Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Supercomputing

How a Cray-1 Supercomputer Compares to a Raspberry Pi (roylongbottom.org.uk) 145

Roy Longbottom worked for the U.K. covernment's Central Computer Agency from 1960 to 1993, and "from 1972 to 2022 I produced and ran computer benchmarking and stress testing programs..." Known as the official design authority for the Whetstone benchmark), Longbottom writes that "In 2019 (aged 84), I was recruited as a voluntary member of Raspberry Pi pre-release Alpha testing team."

And this week — now at age 87 — Longbottom has created a web page titled "Cray 1 supercomputer performance comparisons with home computers, phones and tablets." And one statistic really captures the impact of our decades of technological progress.

"In 1978, the Cray 1 supercomputer cost $7 Million, weighed 10,500 pounds and had a 115 kilowatt power supply. It was, by far, the fastest computer in the world. The Raspberry Pi costs around $70 (CPU board, case, power supply, SD card), weighs a few ounces, uses a 5 watt power supply and is more than 4.5 times faster than the Cray 1."


Thanks to long-time Slashdot reader bobdevine for sharing the link.
This discussion has been archived. No new comments can be posted.

How a Cray-1 Supercomputer Compares to a Raspberry Pi

Comments Filter:
  • ...input latency though?
  • I'm old enough... (Score:5, Insightful)

    by Baron_Yam ( 643147 ) on Sunday December 31, 2023 @11:01AM (#64119255)

    To me, 'Cray' was synonymous with 'ultra fast supercomputer'. It was the go-to thing to express computing power that you, as an individual, could never hope to access never mind own.

    Contrasting it with a Pi and having it lose by a large margin is somehow extremely disappointing. I have a Pi in my home that's just lying there doing nothing, a toy that I'm not currently interested in playing with.

    • by dgatwood ( 11270 )
      In terms of floating-point performance, the pi is more like 100x as fast. I’m actually surprised the number is that low. Computers have gotten a lot faster over the last few decades. :-D
      • I'm not sure about the Cray 1, but the reason they were so fast was they were pipeline vector computers.

        Where GPUs, then or now, operate math on a single number at a time, these operated on an array of numbers, like an assembly line, doing the same operation in broken up stages. Hence the time to pop a new finished calculation (like a car) was the time it took for the longest stage.

        That this little whizzing new pocket computer that's nobody's definition of a number cruncher can outperform that 4.5x is quit

        • Almost all supercomputers and computers, really need you to optimize your code and algorithms specially for best performance. What helped Cray was that the vector computing really helped out scientific Fortran programs without too many changes. The same programs had difficulty adapting to parallel computers. Many modern programs and programming styles don't work well on vector processors because there's so much data and it's spread out, bloatware means performance requires giant caches more than it needs fa

        • by tlhIngan ( 30335 )

          I'm not sure about the Cray 1, but the reason they were so fast was they were pipeline vector computers.

          Where GPUs, then or now, operate math on a single number at a time, these operated on an array of numbers, like an assembly line, doing the same operation in broken up stages. Hence the time to pop a new finished calculation (like a car) was the time it took for the longest stage.

          Most architectures today include vector instructions. They go by various names, but almost all are SIMD - single instruction mu

    • Re:I'm old enough... (Score:5, Informative)

      by DarkVader ( 121278 ) on Sunday December 31, 2023 @11:20AM (#64119295)

      The Cray is still better in terms of seating capacity.

      (Yes, I have actually sat on a Cray once.)

      • Re: (Score:3, Informative)

        by Anonymous Coward

        Please do not sit here - https://www.clustered-pi.com/a... [clustered-pi.com]

      • In 1981 I was part of a 10-person group of science fair winners who got to tour an air force computer facility that featured a Cray 1. We were all allowed to sit on the bench so that we could say we had. Also, Robert Redford sits on the bench in the movie Sneakers.
    • It is sort of ironic how things change. Even the OS has a lot more capabilities. UNICOS was good enough for a few tasks, but it pales to what is offered with Linux, QNX, and other modern UNIX/UNIX-like variants, especially with modern scripting languages.

