problem.

2. I boot from the MS-DOS 6.22 1st disk and run "sys a: c:", and also "fdisk /mbr". These attempts run successfully but didn't solve anything.

3. I'm using a Trancend 8GB CF 133x card. I tried to replace it with another one of the same model that I purchased together - no luck there either.

4. I replaced the Syba CF to IDE bracket adapter with a StarTech 3.5" bay CF to IDE adapter and tried deleting any pre-made partition on the CF card and installed MS-DOS 6.22 once again from scratch - same problem during boot.

The one odd thing I did noted is when I connected the CF card to my Windows 7 machine: I found that some of the files and folders in the file structure seem to be corrupted from some reason. See screenshot:

Could it be the problem that prevents the system from booting correctly?
Did anyone else here examined the CF card on another machine and also found the 'broken' files and folders?

At the meantime, I just connected a regular 2.5" HDD to my build and installed MS-DOS 6.22 on it. But it's mainly intended to be used for a stand-alone Win98SE installation. So I'd really wish to be able to install and run MS-DOS 6.22 from a CF card.

Any assistance would be much appreciated!

Thanks,
goldeng

[agent_x007]

How did you format your 8GB card (partition sizes) and what program did you use ?

[goldeng]

Hi agent_x007, thanks for the fast response.

I tried both fdisk and Partition Magic 7. Partitions were formated in FAT so it recognizes much less of course.
Drive C:\ is 25% from the available capacity (118MB), and the reset 75% is the extended partition which divided into drive D:\ and E:\
The purpose of this division is so drive C:\ will be only for the OS, drive D:\ for games and the rest of the files, and drive E:\ for the image backup of drive C. But I also tried installing the OS on a single partition where I let the DOS installation to take care of the partitions.

[agent_x007]

I would try installing DOS on CF with other PC, and then checking if other PC boots from it.
Also are ribbon cables themselves 100% fine ?

[goldeng]

Mm.. Most of my computers don't have an IDE connector. And in this specific case it's a brand new motherboard that I recently acquired of eBay. New-old-stock. All cables are new and intact. If anything was with the cables then the physical HDD wouldn't have worked too.

I'm not sure but I think that the brand of the CF cards that I have just don't fit for this build. Maybe it's the capacity of the 8GB, or maybe even their speed (133x). I'll try to find another (small) CF card and will update here how it goes. But if you'll think about any other ideas please let me know.

[Malvineous]

I've encountered this problem on a few PCs. The issue is that older computers address data on a disk by using a cylinder, head and sector number. This CHS addressing was left over from the earliest hard drives where it referred to the physical location of the data on the magnetic platters. When IDE was introduced, the drive itself handled the location of the data and so the PC only needed to supply a block number and the drive's electronics would figure out where on the magnetic platters it might be. This was called LBA.

The complication is that DOS still uses CHS to access the disk, and so the BIOS needs to translate that into LBA addresses to send to the IDE drive. A standard formula was created that did this conversion, and so everything that uses this formula will consistently see the same information - specifically if one system writes to C=1, H=8, S=32, then another PC reading C=1, H=8, S=32 will see the exact same sector.

The problem is that it seems some BIOSes implemented this formula incorrectly (or perhaps did so before it was standardised) so on those PCs, the LBA blocks on the disk will appear to be out of order because of the different translation result. The boot sector at LBA=0 (CHS=0,0,0) will always work, so you'll always see "Starting MS-DOS", but the next sector read tends to be wrong and the system will hang.

When you wrote to the disk, the data you wrote ended up in the wrong sectors, and this is why you see corruption when you put it back into your Windows machine.

There are only two solutions I know of for this problem:

1. Don't move the CF card to any other computer. Leave it in your DOS machine permanently. Format it and install DOS on the machine itself, using floppy disks if you have to. Although the CHS-to-LBA formula is wrong on that PC, it is consistent so the card will work fine, it just becomes incompatible with other machines and will become corrupted if you try to access it elsewhere.

2. Install the XT-IDE BIOS, e.g. in your network card's boot ROM socket. XT-IDE will take over from the system BIOS and handle the IDE drives, and it uses the correct CHS translation algorithm so moving the card between your DOS PC and another system will work just fine. Note that in this situation you might need to configure your system BIOS to have no hard drives, as you don't want it accessing the CF card. That should be handled entirely by the XT-IDE code.

There is a third option - patching your system BIOS to use the correct translation algorithm - but that is probably impractical for most people, myself included

[matze79]

You can also use a Network Card and place a Eprom with XTIDE Bios on it.
This will remove your Limitation and you don't have to use a damn Software Solution.
If you want to pull out the card and use it under windows to copy Games on it, the DDO will eventually prevent you from doing that.
Because Windows will not recognize a FAT Partition, instead it will find a DDO Partition that it can't recognize.

