8 ms·
A Quick Tour of the HP-9000 712/100 NeXTSTEP Workstation (2016)
- ajross 7y agoJobs' 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.
- gok 7y agoNeXT 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.
- spitfire 7y agoProbably 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.msg7723#msg7723 http://www.nextcomputers.org/forums/index.php?topic=1271.msg... http://www.nextcomputers.org/NeXTfiles/Images/Rare_NeXT_Hardware/NRW-NeXT_RISC_Workstation/ http://www.nextcomputers.org/NeXTfiles/Images/Rare_NeXT_Hard... EDIT: Changed powerpc to m88K.
- pinewurst 7y agoThe 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.
- qubex 7y agoNeXT 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=sfti1 https://en.wikipedia.org/wiki/NeXT_RISC_Workstation?wprov=sf...
- FullyFunctional 7y agoA 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).
- opencl 7y agoMIPS 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 https://books.google.com/books?id=OzoEAAAAMBAJ see page 44
- pinewurst 7y agoThat $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.
- smacktoward 7y agoI'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.
- trasz 7y agoAs 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.)
- flyinghamster 7y agoI 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.
- p_l 7y agoSoviet 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.
- pinewurst 7y agoIt 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.
- 7y ago
- mschaef 7y ago> 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.)
- wpietri 7y agoAnd 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.
- mschaef 7y ago> 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.
- fulafel 7y agoThe 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.
- pinewurst 7y agoIronically, 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...).
- bluedino 7y agoThree great things came out of NeXT: Doom, the WWW, and modern Mac OS
- saagarjha 7y ago(Mac OS X/OS X/macOS. I normally let these slide, but here it’s important which one you’re talking about.)
- trilinearnz 7y agoAnyone 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.
- bluedino 7y agoThe map tools ran on NeXT: http://www.gamers.org/dEngine/quake/QuakeEd/quake_ed.html http://www.gamers.org/dEngine/quake/QuakeEd/quake_ed.html
- twoodfin 7y agoModern 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.
- at-fates-hands 7y agoAs 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?
- pengaru 7y agoHP'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.
- mikestew 7y agoCrappy, 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-his-way-to-an-internship-at-hewlett-packard.html 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?
- smacktoward 7y agoHe 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/ https://nextshark.com/steve-jobs-12-year-old-summer-hp/
- mzs 7y agoarticle explaining the 'Color Recovery' DSP used https://www.hpl.hp.com/hpjournal/95apr/apr95a6.pdf https://www.hpl.hp.com/hpjournal/95apr/apr95a6.pdf
- rasz 7y ago3DFX 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 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.
- FullyFunctional 7y agoBased 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.
- rasz 7y agoSimilar. Voodoo does color recovery on chip, not in external DSP/RAMDAC, you dont need a line buffer then you have whole framebuffer in ram.
- FullyFunctional 7y agoDefinitely 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.
- 0-_-0 7y agoI 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 https://fabiensanglard.net/gebbdoom/index.html