We've all seen audio amps with a rats nest of wires. How about a 35 year old supercomputer with a snake pit. I have no idea how anyone could possibly troubleshoot it.
Attachments
There is a Cray1 XMP in the science museum. there is a reason for that wiring which is so that latency in every connection is the same. It's a bit worse that the wirewrap protos the older of us remember, but managable. The Cray 1 was installed at the Harwell labs near oxford and they claimed it did more calculations the first day than had been done on the site in the previous 30 years. It was a revolutionary product in its day.
As we all know, a computer never fails :-D :-D :-D!
Best regards!
BTW, this was in the Deutsche Museum in Munich
More than one person has mapped the entire Cray-I CPU complex, onto a single FPGA. One guy even build a 1/10th scale replica of the cylindrical chassis to house his FPGA-Cray. (link to article)
I took a training course on the T3D at the Cray Egan facility back in the early 90's. It was winter in the land of a thousand lakes, and on the drive from the airport to Egan you passed a lot of frozen lakes and ponds. But the lake in front of the Cray building wasn't frozen, and you probably could have gone swimming in it--it was the heat exchanger for the machines inside.
The Cray president at the time joined us for one of the lunches, and explained that Cray success in those earlier years was due to the weaving and embroidering skills of the women in the Chippewa Falls area. They needed those skills for all of that carefully cut wiring, and he said that the Minnesota-Wisconsin area was probably the only place in America where they could find those skills.
The Cray president at the time joined us for one of the lunches, and explained that Cray success in those earlier years was due to the weaving and embroidering skills of the women in the Chippewa Falls area. They needed those skills for all of that carefully cut wiring, and he said that the Minnesota-Wisconsin area was probably the only place in America where they could find those skills.
Last edited:
I took a training course on the T3D at the Cray Egan facility back in the early 90's. It was winter in the land of a thousand lakes, and on the drive from the airport to Egan you passed a lot of frozen lakes and ponds. But the lake in front of the Cray building wasn't frozen, and you probably could have gone swimming in it--it was the heat exchanger for the machines inside.
The Cray president at the time joined us for one of the lunches, and explained that Cray success in those earlier years was due to the weaving and embroidering skills of the women in the Chippewa Falls area. They needed those skills for all of that carefully cut wiring, and he said that the Minnesota-Wisconsin area was probably the only place in America where they could find those skills.
I wonder why the Belgian lace makers didn't beat 'em to it. Nice story!
One of those PBS shows from the late 1970's, I think it was "Connections" with James Burke, traced the punch card back to the jacquard looms. (The entire series is available on youtube.)
"I had a friend..."
A classmate of mine went to work for Cray in Chippewa Falls. His first job was to troubleshoot and bring up a machine after is was first assembled. They had two basic problems; bad wires and bad chips. They would pat the wire mat to isolate the wire problems and there were folks whose job was to replace ships on the boards. The boards had a copper core which extended past the sides of the PCB and was held by a refrigerated (freon) rack for cooling.
Once the machine was running properly in Chippewa Falls, which would take about 6 weeks, it would be disassembled and shipped to the customer facility. My friend would follow the machine there and make sure it was running properly for the customer.
A classmate of mine went to work for Cray in Chippewa Falls. His first job was to troubleshoot and bring up a machine after is was first assembled. They had two basic problems; bad wires and bad chips. They would pat the wire mat to isolate the wire problems and there were folks whose job was to replace ships on the boards. The boards had a copper core which extended past the sides of the PCB and was held by a refrigerated (freon) rack for cooling.
Once the machine was running properly in Chippewa Falls, which would take about 6 weeks, it would be disassembled and shipped to the customer facility. My friend would follow the machine there and make sure it was running properly for the customer.
There was the story that apple used a cray to do some design work,
and cray used an apple to design the cray. Nice story even if not true!
The first part is definitely true, running the Mac OS as a virtual machine.
Don't know about the second part.
Last edited:
I took 2 thirty two K core memory modules to a air port shipping terminal in 1968 for a RCA Spectra 70 Model 45. Both about the size of a small bread box. I declared the value to the shipping agent of a half a million dollars. This was right down the street from McDonnell Air Craft company which had a Huge, running in the middle of winter, icy cooling pond in front of the MCAuto data center on Brown Road. Computer Techs back then were treated like true kings and paid pretty good too. Our biggest machine was a 64 bit 512 K of which only 4 were built. No IC's to be found, all discrete logic.
> 1968 for a RCA Spectra 70 Model 45
My dad was in RCA Camden in that time.
The Cray layout kept things close for low propagation delay. Well before Cray left CDC and rolled his own, my dad explained to me that when computers approached "nano"Second speeds they would have to be less than a foot across. 1nS ~= 1 Foot. Nanos were not happening in 1965, but some thinkers were looking ahead.
My dad was in RCA Camden in that time.
The Cray layout kept things close for low propagation delay. Well before Cray left CDC and rolled his own, my dad explained to me that when computers approached "nano"Second speeds they would have to be less than a foot across. 1nS ~= 1 Foot. Nanos were not happening in 1965, but some thinkers were looking ahead.
Our biggest machine was a 64 bit 512 K of which only 4 were built.
No IC's to be found, all discrete logic.
I remember the EE prof teaching us discrete RTL and DTL, while saying
"now you guys will be using ICs with this stuff all in one package".
> 1968 for a RCA Spectra 70 Model 45
My dad was in RCA Camden in that time.
The Cray layout kept things close for low propagation delay. Well before Cray left CDC and rolled his own, my dad explained to me that when computers approached "nano"Second speeds they would have to be less than a foot across. 1nS ~= 1 Foot. Nanos were not happening in 1965, but some thinkers were looking ahead.
Our basic Clock time for the system was 200 ns. The Spectra 70 45 was the first computer ( as in main frame) to use IC's. These IC's were made of ECL. Emitter coupled logic was "SUB" nano second. Pico second switching times required very fast scopes to work on them. The "sexy" was that the logic for the main frame machine code was in an E core ROM.
RCA Spectra 70/45 computer brochure - CHM Revolution
You can not see on a scope the 4 or 5 gig hertz signals on today's chips and make any sense of what you are looking at.
My first computers were made of vacuum tubes, Analog Flight Simulators.
Last edited:
It seems that as technology is getting more advanced, the applications are getting more stupid. I mean, why on earth do we need smartphones that are more powerful than a Cray-I to play Scrabble nowadays?
Look at if the other way, its scary.
Grad student "I need about 2 hours of time on the cray"
Director "why?"
Student "I want to fling some birds at some pigs that stole the bird's eggs. Dont worry, it has some pretty good real time ballistic calcs and even some cool space gravity sims"
Student " I also want to listen to some high quality audio, but it's not fast enough"
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
- Home
- General Interest
- Everything Else
- A supercomputer with a snake pit of wires