[Jo22]

It's just a wild guess, but maybe both BIOSes differ in the total amount of usable cylinders they do report via the old function calls.
This could result in some kind of garbled data, as only one PC sees the "other half" of the card.

The Windows NT machine is likely not bound to the usual DOS or in13th limits.

https://www.win.tue.nl/~aeb/linux/Large-Disk-4.html

[goldeng]

Hi Malvineous, matze79 and Jo22!

@Malvineous:
I very appreciate your detailed comment!
I tried messing around with the LBA settings in the BIOS as I assumed that hey, this is not a mechanical hard drive and so there's probably a setting to tell the BIOS that this is a fixed and non-moving drive. But nothing that I did really helped. =\

1. Don't move the CF card to any other computer. Leave it in your DOS machine permanently. Format it and install DOS on the machine itself, using floppy disks if you have to. Although the CHS-to-LBA formula is wrong on that PC, it is consistent so the card will work fine, it just becomes incompatible with other machines and will become corrupted if you try to access it elsewhere.

I don't think that it's the case, as the last thing that I did when everything didn't work is to pull the card out from the retro PC and to see if there's anything on it on my main Windows 7 machine. It was connected to the retro PC all the time when the issue occurred.

2. Install the XT-IDE BIOS, e.g. in your network card's boot ROM socket. XT-IDE will take over from the system BIOS and handle the IDE drives, and it uses the correct CHS translation algorithm so moving the card between your DOS PC and another system will work just fine. Note that in this situation you might need to configure your system BIOS to have no hard drives, as you don't want it accessing the CF card. That should be handled entirely by the XT-IDE code.

Weren't aware of this. So I just watched some videos on YouTube but found that it's not something I'm willing to get into. I wanted something simple and thought that a CF card would be it. Apparently it requires more work than one thinks.

There is a third option - patching your system BIOS to use the correct translation algorithm - but that is probably impractical for most people, myself included

Yep, I'm with you on this - patching the BIOS is not something that I'm willing to do, although I have experience with BIOS firmware updates. It just doesn't worth the risk (especially with retro motherboards) as you can end up with a non-booting machine if something goes wrong, and there's no way back.

@matze79:

You can also use a Network Card and place a Eprom with XTIDE Bios on it.
This will remove your Limitation and you don't have to use a damn Software Solution.
If you want to pull out the card and use it under windows to copy Games on it, the DDO will eventually prevent you from doing that.
Because Windows will not recognize a FAT Partition, instead it will find a DDO Partition that it can't recognize.

As I wrote to Malvineous, this is not something that I'm willing to get into. But many thanks for the suggestion! It's always nice to learn and get aware of new things! Appreciate it.

Jo22:

It's just a wild guess, but maybe both BIOSes differ in the total amount of usable cylinders they do report via the old function calls.
This could result in some kind of garbled data, as only one PC sees the "other half" of the card.

The Windows NT machine is likely not bound to the usual DOS or in13th limits.

https://www.win.tue.nl/~aeb/linux/Large-Disk-4.html

Thanks for the reply! I guess that you're referring to the way the file structure of the card showed in my main machine. I'm not sure if that's correct as I saw many people that actually move the CF card from their retro machine to their modern one for transferring data without any problem.


I think that I'll just try another card brand and perhaps in a smaller size and will update the results so we could all learn from it.

Many thanks again to all of you for looking into this.
And if anyone else see this post and have more information, I believe that we'll all be glad to hear and learn from your knowledge and finding.

Cheers

[Malvineous]

I tried messing around with the LBA settings in the BIOS as I assumed that hey, this is not a mechanical hard drive and so there's probably a setting to tell the BIOS that this is a fixed and non-moving drive. But nothing that I did really helped. =\

Yep, solid state drives weren't really a thing until many years after these PCs came out.

1. Don't move the CF card to any other computer. Leave it in your DOS machine permanently. Format it and install DOS on the machine itself, using floppy disks if you have to. Although the CHS-to-LBA formula is wrong on that PC, it is consistent so the card will work fine, it just becomes incompatible with other machines and will become corrupted if you try to access it elsewhere.

I don't think that it's the case, as the last thing that I did when everything didn't work is to pull the card out from the retro PC and to see if there's anything on it on my main Windows 7 machine. It was connected to the retro PC all the time when the issue occurred.

