A Quick Tour of the HP-9000 712/100 NeXTSTEP Workstation (2016) | Hacker News Reader

A Quick Tour of the HP-9000 712/100 NeXTSTEP Workstation (2016)

bytecellar.com

69 pointskristianp55 comments

Jobs' decision to ship NeXT on 68k at the very moment when the architecture fell behind basically crippled the platform. Superscalar pipelined architectures were taking off, and where Intel managed to catch up with the Pentium and pull ahead with the P6 and derivatives, Motorla just never got it done. So while the NeXT was adequate (if somewhat slow) in 1988 by 1993 (when Intel released the Pentium, meaning that the whole industry was now expecting >1 IPS performance) it was kind of a joke.

While SPARC was proprietary (and PA and POWER were still in the future) MIPS was actually shipping chips at the time for general sale. Why NeXT didn't go with them I don't know.

Probably because Jobs knew M68K already and had relationships with Motorola? Just a theory.

NeXT did have a working M88K[1] system, that booted and ran. The case was a slightly taller version of the NeXTStation. Then they killed all hardware.

However the pa-risc systems were the best you could run NS/OS on. Intel might be faster but it was a pain to setup. 712/100 just ran and ran smoothly.

[1] http://www.nextcomputers.org/forums/index.php?topic=1271.msg...

http://www.nextcomputers.org/NeXTfiles/Images/Rare_NeXT_Hard...

EDIT: Changed powerpc to m88K.

The Motorola relationship is how Jobs got a decent price on the 68030/56001 bundle. It was intended as a sort of mutual offset against the coming RISC processors known to be the competition.

NeXT was working on a Motorola 88110 (later revised to Motorola 601) RISC processors., but NeXT abandoned the hardware business in 1993 before they went to market.

https://en.wikipedia.org/wiki/NeXT_RISC_Workstation?wprov=sf...

A shame the mc88120 never made it to market. It was designed as a 2-3 way superscalar out-of-order processor with a lot of impressive ideas (considering the time).

NeXT was out of the hardware market about a month before P5 shipped, and they were already in the process of trying to move to PPC. The CPU was just about the least interesting chip in NeXT hardware though.

MIPS would have put them in a completely different price bracket. The MIPS-based DECStation 3100 was $11,900[1] compared to the original $6,500 NeXT computer.

[1] https://books.google.com/books?id=OzoEAAAAMBAJ see page 44

That $6500 was without any disk beyond the unusably slow (as main storage) NeXT Canon optical drive. I suppose the DEC price was diskless but that was also full retail. Just because NeXT refused to discount didn't mean DEC (or Sun) were limited by that.

I'm not sure a different architecture would have mattered much. NeXT was trying to launch a Unix workstation right at the twilight of the age of Unix workstations. They got clobbered using 68K, but then at around the same time Sun got clobbered using SPARC, DEC and Compaq got clobbered using Alpha, etc.

Intel and Microsoft were busy eating the world, and one of the things that got gobbled up was the market segment NeXT was hoping to sell into.

As Jobs mentioned, "We are going to be the last company that gets into the workstation market, or the first to fail". (He probably worded it better.)

I always wonder what personal computing might have been like had DEC not so greatly feared cannibalizing their mini/mainframe business. By the time they realized their mistake, it was already way too late. If Digital had played its cards differently, x86 might have been just a footnote.

I found it very ironic that the biggest market for PDP-11-compatible micros was one not actually open to DEC: the Soviet Union.

Soviet Union even made pocket computers that were electronically (not just software!) compatible with PDP-11 peripherals, if you dared to open the case and solder a connection.

There are also rumours that around 1980, there was a design at Digital that could have led to a PDP-10 workstation, but Digital wanted to kill PDP-10.

It wasn't a rumor, it was the KT20 "Minnow". It was a workstation sans display. Given the memory addressing limits there were reasons why it was stillborn.

I assume Minnow didn't use Model B addressing, then? My understanding was that Minnow was important step towards miniaturization of the PDP-10 line, in tandem with Jupiter (IIRC?) which spearheaded high cpu speed initiative...

