This is a fun, and I enjoyed the "CRT Dude" video because I resonate with that need to understand :-). One of the things I learned during that era was that there were a lot of computer makers but relatively few factories in China that were making things to assemble them. Because it was simpler to take the sheet metal work that was already designed and being made for brand 'x' and then differentiate on the case molding and electronics, a lot of the mechanical components were "re-used" (the factory can make 1000 or 10,000 with the tooling and the more they make the easier to amortize the tooling costs so the cheaper they can offer them).
I suspect when a company spec'd out a new design and then got all the tooling done, unless they explicit language in their contract about selling stuff made with the same tooling to others, the factories could pitch "we will do the basic case with no NRE[1]" and that was a bargain. As a result a lot of things ended up being "magically" similar in those days.
[1] NRE = Non-Recoverable Engineering which is the cost label for the engineering work to build the jigs and tooling that the factory will use to make the parts you want. Example an injected molded switch cover might cost $10,000 in NRE to make the molds that can produce 10 switch covers each and be used up to a 10,000 times. Then if you make 10,000 switch covers, you have used the mold one 1000 times and used up 10% of its lifetime. Cost of the plastic plus $1,000 (the 1/10th of the cost of the mold) are the real cost of those switch covers.
I suspect a big factor was PSU design. Even large OEMs contracted it out, and there probably reached a point where manufacturers started to offer catalog items in the 200-400 watt range where the primary size constraint was the 80mm exhaust fan. Once that's the constraint, it's going to look a lot like LPX, and you get a reinforcement factor around specific dimensions when vendors say "If we use the exact same screw positions we can compete as a drop-in replacement"
It's interesting that PCs adopted the LPX PSU but not the cases and motherboards. I had always seen LPX described as a bit more proprietary than (baby-)AT designs-- you could fit anyone's mainboard in a generic AT case, but even a HP "LPX" mainboard and riser card might not fit in a Packard Bell case.
In the 1990s none of that stuff was made in mainland China, not even the cases. A great deal was made in Taiwan. Even more was still made in the USA.
Personally I remember AST -- a PC clone vendor who specialized in vertical integration, all in the USA. USA-made accessory cards, fit into USA-made motherboards, fit into USA-made cases, run by USA-made power supplies.
All this dried up and blew away, including the parent company, as both mainland and Taiwan capacity improved and prices fell, but it was not that long ago that you could literally build an entire PC with American components. 25 years, not 50 or 70.
Hmm...makes me wonder if a PA-RISC HP 9000 712 is a better NeXTStation than a NeXTStation is today, in terms of longevity, supportability, performance, etc.
I guess its missing the DSP and fancy printer interface of the NS, maybe the overall sound quality.
When I was an intern at Apple in late 1997, my main machine was a PA RISC workstation running OPENSTEP for Mach (can’t recall which version). It was a lot faster than the PowerPC I had on my desk so I did much of my development on that before testing on Rhapsody.
Back in 1993, a $5k SGI had ~5x as much CPU power as a $10k x86 PC. Not a joke. You could get a 100 MHz R4000PC that would happily drive a display at 1280x1024 for the price of a 33 MHz 486 that would run Windows 3.x at 640x480
They worked very hard to make IRIX a competitive advantage over other UNIX vendors, but it was never a reason to buy SGI over "commodity"
If anything, IRIX was a hindrance to adoption, because UNIX was notoriously RAM-hungry, and the early 1990s had a horrible, horrible price crunch on RAM.
The early 90s were the era of (1) Pentium (2) PCI (3) multiplatform Windows NT. Most vendors switched to desktop/deskside designs based on PC bus and PC components, it would be more cost competitive (volume economics) and could run NT/Alpha, NT/MIPS, NT/PowerPC if the market went that way.
Windows NT was unlikely to be in the minds of many of the OEMs of the early 90s. It was marketed against UNIX, and the UNIX market was losing more and more users to the ever-more-powerful PCs running DOS and (usually) Windows 3, with a Novell setup if you wanted to share files and printers.
It really wasn't until NT 4 that Microsoft started pushing NT Workstation for general office use.
In the end it was GNU/Linux under Intel the one which ate propietary Unix workstations. By late 90's (and even more in early 00's) MESA was getting good
enough to surpass Irix machines.
It makes sense, if one considers them the evolution of the Apple II and the IBM PC.
The IBM had the motherboard on the left, along with the expansion cards, and the drives and power supply on the right. The AT continued this.
Clones wanted mostly compatible cases, and motherboards wanted to be mostly compatible with IBM cases and clone cases.
Then we had Amigas like the Amiga 3000, which had a similar layout but a riser to take horizontal expansion cards.
While some more bespoke PCs had vertical risers, most PC cases in the early to mid '90s were large. It was the machines we paid a bit more for that made being smaller in to something a bit premium.
While taking apart my Amiga 3000 is a bit of work, the design is absolutely wonderful, and more than once I thought about the design of it compared with later machines like the Sun Ultra 5, the Motorola StarMax (PowerPC Mac clone) and others.