So to confirm, you originally formatted the CF card in your retro PC ("format c:"), installed DOS on the retro PC, and the CF card never left the retro PC until you moved it to the Windows machine? If you didn't format it in the retro PC then that's almost certainly the ultimate cause as the filesystem would have been corrupted from the beginning. If you did, then it's very odd that it doesn't work. I'd even suggest running fdisk, deleting the partition and re-adding it, so the CHS values in the partition table match what the BIOS is using.

I have also seen this kind of behaviour when the BIOS settings for the first hard drive have changed after it has been in use, so if you have them set to 'auto' it might be a good idea to set them manually. Hopefully your BIOS is new enough that it has an autodetect feature you can use to get started. Whatever settings you use, make sure you *write them down* because if you ever change the settings (different number of heads etc.) then you'll have to reformat the CF card and start from scratch. This can happen if your CMOS battery goes and the BIOS gets wiped, and you need to re-enter the settings.

I think that I'll just try another card brand and perhaps in a smaller size and will update the results so we could all learn from it.

Pretty sure a different card won't help, unless the one you have happens to be faulty (pretty unlikely, but you could run Scandisk on it in Windows to check.) There's definitely a solution though, so best of luck!

You don't say what motherboard, CPU and BIOS vendor you have. I've only seen 486 AMI BIOSes with this problem, and I'd be surprised if a Pentium 2 or newer had the issue, so I'm curious what platform you're using.

[goldeng]

Actually I left a YouTube link to the 1st part of my Retro PC Build project at my first post in this thread. It contains all the information of the items that I used in this build. But anyway, it's a Pentium II 400Mhz slot 1 processor, on an Intel SE440BX-2 motherboard which has the PhoenixBIOS 4.0 Release 6.0.

And yep, formatted the card automatically via MS-DOS 6.22 installation, and afterwards manually in FDISK after deleting and re-creating all the partitions. Not sure what's going on but it's sure frustrating...

[Malvineous]

Interesting, by that era the BIOS should have full LBA support so you should be fine. Maybe trying another CF card is indeed the way to go...

[Jo22]

Yup, I'm also a bit surpsried. I wonder if the BIOS is configured to force LBA mode or if it is still set to "AUTO"..
Since the Compact Flash card is an ~8GB model, it might be still within the realms of the old L-CHS/P-CHS translation model.

http://web.allensmith.net/Storage/HDDlimit/Address.htm#Enhanced

Also interestiing, but loosely related to this: The 4096 Cylinder BIOS Barrier

PS: Patched BIOSes used to be available at http://wims.rainbow-software.org/ and https://www.wimsbios.com/large-hdd-patched-bios.jsp

[goldeng]

Thanks guys. I'll update as soon as I have information whether it works or not with a different CF card.
Patching BIOSes is very tempting due to the many features that are added afterwards, but also risky. I don't want to end up with a 'bricked' motherboard and prefer to stick with the state that it came in.

The installation with the CF cards is mostly for experiment. Eventually I'll just use regular HDDs with the SATA to IDE converter described by @PhilsComputerLab here https://www.youtube.com/watch?v=Edmg43t28jg&t=11m6s (11:06)

I'll probably use 2 HDDs: 1st for native DOS 6.22 with Norton Commander, and the 2nd for Windows 98SE and to change their availability in the BIOS according to which drive I wish to boot from. Disregarding the above adapter for a minute, I was thinking about going with a regular 2.5" IDE drives over 3.5" as they are more portable and doesn't require power adapter. I already have 1 IDE drive which I connected to my Retro PC with an IDE 2.5" 44pin to IDE 3.5" 40pin adapter. Works great.

Any recommendation regarding 2.5" IDE HDDs? Specific model you prefer? Capacity amount to pick?
eBay is loaded with new IDE drives in very decent prices. I was thinking to go with Seagate brand as most of my hard drives are from this particular brand and almost never let me down. The capacity I was thinking to go with is between 80GB to 120GB, as the maximum amount that Win98SE is capable of recognizing is 127GB. Speed isn't a factor here as we're talking about very old operating systems. So 5400RPM will do. Though I also saw 4200RPM on eBay... Or perhaps you prefer to use the adapter Phil presented in his video with a regular 3.5" SATA drive, say 2TB, and to use some software to limit the size that the BIOS will see?

Thanks!

[Malvineous]

What do you mean by a 2.5" drive not requiring a power adapter? A 3.5 drive requires an IDE cable and a molex power connector, a 2.5" drive requires this plus a 44-pin adapter as well, so it seems it would need more parts unless I am misunderstanding something.

Personally I've always avoided 2.5" drives as being laptop drives they were always slower for a given size than the 3.5" version. Back in 1996 in my Pentium 100 I had a 4200RPM 3.2GB Seagate drive and boy was it slow. 3MB/sec IIRC. Even on a P100 it was noticeably slow. When I later moved it to my 300MHz Celeron machine, I got so many coasters out of the CD burner because the hard drive couldn't keep up (and this was burning at 4x which is only 600kB/sec!) So definitely go for the faster drive if you can - retro PCs are slow enough already!

