The Computing Power in Retrospect

When I was a graduate school student in the 1980s, I participated in an NSF¹ sponsored Supercomputer workshop which gave me a chance to use the famous Cray X-MP Supercomputer.

Cray X-MP Supercomputer was so powerful, it was illegal to export it to the USSR. In fact, it was even prohibited to go near there. Even in neutral Finland, the machine was under 24-hour American watch. Soviet scientists were legally barred from the room, and rumor spoke of ‘kill-switches’ embedded to melt the processor if the machine were ever moved an inch toward the Russian frontier. It was the only computer in history that was so powerful, it had to be house arrested.

It was not a case of unfounded paranoia. Cray’s computing power could help the USSR in these areas

Cryptanalysis: Breaking codes through brute-force or statistical analysis.
Nuclear Simulation: Modeling the fluid dynamics of a thermonuclear explosion.
Signal Processing: Real-time sonar and radar filtering for submarine hunting.

And there I was, using that fabled powerful computer from home, connected through recently popularized phone modem. Except it was no picnic. I had to stay up until 3 a.m. for my turn at UIUC’s Cray X-MP. I was not doing any ground-breaking research. I was doing workshop exercises on how to make the code run faster, using techniques such as loop unrolling. As for why I could not have simply submitted the job and gone to bed, there were reasons, but I forgot the details. I only remember I had to stay up because I was at the bottom of the computing pecking order. I was determined to climb up the ladder and enjoy the computing power under broad daylight in the future.

My wishes came true. Then some.

I was granted 24/7 unrestricted access to machines that are 1,000–10,000 times faster than the Cray. In fact, this fantasy-like reality came true for you, the reader, as well whether you wished for it or not.

As it turns out, the 45-million-dollar-in-today’s-money Cray X-MP was very slow. An Apple Watch on your wrist, for example, is three times faster than the too-powerful-to-export-to-USSR Cray X-MP. Yet, we do not ban its export to Russia because, somehow, no one is doing atomic bomb research on it, presumably. In fact, the Americans let it be manufactured in China.

Metric	Cray X-MP/48 (1984)	Apple Watch Series 10 (2024)	Ratio (Watch / Cray)
Peak FP64	942 MFLOPS	~2–4 GFLOPS (CPU est.)	~3×
Memory	64 MB SRAM	1 GB LPDDR	16×
Storage	~20 GB disk	64 GB flash	3×
Clock	105 MHz (×4 CPUs)	~1.5 GHz (×2 cores)	~14×
Power	~345 kW	<1 W	3×10⁻⁶
Cost (2026 USD)	~$45M	~$420	~10⁻⁵

These days, students carry these computing powerhouses in their pockets, even on their wrists. And yet we do not hear about proliferation of student research on weather prediction or fluid dynamic. Instead, we see all that computing power directed elsewhere.

Perhaps the reason for this underuse of the computation power is not indifference or apathy to technology. Perhaps students simply do not know what is possible, or how to start, with what they already have. This book aims to show the way.

To show you can: There are discoveries, to be attributed to your name, waiting for you. See the section on π and circles below.

A note for the use of AI: Even if you mobilize AI in your research, all of the above benefits still accrue to you because the research does not involve solving problems that have answer key. With AI, your research just moves faster while exploring more directions, all the while reaching higher levels. But make sure to credit AI contribution in your paper. You will learn and accomplish more with AI as long as you use it as an assistant. If you use it as a ghostwriter, you will gain little benefits. This book itself was written with AI’s assistance.

National Science Foundation↩︎

# The Computing Power in Retrospect {#sec-intro-power .unnumbered} When I was a graduate school student in the 1980s, I participated in an NSF^[National Science Foundation] sponsored Supercomputer workshop which gave me a chance to use the famous Cray X-MP Supercomputer. Cray X-MP Supercomputer was so powerful, it was illegal to export it to the USSR. In fact, it was even prohibited to go near there. Even in neutral Finland, the machine was under 24-hour American watch. Soviet scientists were legally barred from the room, and rumor spoke of 'kill-switches' embedded to melt the processor if the machine were ever moved an inch toward the Russian frontier. It was the only computer in history that was so powerful, it had to be house arrested. It was not a case of unfounded paranoia. Cray's computing power could help the USSR in these areas 1. **Cryptanalysis:** Breaking codes through brute-force or statistical analysis. 2. **Nuclear Simulation:** Modeling the fluid dynamics of a thermonuclear explosion. 3. **Signal Processing:** Real-time sonar and radar filtering for submarine hunting. And there I was, using that fabled powerful computer from home, connected through recently popularized phone modem. Except it was no picnic. I had to stay up until 3 a.m. for my turn at UIUC's Cray X-MP. I was not doing any ground-breaking research. I was doing workshop exercises on how to make the code run faster, using techniques such as loop unrolling. As for why I could not have simply submitted the job and gone to bed, there were reasons, but I forgot the details. I only remember I had to stay up because I was at the bottom of the computing pecking order. I was determined to climb up the ladder and enjoy the computing power under broad daylight in the future. My wishes came true. Then some. I was granted 24/7 unrestricted access to machines that are 1,000–10,000 times faster than the Cray. In fact, this fantasy-like reality came true for you, the reader, as well whether you wished for it or not. As it turns out, the 45-million-dollar-in-today's-money Cray X-MP was very slow. An Apple Watch on your wrist, for example, is three times faster than the too-powerful-to-export-to-USSR Cray X-MP. Yet, we do not ban its export to Russia because, somehow, no one is doing atomic bomb research on it, presumably. In fact, the Americans let it be manufactured in China. | Metric | Cray X-MP/48 (1984) | Apple Watch Series 10 (2024) | Ratio (Watch / Cray) | |---|---|---|---| | Peak FP64 | 942 MFLOPS | ~2–4 GFLOPS (CPU est.) | ~3× | | Memory | 64 MB SRAM | 1 GB LPDDR | 16× | | Storage | ~20 GB disk | 64 GB flash | 3× | | Clock | 105 MHz (×4 CPUs) | ~1.5 GHz (×2 cores) | ~14× | | Power | ~345 kW | <1 W | 3×10⁻⁶ | | Cost (2026 USD) | ~$45M | ~$420 | ~10⁻⁵ | These days, students carry these computing powerhouses in their pockets, even on their wrists. And yet we do not hear about proliferation of student research on weather prediction or fluid dynamic. Instead, we see all that computing power directed elsewhere. Perhaps the reason for this underuse of the computation power is not indifference or apathy to technology. Perhaps students simply do not know what is possible, or how to start, with what they already have. This book aims to show the way. 2. **To show you can**: There are discoveries, to be attributed to your name, waiting for you. See [the section on π and circles below](#sec-intro-circle). **A note for the use of AI:** Even if you mobilize AI in your research, all of the above benefits still accrue to you because the research does not involve solving problems that have answer key. With AI, your research just moves faster while exploring more directions, all the while reaching higher levels. But make sure to credit AI contribution in your paper. You will learn and accomplish more with AI as long as you use it as an assistant. If you use it as a ghostwriter, you will gain little benefits. This book itself was written with AI's assistance.