03/17/2017 at 00:16 •
The CD-ROM drive I ended up using is one I already had on hand. The floppy drive was obtained on March 8th, 2017.
Arguably one of the first things you should do after hooking up a PC's bare necessities--mainboard, power supply, CPU, and RAM--you should get some kind of disk installed so that you can load and test software to actually make meaningful use of the machine. As the motherboard I chose supports PIIX3 (and therefore has integrated IDE and floppy disk controllers, along with some other niceties), I had all I needed to hook up a disk. I started with the 3.5" floppy drive.
At the time of writing, I've not purchased the 5.25" drive yet as finding one with a black bezel (to match the case I'm mounting this in) is proving to be expensive, and I may just end up using a bit of paint to get the result I want out of a beige drive. That'll get its own log post.
The drive's manufacturing date is in 2004, so it's at the tail end of the floppy drive's relevancy in modern computing, but still old enough to be reasonably good build quality. It took a bit of BIOS tweaking and another floppy drive from a donor machine to verify the cable wasn't malfunctioning, because initially, the motherboard refused to boot from the floppy drive! Nothing catstrophic: its default boot order is counter-intuitively set to hard disk -> cd-rom -> floppy. Once I verified the drive wasn't bad, I reconfigured the BIOS and rebooted, and all was well.
It's a floppy drive. That's really about it, to be honest! Nothing terribly special. I'm sure it'll be more interesting when I get the more antiquated 5.25" drive!
Now for the CD-ROM drive.
Oh boy, the CD-ROM. One of the things I remember from dealing with similarly aged PCs is that the CD-ROM drives they're equipped with are often incapable of reading burned CD-Rs from a modern machine, complicating the matter of sending data to the machine when a network connection is not an option. I wanted to head this off completely, so I started shopping around for a CD-ROM drive that was IDE capable and at least 24-speed, as I've never encountered a drive ≥24x that can't take a CD-R. I also wanted one with a black facade, so that it would match the case as well. While shopping for adapters to pair with the PSU in this build, I stuck my head into the optical drive aisle at Micro Center.
SATA. SATA. SATA. All SATA. It was legitimately just stacks and stacks of SATA drives. Most of them even Blu-Ray capable. A handful of DVD-only. A lot of burn-capable drives. But not a single IDE. We've really come a long way, haven't we?
I knew this meant shopping for one online. I was fine with that, so I ordered what was listed on eBay as a Genuine Dell IDE CD-ROM drive seen in their earlier office-intended OptiPlex models. Having worked on OptiPlexes before for work, this seemed like a straightforward route, as I do recall them having 48x CD-ROM drives in their earlier models.
In the meantime, I actually had an OptiPlex on hand that could donate an IDE CD-ROM drive. I figured I'd put it back when this project was done, and use the one I bought, since it was relatively ($10 or so) cheap to obtain, with free shipping. The donor drive worked well, as did the slim IDE cable that I grabbed for the machine to save space (because I really, truly hate those old ribbon cables and how difficult it was to cable-manage with them).
I waited and waited. A week after purchasing the drive, I contacted the seller, asking where the drive was because I'd sent payment immediately after purchase. 24 hours later, I received a reply that they would be shipping my order within the day, a promise that they did end up keeping.
Another 4 days later, with USPS's tracking being unhelpful and infrequently updated, the drive arrived. I opened it, inspected it, felt satisfied with it... until I turned it over.
I specifically asked for IDE. The listing was for IDE. It took you a week to ship. I paid you within seconds of pressing that magical Buy It Now! button. I guess in eBay vernacular, this seller gets F---------- WOULD NOT BUY AGAIN.
I ended up just sticking to the old OptiPlex drive, which worked like a charm. Thanks, Dell! Still good for something.
It's nice getting things done.
03/16/2017 at 23:48 •
This part was obtained on March 8th, 2017.
