Slashdot Mirror
A Private Home For Retired Supercomputers
Steve writes "Every geek has wanted to play with a Cray supercomputer. Hexus.net had the rare opportunity to meet up with a man who has something of a fetish for collecting them! They got a look at some of the amazing kit Armari - a systems integration company - have in their possession. Ever wanted to see inside a Cray T3D MPP, or maybe the gargantuan machine that is the T90? Now is your chance!"
I had the luxury of playing with a Cray YMP at the MoD (in the UK)... Just a big number cruncher with a VAX/VMS front end. Lovely to look at though.
Together, We Can Make Slashdot Better. I Do NOT Mod ACs. - Check Me Out
can you imagine the AC (both power and cooling)demand for a good quake bake? :)
I'm told you are what you eat, does that mean I can be you by tomorrow with some A1?
the mummies of the digital age. we're like treasure hunters only instead of jewels and crowns we're looking for gold lined circuit boards.
-Teiresias
this makes me miss the days of trs80's and writing basic code that you saved to a tape recorder.
this makes me miss punch cards and the fear you had of getting them out of order.
it makes me miss...ti calculators where if you held down three of the corner keys, the thing would bypass the on button.
sigh...i miss the old days.
Is it 5:30 yet?
Can anyone tell me how fast these things are compared with, say, an Athlon 2000+?
At my former university, they had a corridor with a glass wall which went past the machine room full of supercomputers, many with flashy-looking blinkenlights arranged into grids or in the form of graphical processor-monitoring screens. There were often some weird and wonderful smaller machines, like some Linux-running, Itanium-powered (according to the labels) SGI workstations - this was late 2000, early 2001 or so, and I haven't seen a single Itanic since...
:-)
The biggest machine was a huge Cray T3E - I don't recall any blinkenlights on it, but it didn't need them! I recently heard that turing.mcc.ac.uk has since been dismantled, presumably because it was no longer cost-effective for its mere few hundred Gflops. I've no idea what was done with it and its parts, or what (if anything) it has been replaced with, but it's what I thought of when I saw this article.
Tedious Bloggy Stuff - hooray?
This is a guy's PRIVATE, PERSONAL Cray collection:
h tm
http://www.digibarn.com/friends/jamescurry/index.
It has to be the most comprehensive collection of Cray systems in the world (including Cray's facility in Chippawa falls?).
(Please do not post it on the front page of slashdot without digibarns permission). Those pictures are quite a bit outdated, as he no longer lives in that state and has added more systems since then.
I believe he had over 11 before. He donated a few to someone, I forget who.
Southeastern Virginia REPRESENT!
I wonder if any computer system collectors have any IBM System/360 machines that are still in operation.
S/360 is interesting because it was one of the first standardized architectures created by a computer company. Before that, each seperate machine had its own instruction set and architecture, and they were incompatible with each other.
A mid-sized functional IBM System/360 is quite a sight. Multiple cabinets of core memory, CPU cabinets, tape systems, consoles with thousands of blinkenlights... A real fun system to watch in operation.
Hopefully someone out there still operates one for fun. It's expensive, but we have rich geeks right? }:)
-Z
A better question is why geeks would have a fetish for Crays at all these days, other than those with a historical bent. Believe it or not its not like you are going to have some transcendent experience logging in to one, unless you really get off on Fortran and vectorizing code. Obviously they have some fine massively parallel machines but so do a lot of other companies. If you have an app that you want to run that fits on a Cray then maybe its still interesting to you but thats not most people.
OS wise they do have some good software for niches, but in general Linux is better and more broadly developed, and of course some of Cray's machines are running Linux too. CrayOS has never had the critical mass of developers you need to be polished, though again in certain niches it has some really great efforts.
Big Iron had its place a decade ago and back. The tyranny of Carer Mead and CMOS processors has led to an age were smaller, cheaper and mass produced trumps big, expensive and custom built for most, though certainly not all, applications.
For example its why most of us are running Nvidia and ATI GPU's and not SGI big iron graphics anymore. And of course SGI's R8000 among others carved up the low to middle end of Cray's market and ended their last incarnation as an independent company.
@de_machina
Good point. Here's how to get the experience, or something close to it. First, get a Pentium III machine, maybe 800Mhz processor. Fill it up with at least a gig of memory, and a terabyte of disk space, implemented as a large array of IOMega Bernoulli drives, 40 megabytes each. Install Linux on it. No, not Redhat, I mean Slackware. Do not install the X server. Install your Fortran compiler. Oh, and I forgot to mention, you will interact with this machine through a VAX front end. You won't log into the machine directly, but will work on the VAX and submit jobs from the VAX to the simulated supercomputer. When the job is done, the results are sent back to the VAX front end computer.
Ta Da! You've got a machine that's very comparable to a Cray I from more than 20 years ago.
Fascism trolls keeping me up every night. When I starts a preachin', he HITS ME WITH HIS REICH!
I recall studying an early (Nazi Germany era) jet engine. It had all kinds of very sophisticate systems (e.g. liquid cooled turbine blades) to get around metalurgical limitations. Some of the features actually went from nearly 50 years before they were implemented again when materials technologies were a limitation and exotic work-arounds were required.
Yes, history may have passed these CRAY machines by, but the engineering problems once solved may be encountered again and it would be a shame to disregard that research because the "big iron" is "old iron".
When the people fear their government, there is tyranny; when the government fears the people, there is liberty.
I visited these dudes a few years ago, when down at my friends employers and visiting their cuppliers (Armari). Got loads of pictures of the Cray T3d right here. Wonderful machine, wonderfully kept. :-)
Dans machine wasn't quite (apparently) 'the first off the production line - Edinburgh uni (where this T3D came from) wanted one that could be upgraded without a lot of hassle. Cray could only offer them this one, which was their testbed unit, wired for a full complement of processors, but not fully populated. That's why it's innards are absolutely stuffed full of wires. Each wire is also a specific length, to ensure that the length of time it takes for electricity to flow down the wire is accurately accounted for in terms of clock ticks.
The power switch that the author wished he'd taken a picture of is here
I loved Dans demo of the differing weight of cooling liquids. He had a milk bottle full of water, weighing a kilo or so, and then an identical bottle weighing about 3 kilos. The plumbing for the liquid cooling was done by a bottling plant systems manufacturer in Daytona if memory serves, and the metal braided hoses that are used in it are of the same type used in Formula One and Nascar cars. Wicked stuff
I helped test out that very pictured T3D machine on the lab floor in Chippewa Falls. I helped simulate the chipset for the T3D and wrote diagnostics to test it. I became the resident expert on the "barrier channel" which was a mechanism used for both syncronization and for low bandwidth one-to-all communication.
Of course, being the very first system, 6001's barrier channel had a bunch of issues to resolve. Armed with just a prom-emulator (each cpu used a serial prom to load it's initial bootcode) and a Tektronics O-scope with a couple probes, I managed to come up with some primitive diagnostics written in assembly to let us determine what was going wrong on the barrier channel. IIRC, there were some banks miswired and some delay signals set wrong which were causing the problems. Once we got past that hurdle, the rest of the bring up went fairly smooth.
The next box built, SN6002, went to the Pittsburgh Supercomputing Center, and I was fortunate enough to go on the install trip with it. I really enjoyed all the work I did on the T3D, as I got to work with it from design stage, thru test & bringup and eventually travel out to customer sites.