The Ultra 5 (desktop) and Ultra 10 (tower) were a cost cutting exercise that put an UltraSPARC IIi (2i) onto what I think was an ATX form factor motherboard. It used ATA drives, USB keyboard and mouse, a VGA port, etc. This was an act of desperation from Sun, not an example of their best engineering.
That said, compared the performance of a $3500 Ultra 10 with 512 MiB of RAM to $10k+ Sun Ultra 30’s and HP C180’s, each with 128 MiB of RAM. These prices were after applying significant edu discounts. The heftier sheet metal, SCSI drives, and nostalgia did not allow these traditional UNIX workstations to touch the performance of the much cheaper Ultra 10 with 4x the RAM.
No, they most emphatically were not. The early 3000 series, like the MIPS based DECstation and VAX based VAXstation, were based on Turbochannel[1], which was DEC's primary bus technology at the time. Later Alphas used PCI as their primary bus. There were Tubochannel to VME adapters, and PCI to VME adapters, but VME was never a primary bus used in alpha.
Before the PC era computers were mostly integrated in the keyboard (many 8-bit home computers), the monitor (Apple Lisa) or both (Commodore PET).
Then they became more or less flat boxes on the desktop. Early PC's were like that (the bulky version) but also early workstations like Sun's SPARCstation (the more elegant version). They were meant to put a monitor atop.
Over time the boxes got bigger and louder and the monitors got bigger which made this design impractical. Some people flipped the boxes, put them under their table and the tower was born. Not long and professionally made tower cases appeared.
Over time the bulky towers got smaller and we had Midi- and Mini-Towers on the PC side and things like the Sun Ultra 24 or the SGI O2 on the workstation side. These could be put on the table again but this time next to the monitor and not below it.
I made a living porting Unix to new hardware in the mid to late 80s so all the busses of the day came across my desk, and designed graphics card silicon in the 90s
S-100 was very much history by the time the boxes in the article were designed (5-10 years before) VME and Multibus were the first generation workstation busses, PCs had ISA->EISA, Macs had NuBus/NuBus90 - all of them converged on PCI once chips with enough pins were packaged cheaply enough (plastic rather than ceramic - 200+ pins)
> Or is the LPX desktop just the crab of personal computer design, and all lineages will converge here eventually. I have no idea.
It's simply a good design for the time. Motherboards needed a bunch of stuff, none of it very tall, expansion busses were busses, so you can put one slot in the middle for a daughter board with slots fot expansion cards. Drives go on the right hand side by tradition, which makes expansion ports on the left.
Depending on how many slots you want, you can make the case taller or shorter.
Let’s not forget Amiga 3000UX - a workstation released with Amiga Unix, a full port of AT&T Unix System V Release 4 (SVR4). Notable users include Free Software Foundation staff programmers who used it at MIT to help further some early development of the GNU operating system.
The Indigo2 family had EISA slots, so it was obliged to hold a full-length ISA card.
There are only so many ways to skin that cat: You want a small, thin desktop with quiet cooling that will also fit a full length ISA card. LPX and Indigo2 look similar because they were solving the same problem.
See also: Octane and Origin 200 vis a vis ATX and WATX. When SGI adopted 64 bit PCI, their desktop-sized systems started to look a bit like ATX, despite a radically different underlying architecture
-----------
The equivalent Sun and HP9000 kit could look quite different because they did not even try to offer EISA as an option
Later Sun and HP9000 kit, with PCI64 slots, started to converge on an ATX "look" for similar reasons to the I2 and LPX
25 comments
[ 2.8 ms ] story [ 66.2 ms ] threadEarly Alphastations were assuredly not VME! They were DEC's own Turbochannel.
Indigo R4xxx was a completely different architecture shared with the later SGI Indy
Both had GIO32 slots
Same case, same slots, totally different architectures inside
I suspect when a company spec'd out a new design and then got all the tooling done, unless they explicit language in their contract about selling stuff made with the same tooling to others, the factories could pitch "we will do the basic case with no NRE[1]" and that was a bargain. As a result a lot of things ended up being "magically" similar in those days.
[1] NRE = Non-Recoverable Engineering which is the cost label for the engineering work to build the jigs and tooling that the factory will use to make the parts you want. Example an injected molded switch cover might cost $10,000 in NRE to make the molds that can produce 10 switch covers each and be used up to a 10,000 times. Then if you make 10,000 switch covers, you have used the mold one 1000 times and used up 10% of its lifetime. Cost of the plastic plus $1,000 (the 1/10th of the cost of the mold) are the real cost of those switch covers.
It's interesting that PCs adopted the LPX PSU but not the cases and motherboards. I had always seen LPX described as a bit more proprietary than (baby-)AT designs-- you could fit anyone's mainboard in a generic AT case, but even a HP "LPX" mainboard and riser card might not fit in a Packard Bell case.
Personally I remember AST -- a PC clone vendor who specialized in vertical integration, all in the USA. USA-made accessory cards, fit into USA-made motherboards, fit into USA-made cases, run by USA-made power supplies.