In seeking out a power supply, I was more concerned about finding a recently-built power supply with less emphasis on SATA-format plugs and more options in terms of Molex-style connectors, as the majority of components I would be installing in this computer that require a separate power rail utilize Molex/AMP style connectors.
Finding a recent Molex-only supply is difficult in today's part market, but finding one with emphasis on Molex wasn't difficult at all. I opted to use a lower-end Kentec 400W power supply, as it certainly supplied more than enough horsepower to keep this machine running, not to mention it still being used as a field-replacement power supplies in HP and Dell workstations in offices worldwide.
There's not much else to say about this power supply, except for a particular saga of my own impatience. In setting up the machine, I discovered that the supply only contains a single floppy connector. Initially, I didn't find this to be an issue, because even when installing a 5.25" drive (which use standard-size Molex/AMP connectors), there's only one place a floppy connector would be used: the 3.5" drive.
Unfortunately, the CF to IDE adapter uses a floppy connector, so I could only ever have the floppy drive or hard disk connected at any one time.
This was unacceptable, of course, so I stepped out the door to the nearest closing RadioShack to see if they carried a standard Molex to floppy-sized Molex adapter, as they do tend to carry older technology unlike many other electronic stores. They did have well-built Molex splitters and extensions, but unfortunately, not a single floppy connector could be found--not even in their connectors drawer.
I chalked this up to RadioShack having supply problems all the time due to routine bankruptcy, and tried Micro Center instead.
Nope, still nothing. Plenty of Molex to SATA adapters, though--and you know what they say--Molex to SATA, lose all your data. I was fairly discouraged and thought I'd have to stop working mid-test-build because I couldn't run a floppy drive and the CF card at the same time. That is, until I noticed this oddity hanging out behind the power cable extensions (product photo, not my own):
It was two male Molex plugs and a daisy-chained floppy connector, powered by a plastic-molded female SATA power connector. This obviously wasn't ideal, but it was the first floppy connector I'd seen in person since opening my power supply box. I wasn't keen on butchering another computer's power supply in the house just to retrieve the floppy connector from it, but this? This was a cheap adapter, and I felt adventurous that day. I bought the adapter and an additional Molex extender, and went to work with the pliers and soldering iron at home.
For those unaware, the floppy connector is literally just a minified version of the standard plug. Nothing special. No voltage or current difference. 12V/22A for the yellow wire, 5V/18A for the red wire on both ends. Super-simple stuff.
I didn't take photos as I didn't have an extra hand to operate the camera, but I performed a Western Union-style splice between the female end of that extra Molex extender and the floppy connector, which amounts to just hooking the two exposed ends of each wire together and then twisting them against themselves to tighten the bond, as seen in the picture from Appropedia below:
I reinforced each bond with some lead-free solder, sealed the exposed metal with electrical tape, and then taped over the bodge with more electrical tape across the whole assembly. The end result:
No excess heat, no fires, none of that. If it works and it's stupid, it's not stupid, right?
03/16/2017 at 05:31 •
The CF Card was obtained on March 8th, 2017, and the CF to IDE adapter was obtained on March 9th, 2017.
Just a warning: this is a longer log post compared to the others.
IDE is one of those legacy technologies that you see everywhere in spite of its age. ATMs, embedded industrial systems, and old office PC installations still utilize IDE hard disks and IDE/ATAPI optical drives at the time of writing. The technology was alive for the better part of two decades, and was a significant improvement over the SCSI drives that predate it, even if it has been largely surpassed by SATA in nearly all configurations.
So why is it this hard to find a good, low-capacity IDE hard disk?
It proved much easier to just use a slightly less antiquated storage format that just happened to be natively compatible with IDE, with a significantly lower failure rate.
During the week leading up to me receiving these parts, I worked with an older Windows 98SE machine I had lying around, installing a new network controller, sound card, and graphics card in it, and configuring it to dualboot Win9x and Win2k in order to give it maximal compatibility with both older and newer systems, as a sort of middleman when sending data between the Reverse-Sleeper and my studio or gaming PC (in addition to just making a great Windows 98 gaming PC, which incidentally, would have been just fine playing DOS games, but where's the fun in that?)
