Trying to Build a PDP-11 Supercomputer - Part 1

[@tommythorn]

“Fully DEC’ed out” very punny

[@marcwolf60]

Your PDP 11/44's bus will go PLAID!!!! after hitting Ludicrious speed. :)

[@Alexis_du_60]

"IC" what you did there... Sorry, that's just me trying to make a electronics pun.. 😂

[@TomFynn]

DEC the halls in ones and zeros...

[@ChristopherHailey]

Lol I thought the same thing

[@SenileOtaku]

I noticed that one.

[@TheAtomstrike]

Hi. I'm from Moscow. In 1989 I had to repair a Bulgarian clone of microVax by cleaning and contacts and tightening nuts on the power supply busbars. You reminded me of this with your similar repair.

[@radiotimofej]

А я из Новосибирска

[@grishka212]

А я из Питера

[@Aeduo]

Neato. I have an MK-90 calculator which is a soviet LSI PDP-11 clone, which I've done some assembly coding with in an emulator. но я из америки. :p

[@TheRealRaddicalReggie-o9l]

Why do Russians steal everything….. and then make it worse?

[@SteveGordon43]

@@Aeduo Which LSI PDP-11 model is MK-90 a clone of?

[@williamhoward7121]

I can just imagine Tron sitting in this abandoned memory section for the past 50 years when suddenly everything lights up around him and section by section his components come back to life! I wonder who the MCP is in this case?

[@theDarthFader42]

7:25 "But we're gonna try". And that's why I love this channel. This is going to be a fantastic project. I can't wait to see where this goes.

[@UsagiElectric]

Thank you so much! And heck yeah we're gonna try!

[@ChrisDreher]

20:14 "If your computer isn't heavy enough to murder you, can you really call it computing?" 😅😅😅

[@denniseldridge2936]

If any quote begged to be put on a tee shirt...

[@SenileOtaku]

Sounds like our old pSeries 690 / Regatta systems. One time trying to swivel one around, one caster dropped into a cable opening in the raised floor. Luckily I was there to help push it back before it tipped over.

[@ibanezleftyclub]

This needs to be on a T-shirt with some old school PDP11 graphics.

[@UsagiElectric]

This could make a fun t-shirt for sure!

[@poofygoof]

"my computer has more hitpoints than yours"

[@computeraidedworld1148]

Man, I love your unwavering positivity. It’s incredible.

[@jazzathoth]

Usagi here running ancient minicomputers like a pro, but getting confused by a linux shell. 😂

[@mrt1r]

It's probably too modern for him 😂

[@RichardBetel]

@@mrt1rUnix was developed on the pdp11 so I’m gonna chuckle at the joke and say nope!

[@ChristopherHailey]

Unix actually started in the PDP 7 and then the Interdata 7/16 and the PDP-11

[@FryAndLeelaAndBender]

Remember that he had used a lot of different operating systems for all of his computer collection, so, yes, it is really difficult to remember all the commands, syntaxes and operations for each hardware... Very impressive indeed 😅😅😅

[@ChristopherHailey]

@@FryAndLeelaAndBender I remember when each machine had it's own OS and you had to learn everything else too. If he is going to go with RSX-11M he's going to have to learn PIP commands (!) He really does have an amazing range of knowledge

[@brianatbtacprod1989]

I worked a lot on equipment from this era, and the only real thing I don't miss about it is the cacophony of fan whine. There was a reason all this stuff went in a different room/floor. I actually know people who have hearing problems because they spent hours in rooms with that whine. The only thing worse is the printer room. Very cool to see it again. Brings back lots of good memories.Thanks

[@Derundurel]

Everything I have encountered from this era and quite a while afterwards runs the fans at full speed all of the time, which makes it much worse. Even unoccupied slots in the backplane get full cooling for the worst case conditions continuously.

[@Brian-L]

After several years working in data centers as a field engineer for HP, I believe my substantial hearing loss and almost debilitating tinnitus is a result of my work. I know many data centers now require hearing protection, it’s just too bad we didn’t figure this out 20-30 years ago.

