There have been components for an Altix XE cluster on ebay.co.uk for an age (XE 320 nodes, XE 270 head). Might as well paint over the hypercube with a big red Supermicro dot. The only thing sgi about the Altix XE is the facade (and I'm not even talking about the LED/blikenlights on the front), and the bankroller of the sales-person. o.0
Nice work on the videos. We need more 'blipvert-y' videos on youtube
. The place is full of clickbait top5 <whatever> videos which take too many minutes of my life to regurgitate about 6 sentences.
SiliconGraphics was diagnosed terminal about '98 imo (although I am a fan of the 320/540 Colbalt boxen, Clark was long gone by this point living in a lighthouse or something like that), the next few years were inevitability, running mainly on 'legacy' fumes
. 2006 absolute nail in coffin, what emerged was nothing SGI really (as existence of Altix prooves). Rackable did alright out of it though.
In regards to nVidia, it was three guys (mainly ex-SUN) that started it, however later they did buy 3 other guys (all ex-Silicon Graphics) who had little outfit going called 3dfx who just so happened to of revolutionised 3D graphics for the PC consumer market. I doubt they brought much to the nVidia table though, as GeForce was already doing hardware TnL when 3dfx got pwned (literally), TnL is based on the same principles of geometric transformation/clipping pioneered by E&S and somewhat refined by Jim Clark into the Geometry Engine. Ironically not actually accelerated by 3dfx boards, which is probably why they failed in the long run, unlike nVidia who seemed to follow the GE memorandum somewhat closer. All originally done by SiliconGraphics already the previous decade. So while all this nVidia/SGI arguing was going on later, the PC graphics world had already taken most USP from them and capitalized on SGI's inability to see the market outside gov pay check.
It could be argued pretty much every hardware accelerated 3D rastered image on a PC is only there because of Jim Clark's GE and SGI heritage. And then there's OpenGL heritage o.0... dead now, and it would seem that later advancements to the OpenGL standard did things that even SGI themselves couldn't keep up with. These days we have little left with Vulkan and DX12. Still rely on rasterizing for the transformation/projection/clipping etc, but per pixel operations broke us out of the fixed functionality pipeline, this all combined into a unified shader execution architecture (and now with OpenCL merging into Vulkan, a unified API) leaves me thinking you would really need to dig deep in the bottom of the barrel to find anything SiliconGraphics-ey.
The other SiliconGraphics thing I used to come across quite often was their STL implementation. Which was actually quite good!
RIP SiliconGraphics Inc.
So much of that article is begining to make sense. I heard rumors that the entire design team on the Colbalt chipset got fired day after launch of 320? The rest of the 'Visual workstation' line used nVidias newly clandestine aquired TnL technology, the same tech which allowed nVidia to dominate PC market over ATi and 3dfx. One of the reasons I like the Colbalt is (stereotypically) because of its unique UMA architecture, it was like a gfx system with expanable memory and upgradable GPU
CPU. Part of me feels it could have been the future had it not been let down by limited FSB/Socket options and it came at a time that the PC environment changed every second you blinked (in about 4/5 years we went from PII/233 to PIII/1400 o.0). Cobalt could have been the last true homegrown SGI architecture, not finished, rushed and then superceeded by nVidia's implementation of SGI's own technology with an SGI label.
If you can do some research on the Colbalt graphics (all aspects of) that would be very much appreciated by (at least) me.