A similar problem has been bugging me for some time.

I found that DOS, win95 and win98 don't like it when you partition HDDs with anything other than FDISK. Make a win95 boot disk and partition the drive with FDISK - otherwise win95 will keep bugging you and telling you that the drive needs to be scanned / will not boot / etc - I can't explain why - maybe it's the size difference you mentioned? But partitioning HDDs for win9x machines with winXP and newer causes all sorts of problems like:

- DOS / win95 / win98 and sometimes winME refuse to boot off the drive
- constant recurrent nagging from scandisk telling you the drive needs to be checked - even if the drive has been checked just before
- weird files appearing on the drive
- data corruption
- bad sectors that go away when re-partitioned with FDISK

So I recommend, OCD aside, that you partition the drive using FDISK.EXE and format using FORMAT.COM off a win9x boot floppy.

Unfortunately, it appears as if more is going on here than just the method of partitioning. I connected the drive to another DOS machine and was able to partition the entire thing with FDISK just fine. I stuck it back in the 486, and it's no longer showing any bad sectors. But, it won't boot no matter what I try. It just gives an error that says "No ROM BASIC, system halted". This still suggests perhaps an incompatibility between this specific model of drive and controller card (Or more specifically, this particular revision of this card). If that's the case, then I'm concerned about the long-term reliability regardless of how the drive is partitioned. I've ordered a different controller card, I'll see how things turn out with that one.

Maybe this is a stupid question, but do both the 486 and the laptop detect the drive with the same geometry?
Is the HDD bootable if you partition and format it in the 486? In that case I'd probably leave it at that.
Is it possible that the size discrepancies result from different partitioning tools displaying the size in different units (ie. 1024/1000 based kB)? Maybe it is something crazy as using a 1024x1000b sized MB.

I bet this is a geometry translation error. There should be no 800MB barrier.

Physically the HDD uses 1652 cylinders, 16 heads and 63 sectors.

If the numbers seen by DOS are different, the geometry is translated by BIOS. And because DOS uses standard BIOS, this limits the cylinder value to 1024, so it has to be translated to use the whole disk.

For example, DOS could see the disk with 826 cylinders (halved) and 32 heads (doubled), still 63 sectors, in one machine. Other machines could translate the geometry differently, and therefore partitioning, booting and using the drive won't work across machines.

Check if you have multiple choices in BIOS/controller card how it performs translation. It is possible it performs the translation in a way that discards some sectors.

Other than autodetect or the option to manually enter the CHS parameters, the only BIOS options I can see dealing with the IDE controller are to enable/disable LBA, enable/disable block mode and enable/disable 32-bit transfer. Obviously LBA needs to be enabled since the drive is larger than 512MB, but otherwise trying every combined iteration of the other two settings didn't make any difference.

The drive itself does have an additional jumper that selects between "Conner" and "ATA/CAM" modes, which according to Google simply changes the method the drive uses to determine whether a slave drive is present or not (This system does not have one). I still tried changing the jumper anyway, no change. I currently have the drive manually jumpered to Master. I could try Cable Select, but I presume that wouldn't make any difference.

I also checked the jumpers on the controller itself. For the IDE side, there are jumpers to enable/disable IDE, enable/disable the secondary IDE channel, detect/not detect the VLB bus frequency, change the address of the controller ROM and allow "non-disk" IDE devices to be connected to it (I presume this means CD-ROM, can't find any details about it). I did try different iterations of the IDE jumpers, but every result was either no change, no boot or no POST. One of the settings even wiped the CMOS and I had to set everything back up. 😒

Like I said, the BIOS itself has no trouble auto-detecting the full capacity of the drive. The CHS paramaters it provides from auto-detect match the label on the drive. So, either it's the controller card tossing up this 800MB limit, or for some odd reason it's DOS itself causing this problem. I did try running FDISK from a Win98 boot disk to see if maybe somehow it was the older version of DOS, but it didn't make any difference. Of course, had I remembered that I just a couple months ago partitioned a 2GB drive under DOS 6.22 on a different system, I would have realized this was a worthless test. 😵

This is really, really weird, because like I said this exact same combination of motherboard, controller card and hard drive (Albeit likely different revisions) would allow me to partition to 810MB on the original system from my childhood. Granted, just like what was happening on my current rebuild, there was the large chunk of bad sectors appearing in the middle of the drive on the original system, which implies an incompatibility of some sort. Now, I've been able to partition it to 810MB (In fact actually 812MB) using another DOS machine, and although it's no longer showing any bad sectors, I still cannot make it bootable. I've verified the partition is set active and I've even run FDISK /MBR to no avail.

At this point, I'm just gonna wait for the new controller card to arrive. I ordered one from the same time period (Late 94-early 95), but made sure to get a different chipset just in case it is a rogue incompatibility between the CFA850A and Winbond chipsets.

My 425mb drive gets detected as 405mb on my 486 (VLB controller). I just live with it. Though it is pretty annoying. Free space is little enough after installing win95 and components, drivers, and whatever extras you might require.

Indeed, if this different controller doesn't fix this problem, then I'll have to learn to live with 800MB instead of 810MB. Like I said, I know it's not really much difference at all, but my nostalgic OCD is throwing a fit about this. 🤣 But yeah, if it comes down to it, I'm willing to live with 800MB. I knew going into this rebuild that it would be impossible to build something 100% identical to the childhood original, so I can learn to deal with this. At least I'll still have the comfort of knowing the correct model of drive is installed, even if it's not partitioned the same as the original.

Blurredman wrote:

My 425mb drive gets detected as 405mb on my 486 (VLB controller). I just live with it. Though it is pretty annoying.

If it says on the drive 425MB, it is 425 million bytes (1000000) which equals 405 megabytes (1048576). But this is a different issue, 850 million bytes should be 810 megabytes. If there is any DOS tool to print out the geometry DOS sees you should (chkdsk, MSD, scandisk?). At least you can use Debug to read MBR or DOS partition boot record and see a hex dump on screen.

I noticed there are CFS and CFA models of 850MB Conner and the other has 1651 cylinders and the other has 1652. But still, that would only account for 504 kilobyte difference.

Chkdsk and Scandisk don't display the CHS parameters of a drive. I don't recall if MSD does or not, I'll check that after work. Beyond that, I imagine there are diagnostic tools that will provide this information, I just don't know any of them off the top of my head, primarily because I've never used any such tools before.

Ok...and for mysterious reasons I am unable to determine, it has suddenly begun working perfectly fine. FDISK can now see and partition the entire drive, Scandisk finds no bad sectors, and it is bootable. The ONLY thing I have done with this computer since last night was replace the video card with a different model. That's it. I haven't changed any other settings. I find it highly unlikely that changing the video card would affect how the hard drive works. So, instead of being stumped as to why it wasn't working...now I'm stumped as to why it suddenly IS working! 🤣