Back in 1997 when I wrote my "An Old Man Reminisces" file, I complained about what the internet was becoming. I complained about how more and more people don't seem to care about anyone else's internet experience, as long as they can extract some small benefit for themselves. I asked, "Who knows what lies waiting down the line?" I guess we've seen part of the answer:
Like my SCSI page, this page is a fairly long, technical, and somewhat dull description of what I did to my computer to make it do something. It's probably not of interest to anyone other than people who have similar hardware, or those who are interested in the technical aspects of different computer setups. Technical information comes from many sources - primarily web pages but also from other people and books.
Mini-Disclaimer: What I have typed here is correct as far as I know, but I make no guarantees. Much of my experience is recounted from memory, and (especially early on) I didn't make many notes about how every little setting worked. Use at your own risk. Several names and products mentioned here may be trademarks or under copyright, and I'm not associated with any company mentioned, except sometimes as a customer.
Before I begin, here's a bit of background: my computer uses AMIBIOS 1.00.09.BC0R on a Packard Bell 520R motherboard. It is a Pentium-60 using the Intel Mercury Chipset. You can read more about my SCSI card and SyQuest drive in my "My experiences with Belkin's Parallel to SCSI Adapter, SyQuest's EZFlyer, and Adaptec's AHA2940UW" page. Looking back at that page (especially the last few lines), I see I was much more optimistic than I am now about the life expectancy of my computer.
In August of 1998 I decided that I should really get a larger internal hard drive. Even though the SyQuest cartridges had been very useful in keeping large programs, sound files, etc. off of C:, it had been cramped for several months. (My original drive was a Seagate ST3660A 520MB drive.) I gave in, and bought a 6.4GB Fujitsu MPB3064AT IDE drive. I thought that installing it would only take an afternoon, maybe even just an hour or two. After all, there's no external cables or cards involved. As it turned out, it would be over four months later before I finally got it working properly and was using it as my main drive.
I only had a 5 1/4" drive bay available in the computer, so I had to get a mounting bracket to hold a 3 1/2" sized hard drive (both my old and new drives were physically the same size). My first problem: the holes for screws didn't align properly between the drive bay 'frame' and the mounting brackets - two holes in the bay along one side needed to be about 2mm further in from the edge. Using an old drill bit as a file, I carefully filed an oval-shaped hole using the existing round hole. The hole now let me insert the screws to attach the mounting bracket. Then, screws through other holes on the mounting bracket kept the drive in place. My computer has a desktop (non-tower) case, and the whole 'frame' of bays could be removed. I could file a hole without getting tiny metal shards all over the motherboard and other drives, provided I removed the other drives from the frame too.
It took me a couple of hours just to get the drive in place. Now that both my old and new drives were secure inside the computer, I removed the IDE cable from my original drive and plugged it into to the new drive. I also connected an unused power cable from the power supply. I turned on the computer, and instead of automatically detecting the new drive during the BIOS initialization (as it always did with the old drive), the computer froze. Hmm... maybe I'd also have to change a setting in the BIOS (by pressing F1 when the "Press F1 to Enter Setup" message appears.) But there's a problem - my computer only goes into Setup after it has detected all drives and has let the BIOS on ISA and PCI cards run. It freezes before it goes into setup, so I can't change settings when the drive is connected. Not a major problem, though. I'd just have to disconnect the drive first.
Before I bought the drive, I had read about the 504MB and the 2.1GB hard drive size limits caused by older (generally dated mid-1994 and earlier) BIOSes. Most hard drive manufacturers' web pages explain the situation. One page said that a freeze-up soon after turning on the computer may be due to several causes, most likely either an incorrectly connected cable, or a BIOS which doesn't support drives larger than 2.1GB. This 2.1GB limit occurs when the BIOS can't handle more than 4096 cylinders (they only accept a 12-bit binary number). My cable seemed fine - both a new cable and the old drive's cable produced the same result. I was hoping my BIOS (which was dated 11/09/94) wasn't one of those with a limit. I knew it didn't have the 504MB limit - my original drive was 520MB and the BIOS supported translation modes.
(A side note: Both the IDE interface and Int13 support large drives, but the combination of the limits results in 504MB. Int13 allows 1024 cylinders, 255 heads, and 63 sectors per track. The IDE interface allows 65536 cylinders, 16 heads, and 255 sectors per track. Thus, the combination is limited to 1024 cylinders, 16 heads, and 63 sectors per track.)
