Workstations

Lain, do you own any workstations or high-end hardware? Regardless of age/era. I've always been oddly fascinated by this kind of stuff because it represents the very pinnacle of computing technology at the time it was manufactured. My father actually owns an SGI Indy that he was allowed to keep from his job back in the 90s. I've played around with it a few times and it's probably one of the most UNIXy things I've experienced. On the contrary , I also own a Lenovo ThinkStation P920 that I mainly use for work in GIMP and programming. It's pretty much the base spec model coming in at $4,500, but I have confidence that it'll serve me well for possibly a decade and a half if I care for it and upgrade it along the way. Anyways, I apologize for the spergout.

Ain't no spergout, that's what this place is all about. I have an SGI O2 somewhere around here. I should dust it off and see if it boots. The Indy was so aesthetic looking. They made really bizarrely-styled computers. Ever see a Tezro?

>>51794 Yup, I've heard they're fine machines too. Wasn't the Tezro one of the last SGIs?

>>51795 Seems that way. RIP, SGI. I still wear the hoodie with their logo on it sometimes. They made some good stuff. And you're right, definitely very UNIXy. Wonder if anybody bothered to play around with getting their OS to run on newer hardware or updating things here and there.

Check this ѕһіt out right here—now that's a flippin' workstation. Awooga!

There is quiet a lot of dual Xeon workstations on 2nd hand/refurbished/postleasing market. HP Z series are among most popular. So if you are looking for computer that accepts fugtone of memory with ECC option and can have 12 or even more cores per CPU depending on generation - go for it.

SGI's were really cheap not too long long. You could find high spec'd Octane2's on eBay for $400 at most. Fuels for even $600-800. And then this little soykaf came along and had to ruin all the fun.

>>53010 The Onyx looks like a 90's themed fridge

I used to have a DEC Multia with an Alpha CPU. I could never manage to get it to boot and therefore did nothing with it.

>>53010 And did what exactly?

>b SGI >try to be Windoze / Intel / Mac >dieded RIP They even had a Wine environment to run Windows software on a Linux based OS back in the 90s. They stopped leading and started following, so there was no point to them in the end.

Workstations were killed by PCs. They were never a particularly good deal - they were really expensive and proprietary. The cheap PC market killed them by being more affordable, and eventually even more powerful. IMO, the problem is the software on the workstations, UNIX, was better. Linux fixed this. There isn't really a reason to go back - I don't want a proprietary UNIX on expensive proprietary hardware. We have open source software and open standards for PC hardware - which the adoption of mobile phones is destroying. Now "Workstations" are just a marketing term for high end PC hardware marked up 200% with a long warranty and support guarantee. The ascetic of SUN tho.

>>53363 I would rather call them a hybrid of PC with server - looks like PC and accepts PCI cards but has ECC RAM and server grade CPUs - like Xeons etc. and is built in modular way - the way that allows fast replacement of components including PSU.

>>51792 >get in argument over Unices with a guy on AnimeBytes forum >he PMs me later offering me an SGI O2 I still haven't gotten around to getting IRIX on it though

What's the point of using of these for home use? Rendering? What else?

>>54718 Anything you want to run long enough that you prefer to not re run again because of some small error caused by cosmic radiation or instability of CPU on higher clock speeds. This is why workstations have ECC RAM and lower clock speeds.

>>54677 Workstations where produced by a variety of workstation companies which are all dead now. A variety of PC companies have started offering 'workstations' but these are mostly PCs plus warranties and guarantees on downtime. The mostly don't have redundant power-supplies, but you are correct ECC and Xeon are popular on 'workstations'. Xeon was never popular on real workstations - SPARC, MIPS, Alpha, were.

