Silicon to Supercomputer

The J90 logic is implemented using application-specific integrated circuit (ASIC) chips fabricated by IBM. There are 10 unique ASICs that are found in the processor and memory modules. A typical J90 system could contain about 230 of these CMOS chips. The photo below shows a processor module with the cover removed. Each module contains 4 scalar/vector processors. The space at the top of the board can be used for optional HIPPI interfaces or Y1 Channels to additional I/O Processors.

A Cray J90 quad processor module.

The ASIC chip types are:

• MBI - DRAM memory interface
• MAD - Memory side of memory crossbar for read data
• MAR - Memory side of memory crossbar for write data
• VA - CPU side of memory crossbar for write data
• VB - CPU side of memory crossbar for read data
• CI - Channel interface (I/O)
• JS - Shared registers for multi-CPU applications
• PC - Scalar processor and processor control
• VU - Vector processor
• MC - Maintenance and clock distribution

There is only one chip (called PC) for each scalar processor and one additional chip (called VU) for each vector processor. There are only 8 chips on each processor module for the CPUs and the rest of the 18 out of 26 chips are used for communication between processors or between the processors and the memory banks. This circuitry is the key to a "balanced" system where the memory bandwidth is great enough to sustain the rate at which the processors can operate on the data.

The summit station at El Misti, Peru (19,200 feet)

"The night is passed at the hut, and the final ascent to the summit made on the second morning. This occupies several hours, as the animal stops to rest every fifteen or twenty feet at this altitude. On two occasions I was obliged to walk a short distance to cross snow which had drifted across the path, and realized the extreme difficulty of breathing during the exertion required."

"The effect of the altitude upon me was chiefly to cause headache, sleeplessness and partial loss of appetite. On one occasion while at the summit I experienced a decided feeling of faintness for a short time."

- Winslow Upton, Physiological Effect of Diminished Air Pressure, Science, 27 December 1901

Misti summit station, Jan. 5, 1894.
Shut in by cloud [and] snow looking N. E.

During the academic year of 1896-97 Prof. Winslow Upton took sabbatical from his work as Director of Brown University's Ladd Observatory. He spent ten months at the new southern station of the Harvard College Observatory (elevation 8,050 feet) in Arequipa, Peru. His primary goal was to measure the geographical position of the station before astronomical observations could commence.

During this time he also made four ascents to the summit of the dormant volcano El Misti, which was the site of recording instruments (pictured above) maintained by Harvard. At the time it was the highest meteorological station in the world at an elevation of 19,200 feet.

Geared to the Stars

The telescope at Ladd Observatory uses a clock drive to compensate for the Earth's rotation and track the stars. Modern telescopes use electric motors but this one was built in 1891 before electric power distribution was common. The Observatory originally had gas lamps for lighting and the telegraph system was powered by "gravity cell" batteries. The telescope's mechanical clock drive is weight driven with the speed regulated by a centrifugal friction governor.

A closeup of the clock drive showing the governor in motion.
An optical sensor and precision timer are used to measure the rotation rate.

Power Up

The Cray J916 was featured at our monthly open house this past Saturday. We spent most of the day talking to visitors about the system and showing them the rest of our collection. We did make some progress in the morning before we opened. Dave took some photos while I was working.

Working on the system clock board on the J90 backplane.
The Central Control Unit is in the foreground.

Everything seems to be functional with one exception - there is a system clock PWR FAULT light showing on the Central Control Unit. We suspected a loose cable or board connector and traced the signal paths. We couldn't find the cause of the fault and will need to dig deeper another time.

The Dawn of a New Era

When I was a young child I would watch reruns of the original Star Trek. It wasn't so much the space ships or aliens that impressed me. It was seeing human beings just simply standing on another planet that moved me. It gave me the idea that there were other worlds out there, and that you could travel beyond the Earth to visit them. That sparked my imagination.

My parents would then change the television channel and again I would see people walking on another world. But this time it was on the 6 o'clock news. A grainy video of astronauts in bulky spacesuits standing on a monochrome landscape with the crackling audio of a voice calmly saying "Beautiful, magnificent desolation." It was, arguably, one of the few moments in human history when reality was more amazing than our wildest dreams.

Six months after Apollo 13.
Photo credit: Mom, Halloween 1970.