    • For me, that moment happened when Unisys created their first "mainframe on a chip." I was astounded to see a computer under somebody's desk running the same software as the multi-million-dollar mainframe, and running it faster. At that moment, I realized that computing was about to change drastically. Once businesses figured out that they didn't need to spend those millions, they certainly would not.

      I was partly wrong, they still have to spend millions, now it's just on software instead of hardware.

  • Price per pound (Score:5, Insightful)

    by nastyphil ( 111738 ) on Sunday December 31, 2023 @11:04AM (#64119259)

    is more or less constant..

  • by Rosco P. Coltrane ( 209368 ) on Sunday December 31, 2023 @11:04AM (#64119261)

    It's sexy.

    Back in the days, I remember drooling over pictures of that thing, reading about how they managed to squeeze every little bit of performance out of that thing (if you don't know, read up on why it has this circular shape, it's not random).

    It was the future. It was exciting!

    The RaspPi is impressive. But ultimately, it's the quintessential utilitarian cheap commodity hardware thing. It's as impressive and exciting as a Chinese or Indian space rocket versus the Apollo program: undeniably worthy of praise, but nothing to stay up overnight glued to your black-and-white TV set over.

    • by znrt ( 2424692 )

      the cray made some sophisticated projects possible for the select few that could afford a time slice.
      the raspberry pi made some sophisticated projects possible for almost everyone.

      being cheap and utilitarian is nothing to scoff at, it's just how it rocks, and why it rocks. i'd say the pi is as sexy as can be.

    • by sjames ( 1099 )

      In some ways yes, in others, no.

      Seeing the Cray 1 was really cool. Reading about how it managed to be so fast was fascinating. There existed a possibility you might actually see one in person or even sit on one, depending on your profession and employment or student status. You might even get to run a program on one. But that was only likely for a few. Most would never see one in person, much less touch one or run a program on one.

      OTOH, I have several Raspberry PIs dedicated to doing what I an I alone want

      • One of the things I built using a pi was a grandfather clock with a full octave of chimes. The thing is massively overpowered, but for 35 bucks, why not. And it does use some CPU rendering the slide show behind the hands of the clock on the monitor. I spent I think more on the copper pipe to build the chimes. I've built probably a dozen things with pi's now. I think I am going to try making something with the new RP2040's next. At a buck for the microcontroller and 5 to have someone else put it on a pcb, it
        • by sjames ( 1099 )

          The 2040 does look nice. We've come a long way since the Atmega. Though I do still enjoy the way an Atmega will tolerate nearly anything, the ARM based controllers pack a lot of capability at a good price.

    • It's sexy.

      WTF are you talking about. There's 4 whole movies about having sex with pies, no one has ever made a movie about someone humping a Cray 1.

  • by Eunomion ( 8640039 ) on Sunday December 31, 2023 @11:11AM (#64119285)
    Airlines are crammed like slave ships, and classic books are being edited and censored by revenue algorithms, but my phone oh-so-helpfully reminds me to buy laundry detergent. What a shit future.
    • by znrt ( 2424692 )

      Airlines are crammed like slave ships,

      that's because companies are cheap and users dumb enough to let them get away with it. actually, often users are part of the problem. none of it has nothing to do with cpu power or software efficiency.

      and classic books are being edited and censored by revenue algorithms,

      that's because ... wait for it: companies are cheap and users dumb enough to let them get away with it. actually, often users are part of the problem. none of it has nothing to do with cpu power or software efficiency.

      but my phone oh-so-helpfully reminds me to buy laundry detergent.

      well, it can do way more than your shopping list, and development kits abound. what's holding

      • Consumer-blaming is one of the lazier forms of victim-blaming. Excuse the ordinary human for believing that the options presented to them by industry are an accurate representation of what's practical.