All this dried up and blew away, including the parent company, as both mainland and Taiwan capacity improved and prices fell, but it was not that long ago that you could literally build an entire PC with American components. 25 years, not 50 or 70.
I guess its missing the DSP and fancy printer interface of the NS, maybe the overall sound quality.
It cost half as much and had considerably more CPU power
There is a reason NeXT decided to become a software company...
If you could install IRIX on junk commodity hardware no one would have a reason to pay SGI $100k for one of theirs.
They worked very hard to make IRIX a competitive advantage over other UNIX vendors, but it was never a reason to buy SGI over "commodity"
If anything, IRIX was a hindrance to adoption, because UNIX was notoriously RAM-hungry, and the early 1990s had a horrible, horrible price crunch on RAM.
It really wasn't until NT 4 that Microsoft started pushing NT Workstation for general office use.
The IBM had the motherboard on the left, along with the expansion cards, and the drives and power supply on the right. The AT continued this.
Clones wanted mostly compatible cases, and motherboards wanted to be mostly compatible with IBM cases and clone cases.
Then we had Amigas like the Amiga 3000, which had a similar layout but a riser to take horizontal expansion cards.
While some more bespoke PCs had vertical risers, most PC cases in the early to mid '90s were large. It was the machines we paid a bit more for that made being smaller in to something a bit premium.
While taking apart my Amiga 3000 is a bit of work, the design is absolutely wonderful, and more than once I thought about the design of it compared with later machines like the Sun Ultra 5, the Motorola StarMax (PowerPC Mac clone) and others.
The Ultra 5 (desktop) and Ultra 10 (tower) were a cost cutting exercise that put an UltraSPARC IIi (2i) onto what I think was an ATX form factor motherboard. It used ATA drives, USB keyboard and mouse, a VGA port, etc. This was an act of desperation from Sun, not an example of their best engineering.
That said, compared the performance of a $3500 Ultra 10 with 512 MiB of RAM to $10k+ Sun Ultra 30’s and HP C180’s, each with 128 MiB of RAM. These prices were after applying significant edu discounts. The heftier sheet metal, SCSI drives, and nostalgia did not allow these traditional UNIX workstations to touch the performance of the much cheaper Ultra 10 with 4x the RAM.
No, they most emphatically were not. The early 3000 series, like the MIPS based DECstation and VAX based VAXstation, were based on Turbochannel[1], which was DEC's primary bus technology at the time. Later Alphas used PCI as their primary bus. There were Tubochannel to VME adapters, and PCI to VME adapters, but VME was never a primary bus used in alpha.
1: https://en.wikipedia.org/wiki/TURBOchannel
Before the PC era computers were mostly integrated in the keyboard (many 8-bit home computers), the monitor (Apple Lisa) or both (Commodore PET).
Then they became more or less flat boxes on the desktop. Early PC's were like that (the bulky version) but also early workstations like Sun's SPARCstation (the more elegant version). They were meant to put a monitor atop.
Over time the boxes got bigger and louder and the monitors got bigger which made this design impractical. Some people flipped the boxes, put them under their table and the tower was born. Not long and professionally made tower cases appeared.
Over time the bulky towers got smaller and we had Midi- and Mini-Towers on the PC side and things like the Sun Ultra 24 or the SGI O2 on the workstation side. These could be put on the table again but this time next to the monitor and not below it.
https://en.wikipedia.org/wiki/Altair_8800
https://en.wikipedia.org/wiki/Vector_Graphic
https://en.wikipedia.org/wiki/North_Star_Horizon
https://en.wikipedia.org/wiki/IMSAI_8080
Thanks for coming to my ted talk
S-100 was very much history by the time the boxes in the article were designed (5-10 years before) VME and Multibus were the first generation workstation busses, PCs had ISA->EISA, Macs had NuBus/NuBus90 - all of them converged on PCI once chips with enough pins were packaged cheaply enough (plastic rather than ceramic - 200+ pins)
It's simply a good design for the time. Motherboards needed a bunch of stuff, none of it very tall, expansion busses were busses, so you can put one slot in the middle for a daughter board with slots fot expansion cards. Drives go on the right hand side by tradition, which makes expansion ports on the left.
Depending on how many slots you want, you can make the case taller or shorter.
https://en.m.wikipedia.org/wiki/Amiga_3000UX
The Indigo2 family had EISA slots, so it was obliged to hold a full-length ISA card.
There are only so many ways to skin that cat: You want a small, thin desktop with quiet cooling that will also fit a full length ISA card. LPX and Indigo2 look similar because they were solving the same problem.
See also: Octane and Origin 200 vis a vis ATX and WATX. When SGI adopted 64 bit PCI, their desktop-sized systems started to look a bit like ATX, despite a radically different underlying architecture
-----------
The equivalent Sun and HP9000 kit could look quite different because they did not even try to offer EISA as an option
Later Sun and HP9000 kit, with PCI64 slots, started to converge on an ATX "look" for similar reasons to the I2 and LPX