I went into BIOS setup and changed the "Auto Configured" setting for drive C: to "User Defineable". I kept "IDE Translation Mode" set to "Auto Detected", and then entered the number of cylinders (13410), heads (15), and sectors per track (63) for the new drive. According to what I had read, BIOSes that don't support more than 4096 cylinders usually don't let you enter a larger number into BIOS. Hopefully my computer wasn't an exception, and perhaps the BIOS supported more than 2.1GB after all.
I rebooted using my "Windows 95 Emergency Disk". Good. The computer didn't freeze any more. I ran fdisk, planning on making a 2.1GB partition. (2.1GB is the largest partition size the FAT16 filesystem allows - this is not related to the BIOS limits but is because FAT16 only allows up to 65535 clusters or 'allocation units' on a disk.) Fdisk insisted my drive was a 472MB drive - this works out to 15 heads and 63 sectors, but only 1024 cylinders. (1024 cylinders is the highest the BIOS Int13 interface allows, so translation is used to remap the drive's real parameters into parameters Int13 supports. For example, doubling the number of heads and halving the number of cylinders presents DOS with the same size drive but with parameters DOS accepts.) I eventually concluded that the BIOS didn't seem to be doing the translation properly for drives larger than 2.1GB.
The drive has a jumper which makes the drive act like a 2.1GB drive. Connecting the 2.1GB jumper made the drive, BIOS, and DOS work perfectly, even in "Auto Configured" mode in BIOS. Of course, this meant that two-thirds of my drive was left unusable. I could prove to myself that the drive itself worked, but I still wanted to find a way of letting me use the entire drive.
Most web pages suggested three ways to get around the limits. First was to upgrade the BIOS to one which supported larger hard drives. The second option was to use a drive overlay program, which is a small program which loads before the operating system boots and which provides large hard drive support by taking over the BIOS's IDE routines. Third was to buy an IDE controller card, which plugs into an ISA or PCI slot, and the drive then plugs into into the card. This card uses its own IDE BIOS routines, so the motherboard's old routines are not used. One other option not mentioned on most web pages was that of getting a SCSI drive instead, which has much higher size limits. However, one look at the prices of SCSI drives made me eliminate that option. (They're usually at least twice as expensive than similar-sized IDE drives. Most sources I've read state that SCSI only provides a significant performance increase in situations such as workstations, network servers, and true-multitasking OSes (not Windows 95, apparently). Adaptec's Performance Analysis page disagrees with respect to Win 95.)
I asked Packard Bell about BIOS upgrades and they told me there was not an upgrade for my computer. (The only version of BIOS available to download from their web/ftp site was older than what I already had!). I didn't really want to use an overlay program (each drive manufacturer has its own version for its drives, with names like "Max- Blast", "EZ-Drive" and "Disk Manager"). They have unusual consequences, such as requiring you to boot from a certain floppy disk or press SPACE before you can boot your computer from a different floppy disk. I wasn't sure how well it would work (if at all) if I later installed another operating system in another partition (such as Linux). Would I be able to use an "OS chooser" like LILO or PowerQuest's "Boot Manager"/"BootMagic"?. I had one free ISA/PCI (shared) slot in my computer, and I decided the most "natural" way to go was to get an IDE card, provided they weren't too expensive. They provide a real, large-drive-supporting on-board BIOS which should work with any OS. I could connect my drive to this, and disable the motherboard IDE connectors (the motherboard BIOS allowed them to be disabled.) Or, some of these cards 'enhance' the motherboard BIOS, I could keep using the motherboard connectors with large drives.
After much internet research (and talking to a couple of people), I decided to buy a GSI model 2C. It is an ISA card with two IDE connectors labelled "X" and "Y". They can be used to replace the motherboard's connectors or to add two more, for up to eight drives (two per connector, as is standard for PC IDE). It has an on-board BIOS (version 5.03) which supports large drives and enhances the motherboard BIOS. Its BIOS can also be Flash-upgraded, meaning that I can use software to upgrade the BIOS without having to physically replace chips. Someone I know used an earlier GSI model and had no troubles. There was also a Canadian distributor for them, so I wouldn't need to worry about border fees or cross-border postage. The place I ordered the 2C from (via the internet) was out of stock at the time, so it took a couple extra weeks for me to get it.