        Expecting billions of people to spontaneously self-organize into a universally self-aware, highly imaginative consumer union that efficiently pressures industry to meet its potential is clearly reasonable, while expecting genius-level professionals and tech entrepreneurs who were practically born into exec
        • Whatever else the development of the IT industry etc has reminded us of, it has reminded us that the market will find and sell alternatives if they exist. If there was an easy solution to the slave class experience, someone would have found it and introduced it because the slaves would have been attracted to such a solution. That they haven't strongly suggests that there isn't a solution.

          The fact that there is the option of paying more for a more civilised experience also suggests that the slaves prefer che

          • The alternatives were already clear decades ago, and the large aircraft manufacturers (all two of them) had long-term plans to move in radically different directions from current designs to address it. But the companies that run airlines have very different priorities from airline passengers, and passengers aren't the ones who buy the planes, so the manufacturers abandoned those plans and built planes to meet the needs of airlines rather than passengers.

            Hence, the slave ship experience when the actual s
            • News to me. Want to tell me more - got a link?

              • Boeing "blended wing." Or the supersonics that were supposed to come on line in the '90s. Or the retirement of the Concorde rather than investing in making it more affordable. Etc.
                • Lol, I worked on the Sonic Cruiser project in the 90's. I still have a few pre-production samples of some of the (proposed) hardware. They used to have a big mockup of it in the lobby of one of the buildings down in Renton.

                  It was a dead dog on the runway from the start because it was a solution for a problem that didn't exist.

                  No viable 'audience' or customer base existed that would pay an extra couple of thousand dollars to get to Europe an hour earlier.

                  I didn't care; Boeing paid me handsomely to do whateve

                  • Okay, but you're begging the question. Boeing didn't do jack shit to bring that cost down, because passengers are completely ignored by the large aircraft manufacturing "industry" (hard quotes). And for all airlines care about speed, they'd still be running zeppelins. Nor do they care about spaciousness or comfort, totally abandoning the blended wing plans.
                • Blended wing doesn't seem to offer any particular improvement for cattle class, whilst supersonic planes seem irrelevant to the debate unless their tickets get to cost about the same as our current prices. Or am I missing something?

                  • It was supposed to be more spacious, and they were talking about high subsonic speed, substantially faster than current Mach 0.75. Or I might be confusing it with yet another long-planned project they abandoned out of laziness.
        • by znrt ( 2424692 )

          Consumer-blaming is one of the lazier forms of victim-blaming.

          so who is to blame for you underutilizing the (vastly superior than a 70s supercomputer) computing potential of your phone again?

          and don't get me wrong, big companies are shit, in great part because capitalism is shit. but this has nothing to do with computing power.

          re the current state of the art of software design and writing in particular (and that is a complex and multifaceted reality, but lets just grossly simplify) ... you would have a very good point there. sw nowadays is hugely anti-economical by de

          • "so who is to blame for you underutilizing the (vastly superior than a 70s supercomputer) computing potential of your phone again?"

            Clearly it's the fault of the general public, who can barely do high school algebra, for not getting advanced IT degrees that might let them fully utilize technology. And it's my fault for broadly studying sciences rather than asking two-digit-IQ'd hereditary corporate overlords which narrow engineering field they want their drones focusing on this week.

            I'm saying IT stra

    • I think the thing is you're really only noticing the tech that you directly interact with on a day-to-day basis. There's a ton of tech you don't think about because you're a few levels away abstracted from it. The most obvious example is advancements in vaccine tact that have largely gotten covid under control. But there's huge amounts of new tech that made it possible to maintain our oil supply and therefore our entire economic system as well as maintaining our food supply despite the massive droughts caus
      • That's fair. It is reasonable to say, however, that the benefits of IT have been severely overstated, and in some cases have been perversely incentivized to undermine consumer benefit.
    • Thought I stumbled on a classic Andy Rooney 60 Minutes rant there. https://content.time.com/time/... [time.com]

      Anyhow back to classic supercomputers.

