6 ms·
Unless you were doing 3D graphics, SGI computers of the 1990s were overpriced. At my first startup in 1997 we benchmarked SGI and Suns against PCs of the time,
by rwmj 11y ago
Unless you were doing 3D graphics, SGI computers of the 1990s were overpriced. At my first startup in 1997 we benchmarked SGI and Suns against PCs of the time, and found white-box PCs offered a clear 3 x price/performance advantage. So we kitted out our test network and datacenter with racks of white-box PCs (literally desktop PCs, not even server hardware).
I still have no idea what the advantage of Sun hardware was supposed to be, except some people were locked in, either by software or some kind of Stockholm syndrome.
Edit: Photo of our test network: http://oirase.annexia.org/tmp/testbed.jpg http://oirase.annexia.org/tmp/testbed.jpg
- agumonkey 11y agoReminds me of the Softimage demo in ~96 (after Microsoft bought them) where they compared SGI and vanilla PC box. The guy explained that if the PC being 1/10th the cost of SGI, even if it was half as fast it would still be worth it. Then, to the great satisfaction of Bill Gates, near him on stage, he said that the PC was actually faster than his SGI workstation.
- hga 11y agoWe've been told that the 1988 Sun 386i was quite successful due to its ability to run MS-DOS, and that the Sun 486i was canceled in 1990 because it was faster than the in-house SPARC equivalents. But it took some time for Intel machines to be competitive "big iron", say starting in the 486 days? Including running a serious UNIX(TM) or Linux. ADDED: and serious I/O, we needed EISA and then PCI and its descendants. So by 1997 Sun and SGI had first mover advantages, but neither being well run companies they squandered them even before counting Intel's general excellence in microarchitectures and fabrication. (Plus wasn't SPARC hard to make fast due to an architectural mistake?) One other problem SGI had, their IRIX was a notorious nightmare, which is part of their not being a well run company. Sun also lost more than a little momentum when they switched from BSD to ATT UNIX(TM), and did brain dead marketing where they boasted their UNIX(TM) was less buggy than Windows NT at a time when that simply wasn't true (and if I recall that was at a time they were too often refusing to fix some bugs). A thesis I've just developed here is that the switch to commodity Intel hardware decoupled hardware and software, they helped you avoid getting locked into a single vendor's mistakes (and AMD helped when Intel faltered in several ways). And open source software provided you many mechanisms to get your critical bugs fixed.
- theoh 11y agoAbout IRIX, an old but entertaining document: http://www.art.net/~hopkins/Don/unix-haters/tirix/embarrassing-memo.html http://www.art.net/~hopkins/Don/unix-haters/tirix/embarrassi...
- aswanson 11y agoHow did the management at two successful companies that had delivered great products for a long period of time get so shitty so fast?
- GFK_of_xmaspast 11y agoThe 90s was, in retrospect and sort of at the time, one hell of a decade.
- SixSigma 11y agoBy trying to beat MicroSoft rather than servicing customer needs.
- hga 11y agoThat would be Sun, right? Particularly Scott McNealy: https://www.google.com/search?q=scott+mcnealy+microsoft https://www.google.com/search?q=scott+mcnealy+microsoft (not needed SGI's special sauces, almost all my direct experience here is with Sun and DEC). In addition, the two biggest and interlocked problems with them that I know of are their ruining their reputation in enterprise space when they messed up an off-CPU chip cache design, required customers to sign NDAs before they would try to fix it, and then blamed the customers and the supplier of the SRAM chips for their non-fault tolerant design. Instead of, you know, the IBM ethos of fixing the customer's problem. Related, in terms of zapping the other channels of sales that might have saved them, was a total internal sales focus on enterprise sales. If you weren't making enterprise sized purchases, several million dollars worth, you were sent to a 3rd party like a VAR, but they typically wouldn't give you the time of day, they certainly were difficult to deal with when I worked with companies that bought one or two Sun workstations. Adding a web site that allowed direct purchases mostly added salt to the wound. What this meant was that startups that needed more Sun kit than they could charge to a credit card on that web site were forced to go with Dell, who's machines weren't as good, but had the strange property of being willing and able to sell them (HP's sales operation was too screwed up at the time to be a big player here). So a bunch of companies learned how to make do with Dell servers, and the successful ones never looked at Sun when they got big enough. So they missed out on the big post dot.com sales. A final straw when they started struggling was reducing the quality of their systems, e.g. saving one chip on a motherboard by using a chipset Marvel controller instead of the well tested and trusted Intel one. And I read that Joylent stopped buying Sun hardware when they started silently changing the lights out hardware they were shipping in their systems (IPMI or the like).