Computers like this are called tweeners, or in-betweeners, because they are just old enough to communicate with antiquated hardware but just new enough that they won't be completely SOL when speaking with the latest and greatest machines. I'll write a post about the tweener separately after this is complete, as it deserves its own post.
(Note that the use of Ethernet, if possible with the machine in question, can totally remove the need for tweeners; however this isn't always an option.)
Now, the reason the tweener PC is relevant to this post is because it, too, uses IDE for its drives, and unfortunately, of the 4-5 IDE hard disks I had laying around the apartment, two of them failed catastrophically before I could use them to any effect. One probably had a logic board failure and was simply not recognized by the IDE controller anymore, and the other suffered from a lovely case of click-of-death.
Neither of them were particularly old as far as IDE drives go (1999 and 2004, respectively). When it came time to figure out what to do about the Reverse-Sleeper's storage, the double-failure was fresh on my mind and got me to thinking, what am I going to do for storage on this machine?
DOS 6.22 uses FAT16 as its base file system. FAT16 only supports partitions of up to 2GB apiece, so obtaining a 40GB hard drive of some kind on the cheap would not only be ill-advised due to the problems with cheap hard drives, but also a colossal waste of space, even if I max out the MBR partition limit of 4 partitions (8GB).
Any 2GB (or other small denomination) IDE hard drives still in existence today aren't even close to being new or robust compared to their original build quality, and are certainly due for failure if they haven't failed already; not to mention the price of finding such a hard drive is prohibitively expensive due to collectors just like me compared to other storage options. And what if it got damaged in transit? It's just a glass or platter with a metal substrate on top. Early hard drives weren't exactly shock resistant. I also really, really didn't want to use a SATA to IDE adapter board because it would be the kind of technology you could see inside a modern gaming PC, right here, right now, defeating the purpose of sticking to older components. I was willing to sacrifice old for interesting, however, and that's what influenced my decision on top of sheer difficulty of finding a suitable drive.
As it turns out, Compact Flash cards fill this niche just fine, and while they aren't exactly retro by any stretch, they do their job surprisingly well in the photography field where fast mass storage is needed.
CompactFlash is a NAND flash (originally NOR flash) storage card that, in terms of components, is not too dissimilar from a modern solid-state drive. Speed-wise, it doesn't have the same advantage that a 6Gbps SSD has, but for the purposes of this project, it did have one advantage: CompactFlash is natively compatible with the IDE standard. Someone could totally build a totally passive IDE to CF adapter with a couple of jumpers for drive modes (master/slave), and a power connector, and then all you'd have to do i--
Oh. Hey. That's neat. And $8 with free shipping.
With a bit of f'nagling of the power connection (notice how it uses a 4-pin floppy-sized Molex power connector), the drive worked perfectly. More on that later in the next update, which should cover the power supply and all that went into that.
I did have to use a bootdisk to format and FDISK partition the drive as it arrived preformatted in FAT32, which DOS 6 was not terribly happy about. One of these days I'll maybe find a patch or driver that lets me use FAT32, but until then, this will suffice.
03/16/2017 at 04:41 •
This part was obtained on March 9, 2017.
The RAM I have is an interesting case. On one-hand, I paid for 16MB of RAM. I verified the motherboard supported this configuration before purchasing. Just one issue:
They sent me 256 MB EDO SIMMs. Rather, they sent me two 128MB sticks. Fortunately it was 60ns in its timings, which was compatible with my board. But 256 megabytes? Far more than I was expecting, and certainly enough to make this a cutting-edge information-age powerhouse (in 1994). The biggest problem, though: The motherboard doesn't support any more than 128MB of RAM, per the manual. RAM must also be installed in pairs--a single stick cannot be used.
Bit of a pickle, if I do say so myself.