      • I'm just saying. Why don't people understand the difference between information and knowledge, and between computation and capability?

        Slide rules and room computers: Moon landings and supersonic jets everywhere. Supercomputer phones: Ad bombardment and fuckall else?
    • Airlines are crammed like slave ships

      Don't know what you're talking about. The last flight I took I had several cubic meters to myself and a complete lay flat bed on the plane and the ticket was cheaper than an shitty economy class seat from the 90s. Bonus points that I got to chose when I wanted my meal had a huge choice of food and alcohol, a dedicated lounge with restaurant serving free food, showers, and a quite room to sleep in.

      Airlines are offering you what you want. You just don't want to pay for it. You are making your own future shit,

      • Then I don't know what world you're inhabiting. Airlines are a nightmare. Airports are a nightmare before the nightmare, despite all the pre-this and pre-that booking shenanigans. And I'm sure you can buy books advertised as an unedited, because the revenue algorithm that edited them predicted that calling them unedited would increase revenue. Come on.
    • by AmiMoJo ( 196126 )

      Those books were always censored.

      For example, Charlie and the Chocolate Factory originally had the umpa-lumpas as an African tribe... That Wonka transported... The publisher made him charge them before the first edition could be printed.

      • Editorial demands prior to original publishing are hardly analogous to erasing the history of generations-old classic literature for current markets. And mind you I'm not just talking about the usual edition process, where they include commentary about the changes they made: They're not even telling people what they did now, and are trying to get old copies removed from libraries.
  • by AlanObject ( 3603453 ) on Sunday December 31, 2023 @11:45AM (#64119361)

    There is more to it than the relative benchmarks

    When I was a kid in the 70s I had various jobs programing the CDC 6000/7000 computers at LBL in Berkeley. Hundreds if not thousands of researchers used those computers which ran thousands of jobs every day, submitted on punch card decks either locally or from remote locations.

    Those computers had not only their own behemoth power supplies, but their own support staff to keep them running. Trained operators would fetch 1/2 inch tapes from the library and mount them on those refrigerator-sized tape drives. Dedicated staff wrote and updated the operating systems and even compilers. And there were at least two full-time engineers that had an office where they repaired things that broke.

    They even had accounting staff. All this was paid for by intra-agency billing, where each job charged for each millisecond of processor time and each "kilo-core-second" used. And a purchasing department to order and stock the endless boxes of line-printer and teletype paper, punched tape, hollerith cards, magnetic tape, and so on.

    It didn't stop there. All of this was housed in a special room behind locked doors with special architecture to route all those 30-70 pound cables under the floor. You had to have special privilege to go in there without escort. So its own police force, effectively. Over the hill and Livermore Labs where they did a lot of top secret work (still do) there was a police force in actuality.

    To the general public, the main staff that casually wafted in and out of that room looked more like a priesthood than anything to do with engineering.

    Aside from computing power, there is no way to compare all that to the typical 5W palm-top disposable rig. A lot of what I describe still exists around large operations (where I rent space in a co-hosting location is like that) but it is more that the culture around them is all different.

  • Anything you buy today will be obsolete tomorrow. We know that.

    Any computer from today will be surpassed in speed and efficiency by something that comes out a year from now. The thing is, though, if you need that computation power now, you can't just wait another year.

    Is a computer of today faster than from from 45 years ago? Umm.... duh! Factor 4.5? Yeah, why not. Hardly a surprise. But you had that 45 years ago and I only got it today.

    Depending on what's your plan, 45 years may be a really, really long ti

    • The performance improvement on the ground has perceptibly slowed down of late. It used to be that every generation gave a massive leap ahead in performance, now we are seeing much smaller percentage improvements as we encounter various practical limitations.

      Factor 4.5? Yeah, why not. Hardly a surprise. But you had that 45 years ago and I only got it today.