I set it to no-IRQ mode (so the connectors on the 2C are disabled and the card works in the 'motherboard-enhancing' mode). I connected the new drive up (without the 2.1GB jumper) and turned on the computer. Disappointment. It froze the same way it had before without the card. Disconnecting the drive still caused a freeze. I didn't have any free IRQs to set the card to use, but I tried different IRQ settings on the 2C anyway (keeping the drive connected to the motherboard. It still froze. I tried different BIOS address settings for the card, and still got freezes.
While asking others for advice, I was asked if I had tried disabling the motherboard's IDE connectors and connecting the drive to the GSI 2C. I didn't think there was a setting in BIOS to disable them, but looking through the BIOS Setup menus more closely, I found that I could. I could then set the GSI 2C to use the IRQs the motherboard's IDE connectors previously used (14 and 15). I disabled both the motherboard IDE connectors ("PCI IDE Interface" and "Standard IDE Interface") and connected the drive to the GSI 2C "X" channel. Great news! No freeze-ups! (I did have to change which BIOS address the 2C used. I think C800-CBFF made the SCSI card BIOS stop working, and no GSI BIOS message came up at all with E000-E3FF. CC00-CFFF would later make EMM386 be "Unable to set base frame address" when Win95 was fully installed. D000-D3FF was the only setting that always worked.)
When it detects an unpartitioned drive, the GSI 2C offers to partition and DOS-format it for you. I initially said 'no' to the offer, and went on to use DOS's fdisk. The first time I tried, fdisk told me my drive was somewhere around 3.5GB. I wasn't 'out of the woods' yet. After 'playing around' and making a few partitions with fdisk and removing them, fdisk started reporting another different value for the size of the disk. I deleted all partitions, and then when the 2C asked me about making a partition, I said "Yes". Fdisk always reported the correct size after that. I'm still not sure what was going wrong with fdisk, but after a lot more experimenting, it looked like that problem was gone for good. The drive was now working okay under Windows 95's DOS.
It was now time to copy the contents of my old drive onto the new one. After re-installing the old drive on its original connector and entering the previous BIOS settings, I ran SyQuest's EZ-Backup program. I used EZ-Backup rather than Microsoft Backup because it supported spanning multiple cartridges, and it also had a DOS-mode recovery program. (I didn't realize MSBACKUP could be run from DOS). EZ-Backup was also free, while other commercial programs were not.
More problems! The first time I ran EZ-Backup, I couldn't eject the cartridge to insert the next one, so I had to restart. Once finished, I re-attached the new drive and card, booted Win 95's DOS, and loaded Adaptec's and SyQuest's DOS drivers. It took a couple of hours to extract the files back onto the new drive. It then gave me a list of 60 files it could not properly restore. They were all part of the \WIN95\SYSTEM directory too, so it wouldn't be a case of simply re-installing one application. Rather than make another backup, I decided to attach both drives to the GSI 2C and try copying back the files one-by-one. Just my luck - many were 'read-only' or 'system' files, so a simple 'copy d:\win95\system\m*.*' would skip hidden files and abort the copy operation when a destination file was read-only. Using DOS's Attrib command to switch settings back and forth for so may files was a hassle. I finally gave in and decided to do another backup.
Scandisk reported surface errors on the SyQuest cartridge I used - the first time I'd ever had surface errors on a cartridge. After trying to backup onto a different cartridge (which was error-free initially), I found surface errors on it, too. It seemed EZ-Backup was somehow physically damaging the cartridge. I spent several days with SyQuest's verify, format, and bad-sector check utilities, and cartridges fixed or deemed 'good' by them would sometimes go 'bad' after EZ-Backup. Finally, I successfully made a backup and recovered it onto the new drive. I don't know why the cartridges would go bad like that. A large file shouldn't damage the disk, even if the head stayed over the same region of the disk for several minutes writing data (should it?)
It was now nearly three months since I had bought that drive, and already I could see drive prices dropping below what I had paid. I thought of the additional cost of the GSI 2C. Add to that the many hours I had spent trying to get things to work and waiting for long formats and scans and copies to complete. The whole SCSI incident had annoyed me, but this IDE matter was even more irritating. At the University I had assignments to do and exams to study for. The last thing I wanted was to spend several hours every week fighting with the computer.