Apparently I’m wrong about the address limits but it’s running on a really narrow microarchitecture with very limited (probably deliberately) performance.

Another massive market for PDP-11's was McDonald's. They ran the cash registers in a lot of the restaurants.

I knew a guy who earned a living rehabbing PDPs that had done service in McDonald's. I guess they weren't always located in the best place in the building because said he spent most of his time cleaning the dried soda off of circuit boards.

A place I worked replaced its Suns with Dells. I couldn't believe they would do such a thing, since each Sun had to be replaced with three Dells to do the equivalent job.

When I asked about it, the answer was simple: Three Dells were cheaper than one Sun.

I wouldn't call another 5+ years of growth "right at the twilight". NeXT launched a 68030 box near simultaneously against the SS1 & DS3100. Both had faster follow-ons, goosed by the then-amazing HP PA machines (720/730). They (Sun, DEC, HP) raced each other til the Intel machines caught up. NeXT was never a player, despite Jobs attempt to define himself as the winner in the self-defined "academic workstation" segment.

> Jobs' decision to ship NeXT on 68k at the very moment when the architecture fell behind basically crippled the platform.

There were a bunch of dropped balls there. To rattle off a few:

* They targeted academic institutions first, which very much constrained their market, and didn't align well with the $6,500 base price. Which turned into closer to $10K once you bought a disk and printer. (Slightly more than $20K in 2020 dollars.)

* Disk performance on a base Cube was terrible, thanks to the reliance on an Magento-Optical disk as primary storage. (Seek times x3-4 of magnetic disks.)

* Their hardware was very good, but it's expensive to be in the hardware business and others had the resources to vastly outcompete them.

* Choice of 68K didn't help. NeXT got two chips (030 and 040) before they started having to look elsewhere. (There was an 060 that even got used in some accelerator boards, but the market was in a wholesale shift over to RISC by that time. It also can't have helped that Motorola's own RISC follow on to the 68K, the 88K, got almost immediately replaced by their PPC venture with IBM and Apple.)

And as somebody who wrote NeXT software for a number of years, they never really aimed to solve anybody's problem particularly well.

If I had to sum up the NeXT, I'd call it "a much better Mac". Which isn't surprising given that Jobs started right after getting forced out of Apple, and took a number of Mac folks with him. But as you say, $6,500 was way too much for the Mac's strongest market, education; that's like $13k in 2020 dollars.

Instead it was priced like a workstation, but it also wasn't great for that, in that it was a totally new OS and didn't have much software available out of the gate. I've heard of multiple schools that ended up with NeXT labs where almost nobody used them.

The only niche where I really saw adoption was with financial traders, who had money to burn and liked the rapid development speed you could get compared with other platforms of the day. But that was never a goal, just a happy accident.

So in retrospect, I feel the whole thing was not about helping any specific customer, but about Steve Jobs and his feelings about getting booted from Apple.

> If I had to sum up the NeXT, I'd call it "a much better Mac".

Which is pretty much how it wound up, eventually.

> I've heard of multiple schools that ended up with NeXT labs where almost nobody used them.

My university was one of these. We had a huge lab filled with shiny NeXT cubes... which, for reasons unknown, was owned by the communications department.

Just try to imagine the befuddlement of a bunch of early ‘90s journalism majors who have just been seated in front of Unix workstations. No matter what you imagine, I guarantee the reality was worse.

Didn’t NeXt get sued by Apple and had to abandon the lower end market where they would have competed against them?

The 1990 '040 was pipelined, and the superscalar RISCs came half a decade after it, many years after the x86 NeXT port was promoted, no? And the superscalar '060 came out in 1994 according to WP.

But yep the contemporary RISCs were faster than the 68k's nevermind those microarchitectural features.

Ironically, NeXT, or rather the external perception of NeXT was a big driver/motivator of Sun's SparcStation (the classic, wonderful SS1) development. I was then working with Sun people who told me that even before the NeXT announcement.