[@robertg2129]

I have been following you for a couple years now. I very much enjoy your quest to resurrect old iron from back in the day when I was active building my own system. Eventually my DEC system evolved to a PDP 11/44 with 2 Fugitsu 80MB drives, with a RL02 and a Kenedy 9000 1/2" tape drive for backup. I ran RT11 with a multi user extension called TSX Plus. I wrote my own software in a language called DBL ( a cross between Cobal and Basic with powerful 7 key ISAM file capabilities) The system had 6 VT100's, 3 serial printers, and a Printronics P300 300LPM line printer with a graphics board. I have photos but I don't know how to attach them. All that old iron is long gone now. I wish I would have had the room to hang onto all of it. I am 77 now and have forgoten most of what I knew back then. Your videos bring back memories of all the fun and great satisfaction I had building and programming my great PDP11 system. Let me know if there is some way of sending photos to you.

[@tronmcconnell4465]

As a former system manager for out department's computers, which included a PDP-11/44 running Unix V7, and a MicroVAX 3800 running OpenVMS V5.5-?, you can be sure I'll be watching your efforts with the PDP-11/44 with great interest. Looking forward to the next installment.

[@sarai3055]

Love your program very much. Remind me of the 1960s when I had my first experience with a computer that filled the trailer 30 * 10. I was walking by one day and picked up a circuit board laying outside filled with transistors honor nice little metal cans. The guy inside the trailer go can you hand me that. I said Mr what are you doing with all these transistors. He looked at me and asked how I knew about transistor. I told him I build radios for my friends all the time. He said what are you 10 about 10. I said no I'm nine. He asked me to step in the trailer and I said no. He said I want to show you what the transistors are for. I put caution to the wind and said okay. I walked in to this God awful noise of fans the entire trailer with filled with racks filled boards fill the transistors. Thousands and thousands. I said is this a computer. He said yes kind of. He said it was a automation computer. Later on I showed some of my radios to him. He handed me aboard handed me a bunch of transistors and said see if you can tell me which ones are bad and replace them. Then bring them back in the next day or so. Needless to say I did needless to say he was impressed and needless to say my father the teacher said what. My father was a bit of a douche.anyways I refined my skills at soldering to the point where I can pull the board pull every transistor off of it and put it back in within 15 to 20 minutes. He said I'd offer you a job but you're not and 9 a half years old. Anyways I went on for a while you guys have fun with all your new toys I got to go build a computer for a neighbor take care.

[@MrTherende]

I don't know if you feel like hunting (wild goose chase??), but I worked with Kevin J. Kleese at Kodak. He ran a VAX-11/750 (VMS) with an FPS processor back in the 1980's. I don't recall which FPS model he used other than it was LARGE. His last patent was in 2002 so who knows if he is even alive now. When I knew Kevin, he lived in Rochester, NY and may have moved to Hamlin, NY. He was a bit of a rail-buff too.

[@gertk2303]

120 Amps (aka welding currents) at 5 Volt means every milliOhm of contact resistance will take its toll on those connections. Great to see that solved. These units must also be excellent space heaters 😁

[@ChristopherHailey]

The amount of AC we used to have in our labs in those days was crazy

[@monad_tcp]

that's why the busbar are huge, and are bolted

[@UsagiElectric]

Yup! It's practically an arc welder to power the 7400 logic, lol. The FPS-3000 is even crazier, it's power supply is delivering 5V at 200A!

[@SteelBeastRT]

​@@UsagiElectricreally weird that they didn't try to separate the power lines in like groups of 25 Amps or something. It's only 1kW of total power from 5V bus in the end.

[@binarydinosaurs]

Man, I'm right back to the 80s here. Seeing RT11 V5 boot on an 11/44 for the first time in decades. Awesome. Cheers Dave!

[@sdrc92126]

There was a Gould vector processor on the vax we used for work creating image processing algorithms in the 80's. Growing up paying with with sprites, I thought this was the coolest thing in the world.

[@johnrickard8512]

When I see a KZbin channel making history in front of me, it is, therefore no surprise I would give a lake for the algorithm.