- chrissnell 11y agoI worked for an early cable high speed internet provider in the mid-late 1990s. Amongst other things, we used SGIs to power the proxy-cache servers at the edge of our customer networks. In some places, we even had SGIs on racks down in manholes where some of the headends were located. I also remember some SGIs up at our northernmost outpost in, I believe, Fort McMurray, Alberta. One of them had a hardware failure but they were under service contract and some poor SGI tech (or whatever company they had subcontracted to) had to make a day-long drive up to fix it.
- bluedino 11y agoNot quite. There were no 64-bit PC's back then or even a quality PC workstation Unix. Your alternatives were things like HP workstations or a DEC AlphaStation which were fairly cheap. Eventually apps started showing up on NT but they were garbage for quite a while. Once Linux was available, along with the Pentium, we never bought another big-name Unix server except for some high-end database stuff.
- kjs3 11y agoThat's not really true at any level. There were decent i386 SVR3 releases before 1990 and some very nice SVR4 releases soon after (particularly, Dell). I deployed a lot of them as an alternative to workstations. HP and DEC were not cheap in that timeframe, though you could argue they were more capable. And while not a "PC", a 64-bit desktop showed up in the form of the R4000 in 1991.
- spitfire 11y agoIn 1997 Irix was still vastly better than Linux on the development front. Mips Pro C++ compilers could actually handle the STL and had useful error messages. Irix came with debuggers that you could use. If you were doing perl however there was no benefit. Things changed very rapidly though. By 1999 Irix was languishing badly.
- rwmj 11y agoAll the machines in the picture were running FreeBSD which was our development platform. We were doing kernel & networking development work and BSD was much better documented because of the second Stevens book. Userspace was a lot of crappy C++ code using the bits of the STL which worked across gcc / whatever Sun's compiler was called / Windows VC++ (which was not a large intersection at that time).
- Theodores 11y agoI was doing 3d graphics. We had broadcast television on analog at the time with engineers that did not know how to use computers running the show. The cost of an O2 SGI workstation was £10k or so - you could buy a broadcast spec clock for that sort of money. Digital tape machines - Digibeta - cost £40k, plus you would always need more than one. That whole industry went race to the bottom. Internally you also had people not willing to learn the new PC tools. So there was a while before the PC inertia got going. On SGI machines doing CPU locked tasks, e.g. frame rendering from a render queue there was a simple cost equation with using PCs, off the shelf non SCSI disks were another saving, as was RAM, a tenth the price on the PC. Memory for SGI machines would need board level authorisation! So with many SGI use cases the CPUs would be doing very little. The graphics pipeline would be doing all the real work and you would hit your polygon limit in there. Only if you went CPU intensive - software rendering - would you Max out the CPU. Most SGI use cases were in capital intensive industries where proper money was paid for proper professional kit. Those 6K£ D1 monitors? Yes you could have a wall of TVs instead but that would not happen. So white box PCs took a while.
- drudru11 11y agoSGI computers had value, so people paid $$ for that value that they couldn't get anywhere else. For those people, it wasn't overpriced. Like many companies that do really well, they eventually squandered that edge. To answer you question... why did people used Suns for several reasons. These reasons are not in any quantitative order. 1. Religion - many people watched people smarter than them achieve success with Sun. Given their growing market size, they had a bigger religion. So whenever somebody tried something different and failed, it reinforced the religion. This lasted for quite some time (til early 2000's). Once the internet hit (1995,1996), just believing in this religion was a successful move. 2. Educational market. Sun gave educational institutions discounts. Students and Professors liked Suns and that created momentum when they left. Remember, even to this day, some people will stick with a system just because they know it better. 3. Sun systems were typically more data center friendly than SGI, HP, PCs, and especially SGI. They had serial consoles with OpenBOOT for lights-out-management - this was huge. They also stacked up nicely and fit into racks easily. If you knew what you were doing, you could do everything remote. 4. Many solutions had Sun picked as their configuration. For example, if you wanted to run Oracle or Veritas, you must buy Sun. It was not uncommon to see Oracle, Veritas, Sun, and Cisco all over the place in any data center in the valley. That was the template. 5. Much lower risk for hardware integration at scale. The hardware was targeted for Unix and scale. You knew that you could put 8 gigs in that Sun. You knew that the EMC RAID was going to work. So, spend the $, and move on to the problem. This was not the case with PCs. Sometimes that Intel motherboard, or hard-drive, or mis-spec'd RAM burned you on PC hardware. This could cost a lot of time and $$. 6. OS Maturity - they had just enough of an edge on the competition here. With a small amount of work, you could have something that could reboot quickly and stay up. The monitoring and tracing tools were much better. With skilled people, you could really make these systems work really well. This was in a time when many kernels had bugs and couldn't stay up for longer than 24-48 hours under normal load. SGI Irix was a great design, but many skilled people still had issues. I remember HPUX requiring weekly patch sets. The internet really exposed how bad these OSs were for server usage. 7. Java - once Java took off, that created a halo effect for Sun hardware. This combination created an enormous wave of $$$ to go to Sun. Sun had its warts for sure, but SGI, HP, DEC, etc all blew it compared to Sun. BTW, I was always more of a NeXT guy (we all had our religions), but I respected and liked Sun solutions. However, once I saw what a Pentium Pro combined with the maturity of FreeBSD (1997), I knew the workstation guys were in serious trouble.
