Hacker News new | past | comments | ask | show | jobs | submit login
First USGS supercomputer decommissioning makes way for successors (usgs.gov)
66 points by keepamovin 9 months ago | hide | past | favorite | 46 comments



I really miss the cool-looking computers of the Seymour Cray and Thinking Machines era. The SGI moiré cabinet doors were gorgeous as well.


For the Cray 1 seats and design were an aesthetic / engineering solution. It had to dissipate and have so much power that the seats are basically the power supply. I also like that aesthetic it feels like you could discuss your problems there. Now power and rack units are somewhat standardized. Cooling can somewhat differ in that it could use liquid cooling but most use air .


And the reason for a circular design is so that the wire lengths could all be under 4 feet in order to achieve actual projected clock rates.

I saw the thing in the Computer History Museum in Mountain View. Highly highly recommend the museum to every HN user. I went through it twice, first with the greybeard volunteer guide, and then looped in by myself straightaway.


To offer a second opinion, the Cray-1 was the only thing I found cool about the Computer History Museum in Mountain View. It was otherwise just a like a big website, the history was all dead. In contrast the National Museum of Computing next to Bletchley Had lots of computers turned on, with enthusiast volunteers demonstrating and explaining things. You could punch paper tape and feed it to the Dekatron and see the memory changing as the program ran, or poke at magnetic core memory sheets and someone was right there to discuss all the old tech.


I do miss the old "visible storage" Computer History Museum - where you could walk around and peek into so many great machines. The current version has been musee-ified to death. It might safer from flocks of school kids but a real shame for the rest of us.


They have very strict policies around powering up their devices. I tried to persuade them to look into the original IBM 2260 terminal font design (it is implemented as core memory with missing cores for bits and it resided in the terminal controller). Of course they said no.


My mind was absolutely blown by mercury delay line memory of the UNIVAC 1. The things we did before silicon dynamic memory were works of crazy engineering.

https://en.wikipedia.org/wiki/Delay-line_memory


My favorite are the "glass delay line memory."

https://archive.org/details/TNM_Glass_computer_memories_-_Co...


Is this a Williams tube or a Mellon tube? I have seen ONE or these, and they used to be more numerous than pocket change, at least the ones that no longer worked. Unused they were treated like fine crystal, used like empty beer bottles.

I wanted to hook one up in a tv to see a 1" movie, or maybe a 3/4" movie with the aspect ratio but the refresh rate was really slow, so it was useful as a memory.


I felt the same. It never even occurred to me that you could you use a physical compression wave in a fluid as memory in a real computer.

Here's a delightful snippet from that article btw:

> Alan Turing proposed the use of gin as an ultrasonic delay medium, claiming that it had the necessary acoustic properties.


We should build a computer that uses lasers and the Apollo retro-reflectors as delay-line memory.

It’ll be the largest computer in the solar system. Who’s in?


The European Space Agency is who.

https://www.lisamission.org/


I doubt they’d allow us to use the beams as data carriers…


You might also like the National Cryptologic Museum next to NSA's HQ. The Cray they have isn't as interactive, although you can sit on it for the Sneakers vibes. They have multiple Enigma machines you can play with which I thought was pretty fun


The CHM is cool, but my cousins went to Bletchley, and were very impressed.


Very highly recommended. The Cray was very cool. Had to explain why there was a wire 11.8" long framed. ( Thank you Rt Gen Grace Hopper! ) I walked through twice without a greybeard, and then when I went with a guide... I started smirking when ever he made a mistake... He finally started asking 'Do you have anything to add?' and I only make a few small corrections, until we saw the wire. 11.8". I would think that the Cray would be designed to keep the wires under 3' 11.21" ( 4 nano seconds ) There are PICTURES! https://www.flickr.com/photos/carrierdetect/3599257794/in/ph...


> Cray 1 seats [...] feels like you could discuss your problems there.

A modern LLM supercomputer could take that contemplative bench a step further, to a therapist's couch.

Though, constrained by the modern standard datacenter rack form factor of supercomputers, maybe you'd have to suspend the patient vertically in a cabinet. So long as they're less than 19 inches wide/thick.


Seats 11 comfortably, bit warm during floating point operations. All the crays dissipated an enormous amount of heat. Today's machines, millions of times more powerful, disspipate thousands of times more heat.

What makes a super computer?


I was thinking space heater, but the feeling is the same.


Hard to pack those circular cray machines together though. And people would keep sitting on them.

Check out the Paul Allen auction https://www.christies.com/en/stories/paul-allen-computer-col...