Brand doesn't really matter. There have been studies over the years that show that apart from particular models that are extremely unreliable, each brand of hard drive is on average just as reliable as the next. If you've had good experiences with Seagate, go with them. Personally of the seven Seagate drives I have ever bought new, only the 80MB one from 1992 is still going, all the ones with capacities measured in gigabytes have long since died. Mind you so have all my Western Digital drives older than four years, and two of those died just sitting on a shelf doing nothing. Ditto for Samsung. Guess that's the drawback of having them running 24/7 in a country where 30C/86F is considered a cool summer day. The only drives I have that work are either new and in warranty, or somewhat ironically ones I got second hand for free. So if you pick them up second hand you might have some luck

I've also had compatibility issues with the SATA/IDE adapters so I steer clear of them, but that's not to say they don't work, just that they seem to be a bit picky with the hardware they work with so you might have to experiment a bit. Unlike the 44-pin IDE adapters which only passively convert one plug to another, SATA/IDE adapters need to translate the protocol and this is done with varying degrees of success. Things like using drive manufacturer diagnostic tools almost certainly won't work through one of these adapters as they tend to use undocumented/proprietary IDE commands which won't get translated.

If you want to use SATA drives and you've got a 440BX/BX2 board then a more reliable solution (if you actually run into trouble, you might not) is probably to get a PCI SATA controller as they are fairly cheap at the moment, although they are starting to become less common. If you can get one that will run in compatibility mode (i.e. hardware looks like a normal IDE controller) then you don't need special drivers either, it just appears like a standard IDE controller. Otherwise just pick one that comes with drivers for your chosen OS. I'm sure some will have Win98 drivers.

Don't forget that you can often restrict the capacity of large drives using jumpers on their underside or software config programs provided by the manufacturer, so if you can save some cash by getting a drive larger than you were expecting, you may as well. In my case I find the unreliability of drives just plain annoying, so I went with the SSD option. My old machines (486 and older) use SD cards connected as IDE devices, and my newer (PCI) machines use either CF cards or SATA SSDs. SanDisk make a nice 16GB mSATA SSD for AU$20 (~US$15) that works really nicely in a retro PC, and is as fast as the machine can handle so you're never waiting on the disk. Luckily for me I don't miss the sound of a spinning hard drive, having found it annoying back in the day, so I'm only too happy to have a quiet retro experience

[goldeng]

What I meant is that 2.5" drives don't require any molex cable to be connected to them to consume power, as opposed to 3.5" drive that do. The additional 4 pins on the 44pin IDE connector (actually only 2 of them) transfer the power to the drive. Though when connecting to a desktop machine you should use such adapter like in the picture I provided above, which has a molex connector hooked to these 2 pins that provide power to the drive, as the IDE on desktop motherboards doesn't provide any power.

Thought also to use a PCI SATA controller, but weren't aware that some have a 'compatibility mode' that makes the computer to see the SATA drives as IDEs. But I'm not sure whether it works naively in DOS or only in Win98 with appropriate drivers. Do you have any experience with these PCI SATA controllers?

[Malvineous]

Yep that's what I thought - a 2.5" drive needs an extra part that you don't need with a 3.5" drive

Not a huge amount of experience with PCI SATA controllers. I know for sure Intel ones have compatibility mode, but I have a feeling you can only get these as part of a chipset and not a PCI card. The Silicon Image ones seem to need an onboard BIOS in order to boot from them, so that would provide DOS compatibility however for Windows, which prefers to speak directly to the hardware, you'd ideally want drivers otherwise Windows will operate in a compatibility mode which slows things down as it then uses the BIOS rather than direct hardware access. So it's only worth it if you can find a card with a SATA chipset that has compatibility mode and/or drivers for your OS (onboard BIOS for DOS, driver files for Windows.)

You'd likely have to find the datasheets for the SATA chip used on each PCI card you come across and look it up to see if it provides this. I haven't done this yet but it would be an interesting project, as retro-compatible SATA cards are likely to become more popular in the coming years as IDE drives wear out and SATA ones are all that's available.

EDIT: A quick look (eBay search for "pci -express sata -graphics -power -msata -m.2 -pcie -pci-e -cable") seems to show quite a few with the VIA VT6421 chipset, which hints on that page that it is backwards compatible with existing operating system drivers from XP back to Win9x. Not sure whether they mean only the IDE part (it's an IDE + 2x SATA device) or whether they include the SATA parts in that, but it could be worth investigating further.