Windows 95 booted fine, provided I had the "Type 1" drive setting in BIOS (actually, I had to enter the cyl/hd/sec values for a Type 1 drive). I also had to make sure the drive connected to the GSI 2C was not on the same IRQ as an active motherboard IDE interface. Windows 95 Device Manager reported "Drive C: using MS-DOS Compatibility Mode", but I thought that would be simple to fix. Wrong again.
The motherboard IDE controller is a "CMD PCIO 640 PCI EIDE controller chip", or a "CMD PCI-0640 PCI to IDE Controller" according to the Windows 95 Device Manager. Removing its entry from Device Manager didn't help. During boot-up, Windows would always automatically detect and re-install it (even when either or both motherboard IDE interfaces were disabled in BIOS.) It seemed the CMD640 would then conflict with the 2C, and cause MS-DOS Compatibility Mode. The 2C manual states that the 2C's channels will be automatically detected and configured by running the Add New Hardware Wizard in Windows 95. However, I think the CMD0640 detection process stopped that from happening properly. It seemed to me that by attempting to be 'smart' (by automatically detecting and installing hardware), Windows 95 was doing something that stopped itself from working properly. Hooray for progress.
I tried as many combinations of 2C IRQ settings and motherboard BIOS settings I could think of. In BIOS I disabled both the "PCI IDE Interface" and "Standard IDE Interface", then one, then the other, and then neither. I set "X" as channel 2, IRQ 15 and 12. I set "X" and "Y" as IRQ 14 and 15, and connected the drive to "X". All these combinations would either cause freezes (right after turning on the computer or while Windows 95 was loading) or the MS-DOS Compatibility Mode.
(I found a couple Microsoft Knowledge Base articles which I thought might be relevant, but they didn't help much in my situation. Articles Q136106 and Q151911 mention the CMD-0640 controller. Q126356 discusses GSI IDE controllers, but doesn't say much that the GSI 2C instruction manual doesn't already say. Q126855 discusses how DOS and Windows handle large hard disks (useful for understanding where the size limits come from), and Q130179 is on Troubleshooting MS-DOS Compatibility Mode problems in general.)
Normally, my sound card uses IRQs 5 and 10 (unusual that it should use two IRQs), with the MPU 401-compatible portion using IRQ 9. The SCSI card normally uses IRQ 11, and the motherboard PS/2 mouse uses 12. The SCSI card is Plug-And-Play, and Device Manager wouldn't let me set its IRQ. The sound card (a Packard Bell 'Sound144AM' sound card, modem, and proprietary CD-ROM interface in one, detected as an "Aztech Sound Galaxy Washington 16") can be set (via Device Manager) to use several different IRQs. I played around with BIOS's ability to make certain IRQs appear 'used'/'reserved' so Plug-And-Play cards wouldn't try to use those IRQs. I tried channel "X" as channel 2, 3 or 4, with IRQs 11, 12, or 10, and used GSI's 'gsieide.inf' file for channels 3 and 4, but still got problems. I still got "MS-DOS Compatibility Mode" when IRQ 10 and 11 were freed for the 2C. The mouse wasn't detected by Windows when using IRQ 12. Later in my experiments, I'd just get a freeze-up while booting Windows 95 if channels 3 or 4 were provided by the 2C.
One more problem that emerged: Quite often, the SCSI card would choose an IRQ that was already in use by another resource, even though there was a different unused IRQ available. When it tried to use the sound card's IRQs (9 or 10), I could set those IRQs as 'used' in BIOS, and the SCSI card would move elsewhere. However at one time, it decided to use IRQ 3, which is used by COM 2 and cannot be set as 'used' is BIOS. Every time Windows 95 boots, it sends a bit of data to my external modem on COM 2 (probably trying to detect new hardware). Once the SCSI card started trying to use IRQ 3, the computer would reset itself when it tried to send data on COM 2. I had to disconnect my SyQuest drive, pull the SCSI card out of its PCI slot, and remove the SCSI driver from Device Manager. Then, after re-inserting the card and re-installing the driver, it finally decided to move to another IRQ, and Windows stopped resetting itself. I was not too impressed by this IRQ-choosing ability of Plug-And-Play, where I have to try 'tricking' the hardware into choosing appropriate IRQs. Again, hooray for progress.
