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!"
Here are the images, mirrored:
t3d_2_big.jpg
td3_psus_big.jpg
t3d_wiring_big.jpg
t90_2_big.jpg
t90_system_board_big.jpg
Working Link
Mirror dot has mirrored the link here.
... or, if you're ever near Mountain View, California, why not see them in person (and a whole lot more)?
Computer History Museum website
http://templar.storyinmemo.com/
SIG: HUP
Armari can also put you together a decent high-end workstation.
Back in the day (c. 1999) I needed a new workstation. Armari set me up with Dual PIII-400's, LVD-SCSI HD, lots of RAM. Man that was a dream machine in it's day. Set me back of the (then) equivalent of $5,000 but it cut through my compliations like a knife through butter.
Still running. Man it's a crap machine now though!
http://www.whitepost.pwp.blueyonder.co.uk/mirror/t 3d_2_big.jpg t 3d_psus_big.jpg t 90_2_big.jpg t 3d_wiring_big.jpg t 90_system_board_big.jpg
http://www.whitepost.pwp.blueyonder.co.uk/mirror/
http://www.whitepost.pwp.blueyonder.co.uk/mirror/
http://www.whitepost.pwp.blueyonder.co.uk/mirror/
http://www.whitepost.pwp.blueyonder.co.uk/mirror/
Seymour Cray never designed the T3D supercomputer. Seymour split from Cray Research Inc (CRI) to found Cray Computer Corp (CCC) in 1989. At CCC he designed the GaAs Cray-3 and stillborn Cray-4. After CCC folded in 1995 he founded SRC Computers which was his first attempt at using commodity CPUs. SRC exists to this day but changed focus after Seymour's death in 1996. Other crayons may have better info but I believe that Steve Chen designed the T3D at CRI. Those of us who knew Seymour still miss him. He was quite simply the smartest man I have ever met.
Considering how tech savy the author seems to be, it's interesting that he doesn't understand what an MPP is. The T3d IS an mpp, made in response to a wave of mpp designs in the late 80's taking some of cray's market share (thinking machines, paragon, etc) MPP, incidently, stands for Massively Parallel Processing; massively as in hundreds, not 32.
The T90, on the other hand, is a pure SMP. The processors all sit on a shared bus (actually 256 parallel shared buses). Each CPU was really fast (for the time) and had really big pipes to memory, and really expensive.
Sorry, just picking nits.
A Gibson?!? No one could ever hack that!
[n8.r0n] http://petesweb.spymac.net/
Please give an example of any piece of code that runs faster on a 486 than any machine ever classified as a supercomputer. No? Didn't think so.
Most forms of user-interactive programs or branch-heavy logic. You see, many supercomputers used EXTREMELY long memory pipelines that placed the processor 100+ clocks behind the memory being fetched. A single branching instruction usually resulted in a task switch to the next queued process, thus keeping processor usage high.
The end result is that these machines could crunch streaming data at an extraordinary rate, but couldn't compete directly with scalar processors for branching performance. I believe that's the reason that VAX front-ends were used on Crays. Running a user interface on the Cray would have caused a great deal of interruption in the processor's duties, and thus would have substantially decreased performance.
Wikipedia (yes, I love to piss off you anti-WP people) has an explanation of vector processors. Take careful note of the section on pipelining, as that is the key to both the supercomputer performance, as well as the lack thereof.
I'm confused.
Yes, we know. Otherwise you wouldn't post as an AC.
Javascript + Nintendo DSi = DSiCade
If you want to get up close to the Cray Supercomputers of the past, visit the Chippewa Falls Museum Of Industry and Technology in Chippewa Falls, WI. It is about one hour and 45 min. east of Minneapolis, MN. The museum is open daily. Adults can get in for $3.00
Site: my.execpc.com/~cfmit/
Museum of Industry and Technology
21 East Grand Avenue
Chippewa Falls, WI 54729
715.720.9206 tel
The University of Minnesota also has a Virtual Cray Museum. http://www.cbi.umn.edu/exhibits/cray/index.html
"It is a miracle that curiosity survives formal education."
A 2ghz Athlon can pull just over 3 GFlops.
The T90 had 32 x 450mhz CPUs that could do 4 ops per cycle, which comes out to 1.8GFlops per chip and 57.6Gflops for the whole shebang.
The real differnce however is not raw cpu horsepower, but memory bandwidth, latency, and scaling. I don't know nearly enough about supercomputers to be able to explain that in detail.
-Dan
I haven't actually RTFA but the two models listed in the submission aren't actually Seymour's designs. The T90 is Seymour inspired but I don't think he was anywhere near where it was built. I don't think the T3E has anything to with him, and am not sure he would have liked it much since its kind of a commodity CMOS(DEC Alpha) based MPP and nothing like anything Seymour would have designed. Thats one of the problems with Cray's deity status, his name has been tacked on so many companies and computers at this point its diluted his "brand" and that is unfortunately what his name is today, a brand that that doesn't have anything to do with the man anymore.
"because companies aren't interested in pushing the technological envelope the way Seymour did."
Maybe its because after the Cray-2 and its derivatives Seymour's approach stopped working. There was only one Cray 3 built, it was at NCAR if I recall, and the Cray 4 was never finished. His need to push the envelope pushed him to GaAs and GaAs was hard to make work and very expensive, the antithesis of the commodity CMOS that won. Maybe he could have made it work if he'd had more money and lived. But the fact is commodity CMOS won for some good reasons.
Its also worth looking at his efforts at CDC on the 6600, 7600 and 8600, while offering amazing performance for the time, and quite successful, developing them nearly bankrupted CDC. Its a lot easier to "push the envelope" if you have buckets of money, presumably from government subsidies than it is if you are trying to keep a company afloat. Not sure you've ever run a company but its not easy, and Seymour's, spare no expense approach tended to give the people trying to balance the books fits.
He was a GREAT engineer and he made some major contributions to computer engineering. Not sure I would go so far as to deify him.
@de_machina