[@twol78s90]

FPS was founded by C. Norman WInningstad (I might be wrong on the spelling of his last name). It was essentially a spinoff from Tektronix. Tektronix was a very heavy user of DEC computers both internally for software development, as well as for controls for Tektronix' automated board and chip-testing systems. Winningstad was always looking for opportunity, and he saw it in creating a peripheral for minicomputers of the day that could allow them to perform large scale computing at high speed. He left Tektronix and founded FPS in 1970. There were a couple of key engineers that were involved in the development of the early array processors. One was George O'Leary, and another was Bill Hall. I believe that both of these digital design engineers came from Tektronix with Winningstad. Tektronix was very supportive of Winningstad's efforts, allowing him to utilize Tektronix' advanced circuit board fabrication lines to build prototype circuit boards for his machines, as well as using Tektronix' Control Data Cyber 73 "supercomputer" to do logic simulations of portions of the array processor logic. WInningstad sadly died in 2010 at age 85. I wonder if perhaps Mr. O'Leary or Hall may still be alive if so, perhaps can be found and would potentially be a source of useful information. The first array processor FPS built is outlined in US Patent #4,075,704. It's typical patent language, but it does give some idea of the architecture of the machine, which could be fed a program sequence and bunches of numbers to be operated on, and then it'd go off and perform the processing on the arrays(scalars, I think) and when it finished it could use DMA to push the results directly into the memory of the host machine. That PDP 11 interface board you have has all of the magic on it to interface a Unibus PDP 11 to the array processor. There is probably a big cable that would run between that board and the array processor that you'll need to find. It was probably somewhat specialized as it likely was designed for very high speed inter-processor transfers. The array processor would need a whopper of a power supply, because most of the logic was bipolar (or perhaps maybe ECL) for speed, as were the memories. I think those big TRW IC's are bipolar high-speed multipliers. The original FTS array processor had separate floating point adder and floating point multiplier that operated independently of each other. You may want to check with the VintageTek museum in Portland, OR to see if they might have some information even though they specialize in Tektronix equipment. some of the folks that work/volunteer there are old-timer Tektronix people that may have some recollections of FPS, or, might even have some FPS gear hidden away in their private stashes. It'd be really cool if you could get at least the array processor running on your 11/44. Finding the software library (probably FORTRAN) that would provide the API for the array processor could be the biggest challenge. Or, you could just rig up an interface between your PDP 11 and an nVidia Jetson Orin Nano Super, and get up to 67,000,000,000,000 operations per second for $249 plus the cost to build your interface between it and the PDP 11.

[@stewarthayne8304]

Amazing!

[@venw6811]

I believe the above is all correct. I worked at FPS from 1980 to to 1986. Didn't work on the 120b or 100 though. Yes they did make one 64 bit ECL machine. It used MECL 100K logic ICs. It was a monster and took a large amount of power. It was known internally as the 264, but I think it was released as the model 564. This was the team I was on. Unfortunately I don't have any parts or documentation for this hardware. I do have a stash of the company newsletters, which are quite interesting. Would like them to be preserved, but I don't know who would do it, and of course they do contain personal names and pictures of the employees.

[@Clavichordist]

I always appreciated DEC hardware. As a computer operator, my first gig was a computer room DEC'd out with many VAXs. I used multiple 11/750s, a few 11/780s, and an 8350. These systems were always reliable and rarely had issues. Dave's PDP-11/44 is a testament to the quality of this hardware.

[@EckNotlisted]

As an ex-DEC person aka digit it nice seeing old machines running. FYI the big machine was the 11/70 and the 11/44 was the kid brother. Myself I have multiple DEC systems both MicroVAX-II and PDP-11 on Qbus including the 11/73 card. Oh and XXDP is weird even by most DEC software. The list is long here. And of the OSs RT-11 is a good utility real time system and RSX is the one that ran whole companies.

[@acoustic61]

The PDP-11/70 was the best looking one with the pink and purple switches. I saw one at the air and space museum.

[@daffyduk77]

And then there's DSM-11 which was installed in hospitals, banks & other commercial installations but on a smaller scale than the above

[@waynetemplar2183]

I used (and even programmed) PDP-11s for power plant control/monitoring application. I loved their simplicity and ruggedness

[@Jackpkmn]

It's crazy to think about it like this. How far we've come. Breaking 100 megaflops is a big deal for a single PDP/11. 40 years later we have single gpus threatening to breach the 100 TERRAflops range.

[@IncredibleMD]

He also maxed out the memory at... 4 MB.

[@mediocrefunkybeat]

Remember kids. It's not about the weight of the computer, it's about the determination of the person wielding it.

[@Stoney3K]

It's really wild to imagine that modern graphics cards are basically supercomputers on a chip. They do exactly what classic supercomputers do: Execute a lot of parallel computations on a shared set of instructions. The only difference is that the instructions for the supercomputer is what we today call shaders. Which makes me think: Did anyone ever program a 3D video game similar to modern games to run on a minicomputer/supercomputer combo?