This story has a (mostly) happy ending. I finally found a set-up that worked. First, I disabled both motherboard IDE interfaces, and set the GSI 2C to use its channels "X" and "Y" on IRQs 14 and 15. (I kept the GSI BIOS at D000-D3FF, as always.) Here's the key that I should have thought of much earlier: attach the drive to channel "Y" and leave "X" not connected to a drive. Even though both motherboard interfaces are set as 'disabled', it seems the interface(s) remain somewhat 'active'. A drive(s) connected to either of these interfaces won't be detected by the computer, but it seems the controller tries to make itself part of the primary IDE interface (channel 1) anyway. (The controller was still detected by Windows 95, even when no drives were connected to it.) By not connecting the 2C's channel "X" to a drive, there's no conflict between the 2C's (still-enabled but not used) "X" connector and the somewhat 'active' motherboard connector(s). Channel "Y" is recognized by the computer as a secondary interface (channel 2), and yet does not conflict with anything the motherboard is trying to do. Setting the "Type 1" drive in BIOS seems to have no effect - things work fine with the "Auto Configured" Drive Type. Diagnostic programs (downloaded from drive manufacturers and run from DOS) refer to my drive as 'master' on the 'secondary' interface, with no drives on the 'primary' interface. Linux agrees - it mounts the drive via /dev/hdc (master on a secondary interface).
In Windows 95, I removed the check-mark from the "Original Configuration (Current)" Device Usage box in the "CMD PCI-0640 PCI to IDE Controller" Properties page of Device Manager. The Device Status box on the same page reports "This device has been disabled in the Hardware. In order to use this device, you must re-enable the hardware. ... (Code 29)". A Yellow "!" circle appears next to the "CMD PCI-0640 PCI to IDE Controller" entry. I have two "Standard IDE/ESDI Hard Disk Controller"s present in Device Manager - one is set to "IRQ 15" and the other is "IRQ 14, DMA 07". (This "DMA 07" controller would always re-install itself when the CMD PCI-0640 was detected and automatically installed.) I manually installed the "IRQ 15" "Standard IDE/ESDI Hard Disk Controller" from the "Add New Hardware" Control Panel.
Once I found the way to make everything work, I made another backup of my original drive and copied it onto the new drive. My old drive had changed slightly (I still had to keep using it after the first backup). I also wanted to be sure there wasn't some weird setting or corruption left over in the new hard drive's old restoration of Windows 95. Fortunately, there were no problems this time. Putting the new correct settings in the new restored Windows' Device manager was simple and worked first time.
As for my old drive, I thought it would be handy for making backups of my bigger drive. I set it as slave on the same channel as the new drive. The GSI 2C has a connector for the computer's front panel LED. With the slave connected, only activity on the slave lit the LED. Maybe that's because I'm not using the CSEL (Cable Select) cable that came with my drive. CSEL is supported by the GSI 2C and eliminates the need to tell drives to be master and slave via jumpers. Master and slave are decided by the position of the drive on the cable (at the end or in the middle). The CSEL cable that came with the 2C was too short to connect both my drives. The official limit to the length of IDE cables is 18 inches, but I need a 23-inch cable to reach both my drives (15 inches from 2C to drive, and 8 inches from that drive to the next. A 36-inch (non-CSEL) cable I bought seemed to work, but I wanted to use as short a cable as I could (remembering my SCSI cable length problems.) The cable originally had 24 inches between one end connector (which connects to the 2C or motherboard) and the middle drive connector. I pried the end connector off and attached it 15 inches from the middle connector (and cut off the unused portion of the cable.) (The connector has two rows of v-shaped notches which cut through the ribbon cable's insulation and touch each conductor inside.) I decided not to keep the slave attached. The new drive is quiet, and seeing the LED is useful. It's not hard to re-connect the cable to the slave. This way, I also don't need to supply power to a normally-unused drive, and the drive won't wear out so soon.
I kept using the 23-inch dual-connector cable for the single master drive, instead of using a single-connector cable. I need to remove the 2C card in order to switch cables, and I'm afraid that it won't take much more connecting and disconnecting before the IDE connectors and the ISA slot contacts start to wear out. There's already a little wear on the ISA card edge contacts from four months of insertion and removal.