My employers were signed up as NeXT and Sun developers and for what we were doing, the former was such a disappointment even if was pleasant to use (other than as an NFS server but that's another story...).

Three great things came out of NeXT: Doom, the WWW, and modern Mac OS

(Mac OS X/OS X/macOS. I normally let these slide, but here it’s important which one you’re talking about.)

Anyone know to what extent was the original Quake was developed on Next as well? I recall the BSP map processing taking place on a networked group of HP Next boxes (like the OP), but think the tooling that was eventually released was all DOS-based.

The map tools ran on NeXT:

http://www.gamers.org/dEngine/quake/QuakeEd/quake_ed.html

Modern Mac OS => iOS + iPhone, and with that addition to your list you’re looking at literally trillions of dollars annually that owe their markets to innovations that grew from NeXT.

I’m sure some of them would have happened anyway in some other form, but that’s still pretty amazing.

article explaining the 'Color Recovery' DSP used https://www.hpl.hp.com/hpjournal/95apr/apr95a6.pdf

3DFX used similar trick in Voodoo 3D accelerators, 24bit graphics dithered into 16bit framebuffer producing what they called "22-bit" image quality using 2x2/4x4 ordered mask

https://www.vogons.org/viewtopic.php?t=36548

It was very close to Nvidia 32bit color quality while sustaining/beating Nvidia 16bit speeds, until Nvidia released GeForce card and killed competition.

Based on your link (details at the bottom) it's not the same thing at all as far as I can tell. The Color Recovery buffers a scanline and does a simple edge aware filtering on a 2x16 block, for every pixel. The Voodoo 2 doesn't appear to have a buffer. Also, the HP solution takes 8 to ~ 22 bit, a harder problem.

Similar. Voodoo does color recovery on chip, not in external DSP/RAMDAC, you dont need a line buffer then you have whole framebuffer in ram.

Definitely interesting. The article references an "HP Color Recovery Technology, HP publication Number 5962-9835E" but google can't seem to find it. I'd be interested in what the "minor adjustments required to optimize image quality" is glossing over.

As someone pointed out in the comments, Jobs seemed to have a fascination with HP. Can anybody expound on why he had this fascination with HP and their hardware?

HP's legacy runs deep in the valley, it's not unique to Jobs at all.

It's shocking how far HP has fallen from its illustrious past.

Woz worked for HP, it was his dream job as an engineer. It was a big deal for him to give up his dream job to start Apple with Jobs.

And Woz sold his HP65 calculator to get the money to start Apple. Jobs sold his VW Bug.

VW but not a bug.

Crappy, ad-laden article from CNBC, but that's as hard as I care to look for a URL to back my point. The headline alone should be sufficient. https://www.cnbc.com/2018/07/25/how-steve-jobs-cold-called-h...

Or are you asking why he'd even have wanted an internship there in the first place?

He got a summer job at HP at age 12 by looking Bill Hewlett up in the phone book and cold calling him: https://nextshark.com/steve-jobs-12-year-old-summer-hp/

It goes back to where he was young and talked with Mr. Hewlett at length about a project. There is an interview of Jobs describing this scene.

http://www.stevejobsthebiography.com/extras.html

Good question. AFAIK they had some really high quality products back in the day and also lots of skilled engineers.

HP’s oscilloscopes and other electronic lab equipment (at the time their exclusive focus) were beautifully designed and highest quality. Perfectionists would naturally be drawn to them. Ditto for their first calculators; they were drooled over. Printers came later; all the electronics got jettisoned into Agilent in 1999.

...and Agilent spun them out into Keysight, to focus on slightly random bits of lab equipment instead. What was a long and illustrious history is now...not. I much prefer my Rode & Schwarz machines.

Incidentally, Agilent did the same thing with Varian -- an NMR company that was involved in the acquisition of data ultimately in several nobel prizes. They didn't make enough money in a two year period, and Excel and MBAs killed a multi-decade business [1], leading to unemployment and a monopoly in the field of preclinical MRI, EPR/ESR, and a duopoly in NMR spectroscopy. Ain't progress grant?

