It's such s shame software isn't available. The cray-1, like the colossus are important parts of world history.

I'm more than a little jelous that this guy made this and not me. Oh well, some day I'll own an old altix.

* HN should have more of this, less ify/.ly dotcoms.

http://chrisfenton.com/homebrew-cray-1a/ The detail article with architecture & photos from the building process.

Nice. It's an FPGA board, the Verilog source code for the processor is available. 80 MHz, plenty of vector registers, fits inside one modern FPGA...

Downside: the author couldn't locate any software for Cray yet. Looks like he looked pretty hard and still came up empty. Anyone can help?

Maybe a museum outside the U.S, can help. There is one in my old hometown in Germany. It is the biggest computer museum in the world and they have a cray-1. http://en.hnf.de

I suppose I'd probably read it somewhere, but it reminds me what a huge gap there used to be between supercomputers and anything else in raw horsepower. 80 MHz in 1976! Workstation-class machines only achieved those clock speeds around 1991-92, and the Pentium hit that speed in 1994. Not directly comparable (very different architectures), but still I think a bigger gap than currently exists between top-end consumer equipment and top-end big iron when you look at single CPUs (big iron of course scales up to much higher multiprocessing).

It hasn't been fair to compare single-CPU performance in a long time, machines on the top500 either use commodity x86-64 chips, or (BlueGene) much slower chips with good memory and power efficiency (NEC is a notable exception). But it's worth noting that supercomputer performance consistently outpaces Moore's Law (http://59A2.org/na-slides/Keyes-PetaflopsSeriously-2008.pdf).

What's also very impressive is not just the raw clock speed but the architecture is also deeply pipelined, something that comes in very handy in a vector processor.

> 80 MHz

It runs a bit slower, he's got it pegged at 33 MHz in the article linked elsewhere, the 80 Mhz was the original machine.

Absolutely awesome project, even if not very useful, the guy must have pretty amazing skills to re-implement all this 'on the side'.

Try contacting the Computer History Museum (http://www.computerhistory.org/). They have a fairly large collection of Crays.

The creator already tried this route, as detailed in the the original post that NYC Resistor linked to.

Hmm...didn't follow the links far enough. I find it hard to believe that they don't even have contacts who would know where to find the original source (most likely for a price, but still...).

EDIT: Nevermind... I see:

    They also informed me that apparently SGI destroyed
    Cray’s old software archives before spinning them
    off again in the late 90’s

Well, that's rather...sad

The X-MP will be a tougher nut to crack, but I think the software may be easier to find.

Well - at least one fact can be confirmed now - Cray does scale :) (-down, mostly)