One thing I chould mention about the IDE connectors: they are two rows of 20 pins, and the cable has a corresponding two rows of 20 holes. Pin 20 is missing from connectors, and hole 20 is supposed to be plugged so that the cable cannot be connected backwards. None of the cables I've seen had that hole plugged, and I did once connect the cable backwards and turn on the computer. Apparrently, this can damage both the drive and the computer, but I was lucky. I also once mis-aligned the cable by one row, so that the first row of pins went into the second row of holes. (The GSI 2C IDE connectors have a plastic rectangle around the whole connector which makes this mistake impossible, but my motherboard doesn't.) Turning on the computer resulted in a strange noise from the drive, and no display on the monitor. Again, I was lucky and nothing was damaged, but thinking you've just fried your computer is not a fun experience.
After more than four months of tinkering, it was finally time to put the cover mounting screws back into the chassis. Every slot and every bay is full. I can start filling up this drive now. I'm using a single 2.1GB partition for Windows 95. Things look the same as before; C: is larger but all the drive letters are still the same. Messing around with drive letter changes is one thing I don't want to mess with just yet - I've already spent enough time adjusting the computer. In another partition, I'll probably be installing Linux along with Xwindows (Xwindows wouldn't fit onto my Linux SyQuest cartridge.) I should enjoy the free space while I can - drives always fill up so quickly these days....
I might just be opening it up again soon, though. There's a 20-pin "Serial B (Internal Header)" connector on the motherboard which can (theoretically) be used as a second COM port (COM 1), if I can get the right cable. (Packard Bell doesn't sell them any more, and it's a proprietary connector.) Asking Packard Bell for the pinouts and getting one custom-built or building my own may be the only option.
Speaking of proprietary...
I currently use a 2x "MATSHITA CR-563" CD-ROM drive on a "MKEPanasonic Interface" (MKE is "Matsushita-Kotobuki Electronics"), which is part of my sound card. (I'm told this interface was used on many different sound cards, including the Sound Blaster.) My set-up has an interesting implication if I want a faster CD-ROM drive. Almost all CD-ROMS sold today are IDE devices. The GSI 2C (in ideal situations) allows for up to eight IDE devices connected to a computer, yet I can only have two on my computer (master and slave on the 'secondary' interface.) Putting a CD-ROM on the same channel as a hard drive usually slows down the hard drive (unless you have a very fast CD-ROM). On one channel, before a device can respond to commands it must wait for the other device to finish any commands sent to it. Using two different channels is the usual way a hard drive and a CD-ROM are installed - both can be responding at the same time to commands sent to them. Fortunately, I do have that SCSI card, which I can connect a SCSI CD-ROM to. SCSI CD-ROMS are fairly hard to find in most stores (I've only seen them for sale in a "Used Computer Stuff" store) and are probably fairly expensive, too (just like how SCSI hard drives are more expensive.) I think the chances of finding any CD-ROM with the "MKEPanasonic" interface is almost zero, and my chances of finding a fast "MKEPanasonic" CD-ROM are even less. (Even though the MKEPanasonic connector is 40 pins like the IDE connector, the connectors are not compatible. I compared the pinouts - they're very different.)
In the process of looking for information, perhaps the most disturbing information I found was regarding the CMD0640 chip itself. It has five flaws, all of which can cause it to corrupt data. It seems quite famous, actually. (The other 'famous' chip is the PC-Tech RZ-1000 PCI EIDE controller chip, which only has two flaws, both of which are also present in the CMD chip.) One 'flaw' is that is that it only has one set of registers for both channels, so its two channels cannot be used simultaneously. The main reason for having two channels in a computer is so that we can use them simultaneuosly. Other flaws include the occasional changing of a couple of bytes - the type of error that may not be noticed until some time after it occurred. Windows 95 bypasses the bugs, but true multitasking operating systems like Linux, OS/2 and Windows NT are more likely to invoke them. A good source of information on the flaws is at ftp://garbo.uwasa.fi/pc/diskutil/, in the file named eidete20.zip (the "20" may be replaced by a higher number for a newer version of the file). When searching the web for information on these bugs, most of the information referred to OS/2 and Linux.