      Hmm, now I wonder how much faster a modern machine/cluster would be if you spent the $10M a Cray 1 system could cost... even if you don't adjust for inflation. I think the impressive (if still not surprising) part is that a cute little toy computer of to

    • by serviscope_minor ( 664417 ) on Sunday December 31, 2023 @12:19PM (#64119417) Journal

      Anything you buy today will be obsolete tomorrow. We know that.

      Not as much as it used to be. I'm writing this from a nearly 14 year old laptop (thinkpad W510) with a late life upgrade of 32G of RAM and 1TB flash drive. It's not a speed demon, but it still kicks the crap out of a surprising number of laptops in terms of RAM and storage (it'll hold 16TB easily, probably more). Haha crappy new macbook airs max out at 24G RAM lolz.

      But 14 years ago (2010), 14 years prior was 1996. Top of the line PC desktop was a pentium pro 200MHz, and supported... wait what 64G RAM? WTF? That's some exotic shit there, and apparently the super high end 4 socket server motherboards could hold 4 with late generation SIMMs, though most even high end desktops wouldn't go anywhere near that (16 sockets!!). Anyhow much more expensive at that point than even this laptop.

      Either way, a 1996 computer in 2010 was unusable: far too slow, and low on RAM.

      So not obsolete tomorrow, maybe next week.

      • I'm sitting here on a desktop running an i7 from 2014 (IIRC). Looking at 22" screens that are close to 20 years old now. They still do their job pretty well and yes, back then they all were insanely expensive.

        The thing is, and here the comparison between Cray and RasPi falls flat, upgrading to a modern system with similar specs will not allow me to save on power.

        We have reached "good enough" a long time ago with PCs. I would say no later than 2010. 2015 if you're into gaming (and don't hunt the elusive 8k o

      • So not obsolete tomorrow, maybe next week.

        Funny coming from someone who did a late life upgrade.

        But 14 years ago (2010), 14 years prior was 1996. Top of the line PC desktop was a pentium pro 200MHz, and supported... wait what 64G RAM? WTF? That's some exotic shit there

        You're mixing desktop and workstations/servers. Pentium Pros were not for typical desktops. If you focus on the high end of CPU offerings you'll find they always supported quite incredible RAM. Current Ryzen 7000 Threadrippers support 1TB. Intel Xeon E5v3 supported 768GB 10 years ago.

    • by jmccue ( 834797 )

      Anything you buy today will be obsolete tomorrow. We know that.

      Here is the difference, something that cost millions of USD decades ago you can get better performance for something now for ~70USD. I have not seen one of these for a while, here we go:

      In 1974, you could buy a low-mid range US made auto for about 3700 USD. That equates to about 23,000 USD now. But a similar new auto costs around 30,000 USD (maybe a lot more). Performance wise, not much of a difference. So yes, computers have improved greatly over the years. But in Autos, there were no real improvemen

      • Cars in '74 sucked performance wise, you should have gone with 1969, but even then you would have been wrong; Car are coming of the dealer's lot with 500 and 600 horsepower and street legal cars have ran the quarter in 5.77 sec at 260 mph.

        • More importantly, cars today require less service, use less fuel, pollute less, and are far less likely to kill you in a number of common accident scenarios. It was rare for a 1974 engine to go 100,000 miles without needing to be rebuilt, even if you religiously changed the oil (every 3,000 miles) and otherwise kept it up. You would almost certainly replace a water pump and alternator in that first 100K miles. I am now on my third car that got to 200K miles without needing any major engine maintenance at
      • Uh... yes we are. If I was to get into an accident, please put me in a contemporary car instead of a 1974 death trap. There's a reason modern cars are far heavier than their pre-1990 counterparts. Most of it has to do with safety.

        Also, a lot of handling improvements have happened since, also with considerable effect on safety. Hydraulic steering, anti-blocking brakes, traction control, ... these things were either completely unavailable in the 1970s or a hard to get feature in some of the top of the line lu

    • Is a computer of today faster than from from 45 years ago? Umm.... duh! Factor 4.5? Yeah, why not. Hardly a surprise. But you had that 45 years ago and I only got it today.