- Keyframe 11y agoI've been doing 3D and film/video at the time on SGIs. While true that CPUs gained in performance on other platforms, IRIX machines were 64-bit, had great 3D accelerators (OpenGL was SGI's in fact) and - most importantly - they had bus bandwidth which was very important for large amount of data image processing required. Also, proprietary I/O was available for video and film which was only starting to come out for PCs (FAST and TARGA cards were getting cheap). It's all funny considering there was a void in the market. SGI (and discreet for video and A|W and Softimage for 3D) was great, if you could afford it, for high end broadcast and film and PC was nearly usable for video, but not for film still. Mac was great for video editing, but not for 3D... etc. Amiga was great for everything for video (not film), but it was, at that point, dead for several years.
- gcb0 11y agoSGI got lazy. It had the market for video and audio in its pocket. mostly because of what you said: I/O bandwidth! even if the RISC cpu was garbage (despite what the uninformed article says), moving from PC to SGI was like adding a SSD to your machine! it was night and day when working with video. Then microsoft come around and used its obnoxius tactics to take hold of firms such as Autodesk and others. which all dropped unix support to focus on windows. that was the nineties, when CGI quit being a small niche market to be the holywood standard, i.e. a very profitable niche market. apple scrambled at the last minute and got some companies too. the most important at the time was shake or tremor (never remember what was the product and what was the company) and quickly reworked the products to only use quicktime video exclusively, which obviously killed it down the line. because quicktime was, is and always will be garbage. i think at that time they got what is now their video suit btw. but i'm not sure. anyway, SGI ended up with no excuse for people to pay the premium price, not even on video editing. So everyone moved to PCs or macs. most of the custom software started to be written for linux (e.g. renderman) which SGI tried to corner by moving its server line to linux. but after itel and nvidia crushed them in performance /watt and /price, they just died for good. tl;dr; there is nothing special about MIPS. i give that assembly for mips is much more comfortable than risc. but meh. you always have to code for the lower common denominator anyway.
- 11y ago
- lobster666 11y agoYeah, PCs were cheaper than Unix workstations in 1997, but a Ford Pinto is cheaper than a Ferrari... Linux had serious shortcomings in 1997. It was OK as a web server or mail server or for home-grown number crunching software, but the NFS implementation was full of bugs, and none of the industry-specific commercial software had been ported. Windows 95 was really, really bad - it crashed all the time, barely had a functioning network stack and had no multiuser functionality. BMCs and KVMoIP hadn't been invented, so for PC-based servers you needed a keyboard, monitor and set of hands for each rack. SATA hadn't been invented either and with PATA you were limited to two drives per controller, no external drives and it was always a crapshoot whether any particular permutation of drives and mobo would actually work. Expansion slots were ISA or mixed ISA/PCI so you got to have fun with IRQ, DMA and I/O address settings. The mechanical design of PC enclosures was awful to mediocre. Sun mechanical design was a little better (retaining springs, etc) and SGI mechanical design was much better (tool free servicing). ECC and parity RAM were exotic on PCs but ubiquitous on Unix workstations, and the operating system actually did the right thing after a parity error rather than either ignoring it or freezing up. Also, regardless of the raw CPU performance, Suns running SunOS seemed more responsive under heavy load than PCs running Linux. I'm not sure why, and I don't have numbers to back it up, but that's how it seemed at the time. PC BIOS is still worse today than Sun OpenFirmware was in the 90s, but we put up with it because it's good enough and for many applications there's no other reasonable game in town.