[1] https://www.sciencedirect.com/science/article/pii/S109078071...

https://sci-hub.se/https://doi.org/10.1016/j.jmr.2014.12.001

HP used to be an amazing company up until the arrival of Ms. Fiorina in the late 90s. Their hardware used to be innovative and bullet-proof (HP calculators, early Laserjet printers, oscilloscopes, etc.) Even today, when the company is a shadow of a shadow of its former glory, I prefer buying their business class laptops for personal use (e.g. HP Elite x360, HP X2 Chromebook) since they have an innovative combination of unique features, user upgradability, and reliability. In the old days, HP and Intel used to be synonymous with 'Silicon Valley' as I understand it.

HP was the top-shelf electrical engineering instrument company. Every top-tier lab would have HP equipment. Aspiring computer-makers wouldn't be able to afford it and would lust after it from a distance. This probably includes young Steve Jobs.

HP built plotters and printers because their customers wanted hardcopies of what they saw on their instrument screens. They built computers to control the instruments and calculators to assist engineers with calculations. Then, for some reason, the computers and printers started selling even better than the instruments. HP, which still thought of themselves as an instrumentation company, didn't catch on quickly or aggressively enough to capitalize, so their computers and printers did not keep pace and became... other than undisputably first tier. Eventually they span off the first-tier instrumentation company as Agilent (bio + EE) which in turn eventually span off Keysight (EE).

Electrical Engineers are still salty that the mediocre printer company got to keep the HP name while the best-in-class instrumentation company that still commands respect to this very day had to rebrand as "Keysight."

When I was in college, studying EE, we had HP oscilloscopes and digital logic analyzers. I can't remember for sure, but I think it was the logic analyzers that had an easter egg that if you held down a specific set of buttons while powering on, you could play Tetris on it. I also remember the oscilloscope being so sensitive, that without any leads connected, it could detect the signal of someone walking by it.

Yep! Their legacy continues today. Here's HP (sorry, I mean Agilent, sorry, I mean Keysight) dropping a 110GHz real-time oscilloscope:

https://youtu.be/dB5rg_TqmD8?t=32

I'd like to think that if Bill and Dave (Hewlett and Packard) were still alive they could still take pride in their T&M legacy, even though it's the DRM-hobbled one-quarter-full printer ink cartridges that wound up with their names on the front.

Don't know about the logic analyzers, but I played tetris on an HP scope in a lab on campus years ago.

1. They were the original Silicon Valley startup.

2. They were very innovative in a lot of the areas they touched. Instrumentation, which was their core business, but also printing, plotting, portable computers (I own one of their mid-80s HP-UX portable computers), touchscreens (the HP-150, for example).

3. Quality was hugely important. That 35 year old workstation still powers up. I have a similar vintage plotter that still works. HP laser printers were an absolute gold standard. Everything they made was the absolute highest build quality.

4. They had an rmarkable culture. For example, Hewlett and Packard mandated that managers would sit in an open area, while their engineers and scientists would have offices. This was based on what now gets called servant leadership - the idea that the manager needed to be available to facilitate the staff, not the other way around. Employees could freely borrow equipment for personal projects.

Sadly, all of that was flushed away starting in the 90s.

HP also was a leader in device ergonomics. Their devices were pretty and easy to use, as well as reliable.

I don't think anyone had offices. It was all cubicles. (I worked at HP Labs in the 80's and met both Bill and Dave a couple of times.)

One cool thing HP had was "lab stock" cabinets on every floor that were full of free-for-experimentation ICs and other components. I think it's likely that the Apple I and Apple II computers were designed using components taken from those lab stock cabinets. (The story is that Steve W. tried to get HP interested in producing the Apple computers, but they just wouldn't go for it.)

I was just reading about DOOM's history, and it was developed on a NEXT machine which id software used at the time. A free book on the subject which I highly recommend:

https://fabiensanglard.net/gebbdoom/index.html

Reply on news.ycombinator.com