      That 4.5 quote is about a tiny computer from 10 years ago (and it even says so i the submission). So you got far more than that after 35 years, smarty pants. Or to quote from TFA: "The first PC to reach the average Cray 1 Livermore Loops score is indicated as a 1994 100 MHz Pentium."

  • ... in another 20 years? Maybe they'll have a chip 1000x as powerful embedded in their brains, communicating with the mothership via Musk's satellites or something?
    • by HiThere ( 15173 )

      Pretty sure that means a drastic change in architecture, moreso than everything on punch cards and paper tape (1950's) to everything stored on the disk. If nothing else, the speed of light pretty much guarantees that. You need to switch to essentially parallel to get that kind of speedup, and some things are resistant to parallelization to that degree.

      Note that the GPUs depend on the stuff they handle being processed the same way when it's split into lots of pieces. But not all problems are amenable to

  • by account_deleted ( 4530225 ) on Sunday December 31, 2023 @01:30PM (#64119559)
    Comment removed based on user account deletion
    • I also remember, once, reading about a senior Air Force officer named Turnipseed. Strange names are everywhere, and sometimes you have to wonder where the names came from. In the DiscWorld books, the forensic expert of the Ankh-Morpork City Watch is a dwarf named Sergeant Cheery Littlebottom [wikipedia.org]. Unlike almost all female dwarves, she doesn't hide her gender, wearing leather skirts and high heeled boots. She doesn't, however, shave off her beard.
      • by cusco ( 717999 )

        My father-in-law was a teacher in the highlands of Peru for 32 years. The Inca never had last names so they were assigned them by the Spanish barbarians when they were baptized, and the Spaniards who went to Peru were **NOT** nice guys. A few of the last names the he encountered included:

        Cusihaullpa - Happy Chicken
        Huallpasuya - Chicken thief
        Sucasaca - Slang at the time for 'cuckold'
        Sacacaca - Take out the shit
        And the all-time worst:
        Alcoruntu - Dog balls

      • by hawk ( 1151 )

        my grandmother's oncologist (?) for terminal cancer was the unfortunately named "Dr. Posthumous" . . .

        hawk

  • ... run doom? If it's even 20% of the performance of the RPi it should be up to the task. I don't seem to be the only person who has asked the question, but I don't see a definitive answer.
  • In 1978, the Cray 1 supercomputer cost $7 Million, weighed 10,500 pounds and had a 115 kilowatt power supply. It was, by far, the fastest computer in the world.

    The Raspberry Pi costs around $70 (CPU board, case, power supply, SD card), weighs a few ounces, uses a 5 watt power supply and is more than 4.5 times faster than the Cray 1.
  • Removed encoding/decoding, too.
  • From Byte Magazine, April 1981, Unclassified Ads, Page 414:
    https://archive.org/details/BY... [archive.org]

    MUST SELL: One slighty used Cray-1 -- too big for my apartment. Hardware included: light-dimmer interface, toaster interface, one home-brew 64-bit parallel I/O port with Nixie-tub indicators, coat hangers, extra buffing powder for instruction buffers, one box of bootstraps. Software included: CAL assembler on thirteen cassette tapes in Kansas City format, Morse-code trainer, tic-tac-toe game, 8080 emulator. Price

  • by fahrbot-bot ( 874524 ) on Sunday December 31, 2023 @03:04PM (#64119771)

    According to the tables on TFA, the clock speed of the Cray 1 was 80 MHz and the Raspberry Pi 700 - 1,800 MHz, so, unless the benchmarks account for clock speed differences, if the Cray 1 was (could be) overclocked to match, it looks like it would be faster and that goes of several (many?) of the other systems listed.

  • Finally an article that is actually news for nerds..
  • where are the jobless people? 10^5 improvement in computing/automation should have made a ton of jobless accountants and such

Keep up the good work! But please don't ask me to help.

Working...