Prismatic Analysis

One of the projects that I've been working on is the Brashear astronomical spectroscope from 1891. We're trying to recreate the ability to record spectra using the original photographic plate holder. Here is the wooden plate holder mounted in place of the eyepiece assembly that is used for visual spectroscopy.

A focusing screen was created by mounting the ground glass in a foamcore frame that fits the plate holder.

Jedi vs. the Droids

How does the performance of a 20 year old supercomputer compare to the devices that we use today? Let's compare the Cray J916 to a recent laptop and a smart phone.

According to an archived copy of the Cray J90 Series webpage the vector processors have a theoretical peak performance of 200 mflops each, giving our 8 CPU system 1.6 gflops. But, your mileage may vary depending on the code that is running. One of the standard benchmarks used for supercomputers is LINPACK. Results for a J916 with the same configuration as ours are listed in Performance of Various Computers Using Standard Linear Equations Software by Jack J. Dongarra from June 1995. An 8 CPU system was measured at 1.436 gflops, an efficiency of about 90% of the theoretical peak.

"Just right for you" - it certainly is for us...

Next we'll run Linpack for Android. My Samsung Galaxy Note II has a 1.6 GHz ARM Cortex-A9 with four cores. Running LINPACK multi-threaded gives about 200 mflops, just a little faster than a single J90 processor. So, yes, that is (nearly) a mid-1990s entry level supercomputer in my pocket. At least on paper. We're really just exercising the ability to do floating point calculations, and this is not necessarily a good measure of system throughput on a real problem.

I estimate that the theoretical peak performance of the ARM is about 3 gflops or so, giving well below 10% efficiency. (I'm ignoring the GPU as I have no way to run LINPACK on it to benchmark it.) I should mention that the Android version of LINPACK is based on this Java Version and the low efficiency is in part due to the Java Virtual Machine.

But, overall, the Cray system with a 100 MHz clock speed has roughly 7.5 times the performance of an Android running at 1.6 GHz.

bootp

We are at the point where we can power up the Cray and begin configuring it. Next we need to work on the J90 System Console (or SWS, the Service WorkStation.) This is used to net boot the I/O Processors in the I/O Subsystem, which in turn loads UNICOS into the J90 main memory. The SWS is a Sun SPARCstation 5 running Solaris. We didn't get the original SS5 that came with the Cray so we had to build one.

The SWS with ThinWire Ethernet, FDDI, and graphics.

We had a couple of these lying around, but they weren't in very good shape. We cobbled together a system from the parts which is better configured than what would have been used when the J90 was installed in 1996. This system has the maximum of 256 MB of RAM and a 24 bit S24 TCX graphics card.

There are two Ethernet interfaces. The one on the SBus card is 10BASE2, more informally known as ThinWire. This is only used to connect the SWS to the two I/O Processors in the I/O Subsystem. This allows the IOPs to net boot from the SWS and is also used to configure and manage the system.

This is Arecibo Calling...

"Ironically, the globular cluster at which the signal was aimed won't be there when the message arrives. It will have moved well out of the way in the normal rotation of the galaxy." - It's the 25th anniversary of Earth's first attempt to phone E.T. Cornell Chronicle, Nov. 12, 1999

My estimate of the Arecibo transmission from a couple of years ago.
It looks like we may have missed...

In 1974 a ceremony was held to dedicate a major upgrade to the radio telescope at Arecibo Observatory. As part of the festivities the telescope was used to transmit a message towards Messier 13, the Great Globular Cluster of stars in the constellation Hercules.

"Scientists Hope to Reach Hypothetical Civilization in a Cluster of Stars" - New York Times, Nov. 20, 1974

System Ready

In a previous post I described the power requirements of the Cray J916 and the importance of the Central Control Unit (CCU) monitoring for faults. This is rather critical as the machine won't function properly unless these hardware status signals check out.

The back of the I/O Subsystem in the Peripheral Cabinet. This is the cleanest cable management system I've ever seen, but tracing cable paths while re-wiring the system was time consuming.

With the system re-cabled we can power it up and begin testing the hardware. We noticed that the CCU contains rechargeable D size batteries. Yes, we received a donated Cray and batteries were included! They were, of course, very dead. The system had been unplugged for a long time.