What's a man to do in a situation like this? I didn't want to spend more money on additional RAM, but surely I could find a way to make this work. I ended up rolling with this configuration regardless, and interestingly enough: the computer boots. However, it reports that it has 65536 KB of RAM (in binary, 1111 1111 1111 1111 KB plus one).
This didn't seem to be much of an issue, though I was wary of the potential side effects if a poorly coded program tried to poll for the amount of RAM available, underflowed to 0 KB available, and then asserted that "you have no expanded memory!" This did end up happening with Wing Commander in later tests (preventing me from configuring it to work with the Sound Blaster)--although Wing Commander also suffered from being unplayable due to it expecting the fixed 4.77 MHz of the 8088 that would have been common on most early PCs of the day.
Oh well. It's the seller's problem now, because the computer works and I'm fine with it this way. There exists a handful of TSR programs that can be used in DOS to fix RAM or clock speed issues with older programs.
A small bit of history and trivia for you. This part isn't relevant to the project, but if you like classic computing like this, particularly in the IBM PC world, you may find this interesting.
This RAM specimen is EDO, or Exptended Data Out. EDO is typically optimized for a 66MHz bus speed, which the Pentium MMX--this machine's CPU--is clocked at on the bus, making it a perfect pair. EDO was a breakthrough in that it allowed other address reads to take place while the previously queued read was still in progress, allowing faster CPUs to utilize less latency (and thus incur significantly less slowdown when the CPU's cache fails).
The first chipset to optimize for EDO was the Intel 430TX chipset, which also happened to support PIIX3 (PCI, ISA, USB, IDE, and a bunch of other nice features)... which the motherboard used just happens to be one of the first to support.
It's truly a match made in heaven, and totally by accident, because I didn't learn any of this before building this machine.
03/15/2017 at 06:36 •
This part was obtained on March 10th, 2017.
The CPU used in this project is the Intel Pentium w/MMX Technology, clocked at 200MHz internally with a 66MHz bus speed. It uses an eclectic 2.8V input voltage, which was strikingly high for the era. It's the most capable CPU for gaming by far for this build. A 233MHz variant exists, but the motherboard does not support it as a possible frequency/multiplier/voltage jumper setting. Despite this, 200MHz is plenty of horsepower, to the point that even most Windows 95-era games (which will not be part of this project) didn't require it.
The reason I selected this CPU is a product of what it competed with at the time combined with cost efficacy of finding parts.
In terms of what the competition had, the AMD K5 only boasted clock speeds up to 133MHz, not to mention it being riddled with design issues. Even the later K6, which included MMX instructions as well as the extra 3DNow! instructions, came in time to compete with the Pentium II, which already had it beat in the clock speed and onboard cache department (despite it running hotter than hell). The other competitor was the Cyrix MII, which also boasted clock-speeds up to 233MHz, but had its own share of design issues, not to mention arriving late to market and being forced to compete with the K6 and Pentium II, and later the Celeron to squash the low-cost market. Ultimately, these factors led to the MII being the last Intel 80x86-compatible processor produced by Cyrix.
The other possibility to use would have been a Pentium Pro, but this would require using the oddly narrow Socket 8 form factor (requiring a different motherboard). The Pro suffered from only having a measly 5% increase in power from its predecessor on 16-bit operations. Considering most applications at the time were written for 16-bit operating systems, including those for Windows 95, the Pro was just simply not cost-effective for the era. It did introduce an extra 256KB of L2 cache directly on the die, but due to this being untestable until the CPU die assembly was complete, the cost was driven up further due to manufacturing waste every time a die failed to pass QA.
The Pro also only went up to 200MHz, just like the MMX I ended up selecting, and with only 256K onboard cache (compared to the 512K found on the motherboard I selected), it's really no question as to why the Pro isn't the right choice. 512K and 1024K L2 cache variants exist, but again, price to advantage ratio is not in our favor here.