>>55330 These architectures in workstations and supercomputer sector were murdered by lowering prices of x86 that was soykafty in many aspects but was cheap enough to be competitive when paired with floating point coprocessors around 1990 when we take in account that most code was still compatible with newer PCs that you could update with new cards, ram etc during it's life cycle. PC components were made by many companies meaning there was actual competition meaning also lower prices of these components. Custom workstations you have in mind were often one model boxes that no one else was making components for. Powerful, but expensive and with almost not existing choice of new modules/cards when we compare them to broad rang of PC compatible components. Around 1997 if memory serves me correctly Intel started introducing vector extensions to their processors meaning introduction to the PC market of stuff that was earlier a domain of workstations and supercomputers (and what in many cases was implemented as separate from CPU cards). That made life of remaining hermetic custom architectures even harder. Introduction of hardware 3D accelerators to the PC world was another blow. This is more or less why modern workstations or as you prefer "workstations" are indeed PC platform on a general concept level just in few strategic places it has components that are ensuring precision/validity of data to avoid occurrence of hardware cause data corruption that can ruin results of long extensive calculations. "Normal everyday PC" do not use such components as they are used for other stuff when precision on that level do not matter. If however in the future we are going to achieve technology when ECC like mechanism would be rational and cheap enough to be used in memory for generic computations including these everyday uses by sockermoms and teenage gamers - we could potentially see no longer division between "server"/"workstation" RAM sticks and "home computer" sticks. Modern Crays have (or at least had few years back) own proprietary network cards per their ummm "motherboard(?)" they use to communicate with other boards. If similar or better solution would even enter PC market Cray possibly could abandon own proprietary solution and use third party generic cards as it could potentially lower production cost. When you use generic component you can cut spending on developing and researching own solutions if generic market parts are good enough. That my outlook on why general purpose computing solutions are gradually killing custom proprietary stuff that only limited amount of producers can or/and know how to produce parts for. Economics matters.

I think peak example of what's wrong with idea of workstations on the general market were PIXAR Image Computer and NeXT workstations. Too powerful, too advanced (being too fast forward can make you run ahead the market/industry too much you might be early but being too early means less clients willing to buy your new and expensive hardware, especially when they not yet figured out what uses your hardware might have and would using it be profitable financially), too expensive (very few buyers can afford them), too niche (less buyers) and too isolated as architectures in market where much cheaper (serial production in huge quantities) PC standard (you don't need access to secret proprietary documentation to manufacture a network controller card or graphic card) steamrolled other platforms.

>>55826 >Too powerful, too advanced (being too fast forward can make you run ahead the market/industry too much you might be early but being too early means less clients willing to buy your new and expensive hardware, especially when they not yet figured out what uses your hardware might have and would using it be profitable financially), too expensive (very few buyers can afford them), too niche (less buyers) and too isolated as architectures in market where much cheaper (serial production in huge quantities) PC standard (you don't need access to secret proprietary documentation to manufacture a network controller card or graphic card) steamrolled other platforms. This is in fact not what happened. You have the order completely wrong. Minicomputers where the "cheap" computer and where shared. The new evolution, Workstations bought professional computing to the individual instead of being shared. There were no PCs - it was workstation or shared mini-computer for a while. Eventually micro-computers came about, but they were incapable of being used professionally, and workstations continued to thrive. Eventually, low cost PC hardware became virtually as capable as workstation hardware, but way cheaper, and displaced workstations entirely. One Hilarious Artifact of this, era is Windows NT was designed to run on multiple architectures, including DEC Alpha - just in case. Wikipedia has this to say about it: >Workstations tended to be very expensive, typically several times the cost of a standard PC and sometimes costing as much as a new car. However, minicomputers sometimes cost as much as a house. >By the early 2000s this difference [between PC's and Workstations] largely disappeared, as workstations now use highly commoditized hardware dominated by large PC vendors If you are intrested in this history of computing, Steven Levy's Hackers is an enchanting read.

>>55827 Not exactly. Comparison was between workstations and PC. Not workstations and mainframes/supercomputers or generic microcomputer and mainframes. You eliminated PC from discussion, while post you replied to was developed (good few years after introducion to the market) PC centric perspective. Therefore it makes no sense to apply to pre PC era. You are correct in the further part tho. But it was not stated otherwise by anyone so far. >By the early 2000s this difference [between PC's and Workstations] largely disappeared, as workstations now use highly commoditized hardware dominated by large PC vendors which was already mentioned in this thread