You don't want to pack them too tight for heating/cooling reasons; it was never intended to be 'compact'. Plus, you'd loose the cool factor of having a Cray in your datacenter to show off to visitors.


> Cray in your datacenter

You’ll want it next to your lobby, in the “computer aquarium”.


Given its historical importance and rarity, $5,000–$8,000 for a working Alto seems like a bargain.

Somebody who cares as much as I do, please buy this!


Ah it's got the Altair 8800 !

Cray was fascinating but the Altair changed everything and I'd oddly rather play with a working Altair in a museum than a Cray

https://www.nutsvolts.com/magazine/article/micro_memories_20...

https://www.nutsvolts.com/uploads/wygwam/NV_0103_Driscoll_P4...


You can play with mine.

    docker run -p 8823:8823 rbanffy/altair-mpm


to be fair it does look like a bench seat


Wait, serious question: was it not a bench seat? I've always assumed it was, and a quick Google finds several places claiming they were bench seats, though none of them terribly authoritative.


The "bench" exterior was an aesthetically minded comfortable seat, but inside was much of the cooling system.


Filled with that ... coolant.

Pics or it didn't happen?

The Bench:

https://en.wikipedia.org/wiki/Cray-1#/media/File:Cray-1-p101...


Perfectly functional as a seat.


Yes, you can sit on them. I have.


Have a look at MareNostrum [0] then. It's in the Barcelona Supercomputing Center, in an old de-consecrated church, and looks awesome!

0. https://en.wikipedia.org/wiki/File:MareNostrum_4_supercomput...


Yeah. And to me nothing beats the Thinking Machine's CM-2, here pictured at the MoMA:

https://commons.wikimedia.org/wiki/File:MoMA_Exhibition,_CM-...


That was cool, but there was also a TM @ Pampadu in Paris, and the one at The Army... whew!

https://people.csail.mit.edu/bradley/cm5/corestore/cm5-2.jpg


The perfect modern machine for a 'Das blinken lights' notice.


Denali bears huge Cray logo on top. Is it produced by Cray? What are the specs?


the corporate descendants of Cray, now owned by HPE, still win the contracts to build a lot of supercomputers. but it's not like the days of Seymour Cray himself, hand-designing every aspect of the system. They use off-the-shelf parts mainly, it's essentially building out a highly-specialized datacenter. the main technology that Cray contributes is the specialized interconnect.


Where do these end up? I'm sure there are some places that would be thrilled to get this. Or is the tech so old that it's not viable for research?


I think “not viable for research” is not a thing, haha. At least, if the ratio of compute to communication is right, we can test our algorithms on old hardware using smaller problems.

Edit:

The SuiteSparse collection (formerly UF matrix collection) is really great for this. You can download matrices with their year and some context about their application. The ones smart people were trying to figure out how to solve 20 years ago on clusters. And then you can marvel at how your laptop smashes them using PARDISO or some other direct sparse solver. Or you can try to figure out a clever way to solve them with 10 year old hardware.


ime, They usually get sold, either as full packages or per rack. There are refurbishers who tend to handle the piecewise extraction of all worthwhile components. See also, the recent auctions of Cheyenne

They aren't usually re-used as full compute racks because the infra required (cooling, power) is non trivial.


For the past couple of decades at least, supercomputers are decommissioned when newer HW becomes so much more power efficient that it's cheaper to buy new hardware than keeping the old one running. Usually that lifespan is in the range of five years or so.

Of course, if you have free power (or say, the power bill is paid by a different department), sure go wild and keep that old darling running.


If my local Craigslist is any indicator the parts will be stripped down and resold. There were some very cool and very expensive, when new, SGI machines I found on my local Craigslist page.

I quickly closed the browser tab when I found out about the power requirements and noise.

These computers will probably not make it to CL but eBay and niche auctions might sell them.


i'm surprised the usgs didn't have any supercomputers in the 70s


Consider the top super computers back at the start of Top500 - https://www.top500.org/lists/top500/list/1993/06/ ... which are nearly all various configurations of a CM-5 or Cray/HPE.

I am still amazed at how far we've come. A Raspberry Pi5 is 31.4 Gflops at 8.6W ( https://web.eece.maine.edu/~vweaver/group/green_machines.htm... ). That would have put it on the first page 1993.

(aside: my favorite entry on there is still https://www.top500.org/system/173736/ )


agreed; i wrote the beowulf faq in the late 90s (cf. https://webhome.phy.duke.edu/~rgb/Beowulf/beowulf_book/beowu...)

apple really screwed over the virginia tech folks


They probably borrowed time on NASAs SGI Cray3. They do a lot of geo-modeling, the oil companies also do that.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: