What's the coolest machine you've ever worked on?

Question

What's the most exotic, coolest, unique, or interesting machine you've worked on? Most of us work on machines with x86 architectures using some Windows or Linux variant. I'm sure there are those of you out there who are working on or have worked on machines with experimental architecures, or operating systems. Maybe you worked on a machine that has some sigificance in the history of computing. I'd be interested to hear about it. I'm sure others reading SO will as well.

---

EDIT: I appreciate all of you who took some time to talk about their experiences with interesting or unusual machines. I enjoyed reading your answers. Although it wasn't my intent to get nostalgic, I see that theme amongst the responses.

Answer 1

Xbox360 and Nintendo DS.

I wonder if anyone from JPL or Lockheed-Martin browses StackOverflow...

Answer 2

Imlac PDS-1 .. from the days when you had to toggle in the boot loader by hand. My first "personal computer". It had 2 processors - a regular CPU and a vector display processor - which was a novel design for it's time. And a Lightpen!

Mazewar and Spacewar were never so much fun !

Answer 3

Hands down (for me), the Cray Y-MP. Developing and testing Reservoir Simulators in the 80's (FORTRAN), unix operating system (but timeshare for the users).

Runners up (all in the 80's as well): Apollo Domain systems (sort of an early Unix cluster). Also the HP 9000 (one of the faster minicomputers in it's day. Perkin-Elmer (one of the fastest executing FORTRAN compilers of it's time). Honeywell Multix (again, reservoir simulators and other Chemical Engineering programs in FORTRAN).

-R

Answer 4

IBM System z

Actually, the current machines we work on are the most exotic - they're the IBM System z mainframes. And, before you laugh about dead technology, consider that the hardware has come a long way in the last 40 years. It started in 1964 as System/360 (of Mythical Man Month fame), evolved into the 'zSeries', and finally became the System z of today.

Think of a machine with many 'books' of CPUs, each able to hold 56 CPUs (taskable to primary or automatic hot switching backup use) and all I/O offloaded to dedicated processors using fiber optic channels.

The 'z' in System z stands for zero down-time and this is easily achieved by moving work between virtual machines or partitions.

Also able to run Java workloads with the specialized zAAP CPUs and database-intensive work with equally specialized zIIP CPUs, has its own UNIX subsystem (although it's EBCDIC under the covers - ugh).

It can run zLinux (SLES or RHEL) as well - we've had our relatively puny machine (a z800 used for development and testing) running tens of tousands of instances of zLinux concurrently under the control of zVM - that's a big plus in terms of energy usage.

The current crop of 'dinosaurs' that administer z/OS still prefer ISPF (a green screen user interface) but IBM has produced management and health-checking tools using web servers which make the interface a lot less painful for current graduates.

IBM also have many initiatives to train people up in these systems - they realize that most of the current crop of programmers and sysprogs are all approaching retirement age and that's going to leave a void for anyone smart enough to enter the field.

Answer 5

Sinclair ZX Spectrum 48K

It just had that "New Alien Tech" smell that made a bunch of sleepless nights.

Answer 6

I once programmed integral functions on an analog computer, which mostly involved plugging wires in to sockets and adjusting input gains and watching oscilloscopes draw lissajou figures. It was very cool.

Answer 7

Encore Multimax, 32-processor, shared memory architecture running a variant of UNIX. This used NS32032 processors.

Alliant FX/2800, 16-processor, distributed memory architecture also running a variant of UNIX. I believe this used Intel I960 processors.

HP N-series, 8 processor, distributed memory architecture running HP-Ux on PA-RISC 8500.

All of these systems were used for engineering applications, primarily real-time simulation of mechanical systems and finite element analysis.

Answer 8

Not the coolest, but certainly interesting. In the late 80s I worked on the LEC 16 (a mini made by Lockheed), also called the MAC 16. I was working in the middle-east on air traffic control systems and these dumb-asses had been sold this system (last time it was used in the States was in early 70s). The system had core memory! (4k words per 16" * 16" board). The boot 'ROM'' was a 16" * 16" board with 4096 bits whose value was determined by individual diodes (there or not there). To get it to do anything you had to load a boot-loader through the front panel (like the Altair 8080), although it was easy enough to coax it to load a paper tape from a teletype. Storage consisted of usual IBM style tape drives and a mad fixed/removable disk (two platters, one removable 2.somerthing MB on the fixed platter, 1.something MB on the removable). Even in the 80s this was pile of obsolete crud. The main unit (excluding storage) was a 6 foot high steel cabinet. Mad!

For any history buffs out there here is the only thing I could find : http://www.computerhistory.org/brochures/companies.php?alpha=j-l&company=com-42c1686f07480

By the way, the coolest computer I ever had, the Atari 520 when it was launched!

Answer 9

That should be the fridge. It's pretty cool in there.

Answer 10

AS/400, later renamed to iSeries, then System i. I haven't actually worked on one, but I have sat in front of one and typed commands into the terminal.

It's a brilliant architecture that is still very much alive. IBM is putting out new models every year (although the line has now been merged with what was once called the RS/6000 (later called the pSeries)) and is still selling quite a few of them.

One thing I like about it is that it challenges many "conventional" assumptions about computers. For example, a lot of programmers say that one must learn C in order to understand how computers work. However, the AS/400 actually is pretty much the opposite of what you would learn through C. The basic abstractions of C are pointers, files, processes and unstructured bytestreams. The AS/400's operating system (OS/400) doesn't have any of that: it doesn't have pointers, it has something called "files", but those have no resemblance to C files, they are more like SQL rows, also it has logical files which are like SQL views and a lot of other kinds of files like display files (basically GUIs), device files and so on. It has jobs, which are a kind of process abstraction, but it doesn't have seperate address spaces. And all files are structured and strongly typed, like PASCAL records, unlike C unstructured untyped bytestreams.

In fact, not even the CPU has pointers! It uses tags, which are more like object references in a memory-managed VM than C pointers.

Instead of a file system, the OS/400 has a relational database engine embedded at its core. IBM calls it DB2/400 but it doesn't actually share any code or technology with IBMs DB2 product.

The execution model is based on an abstract machine, called the Technology Independent Machine Interface (TIMI). Programs are delivered as MI bytecode and compiled to native object code when they are first executed. The compiler is a privileged system service that can only be called by the kernel; that way the compiler can be trusted and a lot of the security checks that more conventional systems are doing at runtime (usually in hardware, such as memory protection in the MMU or privilege domains in the CPU) can be done at compile time, thus eliminating runtime overhead. (Microsoft Singularity is an Operating System built along similar ideas and they measured a slowdown of more than 30% when going from compiler security to hardware security. Think about that: your PC could be 30% faster (or cheaper, or cooler) if only your operating system were not written in C!)

Originally, the kernel was implemented in Modula-2, with the rest of the OS implemented in PL/MI (a variant of PL/I that compiles to MI bytecode). It has since been rewritten in a more conventional language, mainly because the OS/400 team at IBM where the only people in the world still needing a Modula-2 compiler and they got tired of maintaining both an operating system and a compiler.

Answer 11

My first job was programming CNC punch pressess. Very similar to moisture vaporators. In most respects.

Answer 12

Digital pdp-11 RSX.

One of the first projects I ever worked on was automating a warehouse.

This had little robot trucks carrying stock around under the direction of a pdp-11 programmed in the (awful) CORAL language.

It was just totally kool to see these trucks shuffling around the aisles under the direction of this tiny eight bit computer.

Answer 13

• TX-0 (the first machine with magnetic core memory) Coded up a light-pen scribble program.
• Raytheon RDS-500. The JSX * instruction could be made to do co-routines.
• Greenblatt-Knight Lisp machines, chess machine.
• Intel 8008, on a homemade circuit board, playing a tune on a speaker.
• the Imlac PDS-1, that Scott mentioned.

Answer 14

Current I program industrial robots. These are fun as I get to work in many different domains. The robots are used in so many different industries.

One of my favorites however was an old Martin Marieta 1200 ATE set. We used them in testing Avionics equipment. The programs were loaded from magnetic tape and you have to hand key the boot strap for the tape loader. The electronics were wire wrapped circuit cards. To achieve the repeatability we needed for voltage measurement (micro volt range) the room had to be keep at 68 degrees plus or minus 1 degree. Quite a trick in Miami during the summer.

Answer 15

HP Newbrain...

some weird machine :)

Answer 16

For me, it was a SGI O2 box in 1995. All purple and toaster-looking. I did image analysis on it for a uni project - video in from a camera, processed it, located an object and sent the location to a delphi app on a PC.

Lots of fun. Nothing else like it at all at the time.

Answer 17

On my first job as a professional programmer in 1980, I did software and system support for a network of 6 Xerox Alto computers that were used to publish Dissertation Abstracts. For those not familiar with the Alto, it was created by Xerox Parc and it was the first computer with a bit-mapped monitor and a mouse, among many innovations. The Altos were connected to a 3 MHz Ethernet, yes 3 MHz. Four of the Altos were workstations. One was a fileserver, and the last was a printserver. Every Alto had a removable 2.5 MB drive that was the size of a garbage can lid. The fileserver also had an 80 MB drive. The publishing application was written in BCPL.

In the evening and on the weekends, I could connect to the Xerox network via modem to download updates and participate in the Xerox Alto community. I made one pilgrimage to Xerox Parc. While I was there, I met a young programmer who was sitting in his empty office waiting for furniture and computer to arrive. His name was Charles Simonyi.

Here is a link to the Wikipedia page on the Xerox Alto.

Answer 18

• SPARCstation 20
• Sinclair ZX81
• Robotron KC85/1
• some Convex machine (vector processor)
• DECstation 3000
• microVAX

Answer 19

My iPhone.

Answer 20

Psion EPOC-16 on the Series 3 range and the HC.

Pure small model code, pre-emptively multi-tasking, with a very efficient library in ROM with much of the C standard library functionality.

It rarely crashed, the OS was rock-solid, and the SDK very mature and powerful.

Answer 21

Not quite a "directly programmed" type system, but an internship I had one summer was with a company called Tandem (bought out by Compaq, then bought up by HP 8^D) that ran a Non-Stop Kernel and powered a bunch of ATM machines and the like. They had their own OS and everything to run it. My project involved porting an emulator they for debugging/troubleshooting purposes from Solaris to NT. So while I didn't get to work directly ON the hardware, I learned OODLES that summer about some really interesting low level stuff.

Answer 22

I had a NeXT machine in my office in 1992. I thought it was the coolest thing I'd ever seen. Prior to that, I'd used Unix and Macs. It was great to have the best of both in one box. In many ways, it really was ahead of its time.

Answer 23

Multics, where the unix folks learned about OS's. ;)

Answer 24

I remember NeXTStep running NeXT when I was in college. It was full color and I was blown away with its hefty requirements -32 MB of RAM where most computers only needed 2 or 4 MB RAM.

Also, I remember Windows NT 3.51 running on IBM PowerPC. It was the first time I ever saw a Win3.1 interface showing full content window dragging and it was extremely smooth. It wasn't possible on Intel machines at that time.

Answer 25

I have to say the Commodore Amiga.

It was a machine with a hardware platform and Operating System that, IMHO, was years ahead of it's time. And I also freely admit that nostalgia plays a big part in it! :)

Answer 26

My first programming experience was on a Commodore64, in Simons' BASIC... was developing games and demos... good memories... :)

Next, it was GWBasic/QBasic on a 286, after that, Turbo Pascal on a 386...

and so on... :)

Answer 27

The coolest computer I have used was a NextCube in ~1990. The NextCube had Display Postscript, Objective-C, the first object oriented IDE I had seen with WYSIWYG forms editing, real lightweight threads, a huge 256MB rewritable optical disk and a lot more. I paid for the NextCube out of my own pocket (they were not cheap) and even though Next Inc. went out of business, it was an excellent investment thanks to the things I learned - things like working with lightweight threads which is certainly valuable experience to have in today's world of multi-core processors.

Those who do not remember the NextCube might have heard of the successor to the Next OS - Mac OS X.

Having said that, no computer ever seemed as cool at the time as my first Atari 800 - purchased to play games - but I quickly learned that programming was more fun / challenging / rewarding.

Answer 28

• ZX81 - my first computer

• Xerox LISP station - during my studies at Linköping Institute of Technology

• VAX/VMS cluster - I remember it had some fun DB way to store record structures (for pascal)

• Coding, compiling directly on Palm PDA with OnboardC

• Emulating Jupiter ACE from 80's (valued $150) on a CAVE with a Silicon Graphics InfiniteReality dual rack (valued $2M) (windows version here)

Answer 29

I am currently working on a cluster of PS3s, that's pretty cool. I'm trying to get OpenAtom to run on the SPEs. Also, a hadoop cluster.

Answer 30

Computer Automation minicomputers - LSI2 / LSI4. 16-bit word lengths from the 70s and 80s. They made great test controllers with A/D cards, a real time operating system, and programmed in assembler or FORTRAN.

The LSI2 had core memory, and so bootup time was nearly instantaneous. The most interesting upgrade was retrofitting the LSI2 5 MB disk cartridge with a SCSI interface - the result of which is that you could plug in a 10G SCSI hard drive but the computer OS would address only the first 5 MB.

Answer 31

It's a tossup between the Tektronix 4051 desktop graphics microcomputer and the HP 250 mini.

The Tek was an astounding piece of hardware for 1978. It used Tektronix's bizarre storage-tube graphics technology, which was kind of like a plotter that used electrons instead of ink. It had a built in BASIC interpreter for programming, something like 64K of memory, and, for virtual memory, a streaming-tape-cartridge drive. Yes, not only were programs broken up into segments that you could load dynamically into the shared buffer, they were loaded off of tape. This made UI design a challenge.

It also had programmable function keys (and plastic overlays that you could label your program functions on!) and, for coordinate input, X and Y thumbwheels that positioned the crosshairs on the screen wherever you wanted. It was immensely heavy and immensely expensive. I think when we finally were able to replace the tape drive with an external floppy disk drive (and man, was that an improvement), the price tag was around 10 grand.

The HP 250 was something else entirely. It was a small-office minicomputer system that could support up to five users (barely), four of them using custom HP 250 terminals, and the fifth seated at the big white plastic desk of the administrator's console. The main console was an elegant thing that may or may not have ever won any European design awards, but that was sure the look they were shooting for.

The terminals all featured a row of eight programmable function keys across the bottom of the screen, which could all be labeled dynamically under software control. This, coupled with the forms-management software, could make for some pretty decent user interfaces for 1980. The other software bundled with the 250's OS was HP's IMAGE database, a report writer, and of course to program it all with, HP 250 BASIC.

It was a sweet machine, and fantastically expensive, and the moment the IBM PC came out, Mr. Hewlett and Mr. Packard looked at each other, shrugged, and took Old Yeller out back and shot him in the head. Kind of unfortunate for those software vendors who'd built their business on it. I believe my company actually got a settlement from HP, though I wasn't privy to the details.

And then we switched over to PR1ME, but that's another story. (Our PR1ME sales rep, an otherwise nice guy, went on to become CEO of one of the most hated malware vendors in the industry.)

Answer 32

DEC Alpha AXP. It was a very fast system for its time. It it where I learned UNIX programming. With the 21064 CPU I was solving 64 bit portability decades ago.

Answer 33

My Mac Pro :)

Answer 34

Multics was (is) years ahead of its time.

Answer 35

I'm currently working on a bank note sorter.

Ok, it sounds boring, but...

It's modular, each module contains an ICOP running vxWorks and the number of modules is scalable, the minimum number being four. It has image processing capabilites ranging from pattern matching to full OCR provided by three or more PCs running Win2K Server. The whole machine is controlled by an XP based machine running a .Net2 based GUI. There are two networks within the machine, an ethernet and a CAN. An when it runs, it makes a lot of noise and can process over 2000 notes a minute. And it's big. And, best of all, it has an attachment that looks like a Stargate^*!

Well, it impresses my four year old boy.

Skizz

* sadly, without the interplantary transport wormhole.

Answer 36

ZX spectrum clone in my school.

Answer 37

Indigo O2 by SGI. Today, every PC laughs at the graphics performance but in 1995, having an animated model of a car engine with 200'000 polys render in real time with four light sources was amazing.

Plus the C debugger on the machine allowed to hot-swap code parts - something which we're seeing again in Java today ... when the VM has a good day.

CPUs on that thing had lots of local RAM (as in on the same board as the CPU) with fast links between the CPUs. Hardware monitors would notice when CPU X would hammer the RAM of CPU Y and swap the memory around without telling anyone.

Good stuff. I walked out of the presentation with a very wide grin on my face ;)

Answer 38

Microsoft Surface. Very cool.

Answer 39

A 100% working replica of the Enigma

Answer 40

Data Flow architecture: the NEC upd7281

Associative processors, like the AMD CADM and 99c10 (latter was on an architecture we designed on and built for image processing/OCR);

Nestor Radial Basis Function accelerator (Neural net engine)

Answer 41

It's hard to say what's the coolest or most interesting. Every few years, I stop and think of how amazing the current crop of hardware is, compared to older systems. Using multitouch to interact with an iPhone is very cool. Thinking about how much power is in this MacBook Pro that I'm typing from is amazing, compared to systems just 10, let alone 15 or 20 years ago.

Coolest machine might be one that we never built. Code-named "glass pipe," it was a VLIW processor at Pyramid Technology, back in the late 80s. The assembler syntax for it was interesting. The other interesting machine I worked on was the MasPar MP-1, SIMD machine. And building custom embedded systems, from hardware all the way to the custom software, on 8085 or 68010 systems, cross-compiling and debugging on the VAX, was also fun in my first job.

But modern systems combine many of the elements of these older systems, including VLIW and SIMD. And that in itself amazes me whenever I stop to think about it.

Answer 42

my very first programming contract was for my mom. I wrote a very simple and rudimentary "to do list" using a TI 994a Coolest machine ever! Specially for a 9 year old :)

Answer 43

That's an easy one! I learned assembly bo toggling into a '70's DEC PDP-11. My old college prof is still maintaining and using the box, as seen in his You Tube videos.

Its one experience I'll cherish forever. I can't believe the level of effort it used to take flipping a switch for every individual bit to do something that we do in 1 word of code now!

Answer 44

I would have to say the coolest one is the one I built myself using FPGAs and other chips for a college class; and by built I mean actually built (wirewrapping), not just downloading a design to an existing FPGA board. My own architecture, instruction set, and assembler. You haven't lived until you loaded/run a program and read back the results using a logic analyzer :)

Answer 45

At the time, the very first corporate server I worked on. It was an HP9000 with 16 200Mhz processors, 32GB of RAM, and 1.5TB in a massive EMC RAID array. This was about 11 years ago. The company paid almost $2 Million for the box at the time, and I was partially responsible for DBA work on it. As a first gig out of college, I thought it was incredible.

I laugh now because I just purchased a home server with 8 2.5Ghz cores (2 Xeons), 32 GB of RAM, and 1.5TB of RAIDed space. It smokes that server in everything except parallel disk I/O. Grand total price - ~$2000.

Answer 46

I did my first programming (in BASIC) on a Poly-1 - a custom computer commissioned by the New Zealand government back in 1980 for the education market. Ahead of their time, but couldnt compete with mass-market Apple-IIs that were making their way into the education sphere at the time.

Here is some more information.

Answer 47

Back in college we did a bit with liquid-nitrogen cooled stuff.

It wasn't computers but the question says "machine" and liquid nitrogen is probably a lot cooler than the cooling medium of any of the machines yet mentioned.

Answer 48

I know I'll be beat up for this but here goes. A couple of days ago some poor schlep gets his question closed as he asked if his old macbook was good enough for programming. It was closed for not being programming related...

I enjoy the question posted here as well as the replies however I ask, how is this question acceptable and the other not?

The closed question I have referenced

Answer 49

The Tandem Himalaya was a ridiculously fast mini/mainframe that I had the opportunity to work on back in the 90s for awhile. Still the coolest machine I ever built applications to run on.

Answer 50

I think the one I was most thrilled about at the time was an SGI Indigo2. SGI Indigo2. When I started that gig I was on a Data General MV8000 with graphics sent to a trusty Tektronix 4014. That Indigo was quite an improvement (though IRIX 5.1, not so much). We had actually transitioned from the DG to some generic unix boxes, to an Apollo, then for a while some Suns before finally settling on the SGIs. That was a fun few years.

This was back when VR was The Next Big Thing, and because I worked on an Indigo at work I had an opportunity to help debug a nifty exhibit with 3D goggles at the local science museum that ran off of an indigo.

Looking back, that was an exciting time for hardware. Things were changing at a remarkable clip.

Answer 51

Amstrad CPC

Answer 52

ZX Spectrum. I liked it when I made cute graphics game in BASIC. Never found a language so easy for starters.

Answer 53

Hmm. I remember working on a Tektronix machine with a special monitor. It wasn't bitmapped graphics, it was some other technology where the CRT would persist whatever was painted on the screen. It was particularly suited to playing around with APL, which is what I was doing, because it supported APL's overstrike capability. That is, somme commands in APL were backspace and the second would be overlaid on the first in order to completely show the character.

Put that in your source control and smoke it. ;)

Answer 54

An SGI Altix 450. It's filled with dual-core Intel Itanium processors and a pretty fast interconnect back plane. A tall rack can get up to a teraflop.

Answer 55

Microcontrollers of all kinds, along with the associated peripherals.

Nothing is quite as fun as building something that can interact with the physical world in a meaningful way.

Answer 56

I wish I could answer this question. But the second best would have to be the Commodore Vic 20. The first computer I ever owned, well, my dad owned it. We had a bunch of games on cartridge and tape, but the real fun was typing in the code from books. I was too young to really benefit from it, but I remember hypnotized by how complex it all was.

If only I knew then what I know now. I'd have made Tetris.

Answer 57

The 1GHz Celeron box I use at work.

Everyone else there has P4 machines, but I'm the one that pushes the AJAX code to run as fast as possible, if only for the sake of my own sanity while using it...

Answer 58

CDC Cyber 205. And TI Lisp machines in production.

Answer 59

A special purpose 80 MHz 24 bit processor in an FPGA, for controlling tests of a communication protocol. It was programmed in assembly, which was generated from the test specifications by some Java software I wrote. Tests where timed to sub-microsecond precision.

Answer 60

I didn't work on it. But just looking at my friend's BeBox was pretty cool. Dual processor with CPU load lights right on the case. You always knew how hard it was or wasn't working. Lamentably ahead of its time.

Answer 61

In the picture is a fridge that cools down to 0.3 deg above absolute zero by evaporating helium 3 - a very rare helium isotope more commonly found on the moon. We used it to measure heat capacity of superconductors but now is being used for other general purpose low temperature experiments.

Incidentally the rack on the right is controlled via a GPIB bus and some hackish Delphi code.

Answer 62

My Laptop

Answer 63

My HP 15C scientific calculator. It's older than me yet certain to outlast me, and it's battery life is measured in years. Despite that, it packs high-precision (56-bit BCD) complex number and matrix support, numerical root finding and integration into a pocket-size package with exactly the right proportions. Did I mention that it's fully programmable, and supports subroutines despite having less than half a kilobyte of RAM? It's truly the only electronic device I've ever owned that would be usable and useful if I were stranded on a desert island.

Answer 64

Don Branson wrote: "I remember working on a Tektronix machine with a special monitor. It wasn't bitmapped graphics, it was some other technology where the CRT would persist whatever was painted on the screen."

Storage Tube display. My dad brought home a machine like that when I was about 10, and we had it for years. It was a Tektronix 4051 graphics terminal, with a 1/4" tape drive. I used it to go onto Compuserve a few times. It was really bad for that, because instead of scrolling, text would wrap back to the top and overwrite the contents of the screen. So you had to hit the clear screen key at the bottom of each screen and try to keep up.

I wrote a dumb little BASIC graphics thing on it. The non-pixelated nature of the screen was interesting.

That was the second computer I had access to. The first computer was a Rockwell AIM 65, with a one line, 20 character display and a little cash register-style thermal printer.

Answer 65

My father's old HP calculator programmable through magnetic strips was pretty cool, mainly because I always loved the sound of the little motor that pulls the strip through.

If I recall correctly, in 1994 I saw NeXTSTEP running on a computer consisting of an HP PA-RISC board built into a 15" LCD panel. It was at an investment bank that used NeXTSTEP, presumably a demo unit being evaluated for use on trading desks. They also had a bunch of old Symbolics LISP machines around.

And at my current workplace we have some $100,000 128-channel direct neural interface boxes, which are really cool, but I wouldn't want to have the required surgery to use them.

Answer 66

The Tandem Himalaya running the NSK operating system.

The acronym NSK stands for Non-Stop Kernel, which seems like one of those typical marketing gimmicks but in the case of the Himalaya, this is not an exaggeration, it is a truthful representation of the system. The Himalaya and its operating system are never shut down other than for relocation to a different building.

It has up to 4096 CPU pairs. Each CPU pair is wired up in lockstep. This makes both CPUs execute the same instruction in synchronisation and if they do not come to the same result, the computation is ignored and the pair is shut down. All CPU pairs on the Himalaya are hot swappable, that is they can be replaced under system load. CPU pairs are assigned to CPU pools and they can be moved between pools under system load.

Operating system upgrades are also performed under system load. Memory is hot swappable, too. In fact just about everything in the Himalaya is hot swappable, even the power supplies.

They are fast, too. When I was working on a Himalaya, we were running a decision support system with tens of thousands of database queries running simultaneously against millions of database entries, in fact against the entire database. At the time the Himalaya held the tpc-D benchmark record on that type of query by a very large margin (there was virtually no other system even coming within 1/3 of the score of the Himalaya).

The Himalaya is a true marvel of engineering. If you really need to be sure nothing goes wrong, there is nothing that beats a Himalaya.

Answer 67

Access to an Alliant Computers "minisuper" FX/2800 series (I think, I believe seem to remember it was Intel i960 based) and remote access to a Thinking Computers' Connection Machine. Very cool way for a 16 year-old high school kid to spend a week of his summer vacation in the early 1991 or 1992. Also working on and programming on a DEC (Digital) Alpha 64-bit system using OSF/1 (later Digital Unix) in the mid-1990s was great and I experienced 64-bit portability issues before many of my fellow students had migrated their own development to true native 32-bit (e.g. Linux, Windows NT, Win 95).

Answer 68

An iPad.

Other than that it's been pretty boring since the TRS80 !

Answer 69

HAHA - I think I'll plus one for myself.