Finally, combining all of these factors and considering the higher cost and difficulty of finding a socket 8 board, it's difficult to recommend a Pentium Pro for this build.
In summary, the Pro was little more than a brief segue between the MMX and the Pentium II, compounding my decision to stick to the 200MHz Pentium MMX.
03/15/2017 at 06:05 •
This part was obtained on March 8th, 2017, for $40.
The motherboard for this computer is the Freetech P5F72 ATX board. The P5F72 uses an Intel 82430VX chipset, is one of the earliest motherboards capable of Plug 'n Play, and contains an equal number of PCI and ISA slots, giving me ample wiggle-room to figure out what expansion cards I needed.
My chief reasoning for selecting this board was the form factor. While a PC/AT board would be more akin to 'the true experience' in terms of what would have been common back in the day, the AT is immensely difficult to find cases for at auction (online especially), particularly when pricing is an issue. Apart from that, one of the key tenets of this project is to create a reverse-sleeper: instead of putting a modern gaming rig in an antiquated case, it's an antiquated gaming rig in a modern case. You won't find AT cases with all the bells and whistles like today's cases--they just don't exist.
Compounded by issues with finding a working, reliable AT power supply with the build quality that would have existed in the early 90's, the best decision was certainly to spring for an ATX board. Knowing that there were ATX boards with sockets compatible with the later Pentium processors, this was the best route to take for cost-efficacy. But the advantages don't stop there.
Additionally, you can't find PCI slots on any PC/AT motherboards: the PCI standard didn't exist then, meaning the Voodoo2 is out. No Quake for you.
ATX it is. Unless you don't care about 3D, and you have deeper pockets, in which case: Okaaaaaaay. But don't say I didn't warn you.
With this in mind, the P5F72 also sports the following features:
- Integraded LPT, COM1, and COM2 ports
- Built-in controller for IDE fixed disks, ATAPI CD-ROM, and floppy drives
- Plug 'n Play compatibility so you don't have to dance with IRQ jumpers for hours and hours
- 512KB L2 Cache, from a time when L2 cache was something you inserted into an IC socket on the board
- Able to supply the strange voltages required by the MMX-capable Pentiums, unlike some Socket 7 boards
- Automatic disk size detection, so you don't have to configure the number of cylinders/sectors on the disk.
For someone such as myself who grew up in the era where these things were an expected standard, and such things would be unthinkable to not include on a modern motherboard, the P5F72 is an absolute godsend.
This particular specimen was not without its problems, however.
Circuit boards are made of fiberglass, coated in copper on both sides (sometimes with layers of copper/fiberglass sandwiched together) and often covered in a silkscreen or epoxy afterwards to keep it clean. This is a durable, reliable, and time-tested setup; but if you introduce its natural predators of heat and stress, fiberglass will bend over faster than a cougar looking for college-aged men on St. Patrick's Day in Savannah.
I think this motherboard might have had a bit of fun a decade ago.
Oh boy. That's not good.
The eBay listing did say that there was "a bit of warping near the parallel port," and upon later review, I did see that the seller noted this bend in extensive photos that I had missed due to the photos being obscured by the 'next' button on the eBay interface. Okay, that's fine. The board is marked as tested and working, so I can get a refund if they lied about that at least.
Still, it was a cause for concern. I cautiously mounted the thing on top of a ceramic bowl to keep it off the granite of my counter-workspace just in case any stray residue I missed in my cleanup were conductive enough to short something in a weird place because of the warping.
ATX boards are typically capable of about a half-inch (roughly 1.3cm) of stress at max before something breaks, in my experience. I optimistically stuck with this board because the additional features and ease of use were too nice to ignore over what might just be a cosmetic lapse.
Ultimately, the warping ended up being a non-issue. The board booted, and with a bit of jumper tweaking (definitely too frustrating for photos, for someone who's never built a computer that needed jumpers to set the voltage and clock speed of the CPU), everything was dandy.