[@stevebabiak6997]

Back around 1994, there was something called a Freedom Graphics 3D accelerator that was a separate box the size of an undercounter mini-fridge. This was connected to workstations that supported use of that box to get “high performance” (at least for that timeframe) 3D graphics. So sort of what you asked for, minus the game part. That hardware was quite expensive, and it would be used for “real work” that needed to be done - games didn’t fit that ;)

[@Clavichordist]

@@stevebabiak6997 The drafting department at a long-gone company I worked for had one of those. I remember being wowed by it back then. Prior to working there, I was a hardware tech that worked on graphics terminals and other peripherals for various systems, and this intrigued me. What's amazing is this is around the same time that companies were running Autodesk's 3d Studio and similar programs on "regular" PCs. A friend of mine, who I ended up working for a bit later, had a designer using 3d Studio R4 for DOS on a Pentium 100-based computer. It was this software and faster desktop PCs and graphics cards that killed off the proprietary hardware.

[@RikkiCattermole]

They are co-processs yes, on the other hand, our standard x86 CPU, is also equivalent to supercomputers too!

[@mikgus]

If you have time and are nerdy enough google "Namco's Magic Edge Hornet Simulator" It is a arcade flight simulator powered by a silicon graphics supercomputer from 1993

[@ChristopherHailey]

I remember looking it up once and my graphics card at the time was 500 times faster than a Cray. Graphics cards are similar to a Cray because the do operations in parallel

[@nickm8134]

Wonderful video - hope you can get the supercomputer together! Worked for DEC for more than 20 years in the UK..I started as a hardware engineer supporting Unibus PDP-11s, so this is very familiar. My patch was 11/04s, 34s, and 84s mainly, but I worked on lots of 44s (and others) on out-of-hours support. After about 3 years I moved across to software, I did quite a lot of RSX-11 support, then VMS. I loved RSX-11 - something very special about that OS. Sysgens were always lots of fun 😪. The build quality of these machines is nothing short of amazing, though that was reflected in the price. Sadly DEC never managed the transition to micro-computing as the mini-computer market evaporated. The later Compaq and then HP years were never quite the same.

[@АндрейМилованов-у9у]

I think main DEC problem was not "the mini-computer market evaporated" but DEC architecture (good for 70s but totally unacceptable for 90s) and management (don't understanding that in 80s). Mini-computer market exists till our days, it's different multiprocessor "light" servers $10000-100000 price (of cause $1000 in 70s and now are different but market volume is much bigger than in 70s).

[@billspooks]

We had three PDP-11/40's before an upgrade. It is a pleasure to see this video. Thank You.

[@neilbarnett3046]

16:04 (& 19:04) Always check your power supply voltages. First thing I was taught about mending PDP-11s. And 5.25V is the maximum for TTL to work reliably, especially 1970s/80s TTL. The memory saying P probably means parity memory. If you have memory chips laid out in a multiple of 9, it's almost certainly parity memory. The CSR may or may not be a control status register or a checksum. If Checksum, then 0 is probably good. I mended PDP-11 PSUs for a few months. And 5.25V is the maximum for TTL, especially 1970s/80s TTL.

[@heckelphon]

Great stuff! This feels like the first course of a rich banquet of PDP-11 we can gorge on throughout 2025. I'm a-working up an appetite!

[@RobertLiesenfeld]

Every one of your episodes brings me joy. Please never change!

[@UsagiElectric]

Thank you so much!

[@acoustic61]

I love the color scheme of the PDP-11/70 with the pink and purple switches. You can always spot them from a distance. I saw one at the air and space museum.

[@WilliamLeue]

Your comments on the FPS AP-120B bring back a lot of memories. I programmed this beast for GE starting in 1976. We used it on the CT systems and the prototype MRI systems.

[@LightTheUnicorn]

You just keep pushing boundaries with even stranger and forever less common machines! Awesome to see the 11/44 running well, very interested in following along with the FPS stuff!

[@mdouglaswray]

GREAT VIDEO! I've worked with the PDP-11 in a lab setting. Really enjoying your enthusiasm with this recovery!

[@fubaralakbar6800]

"Troubleshooting old computers is...95% information gathering" That's exactly what I commented to Adrian on his recent video about the 1541 disk drive repair 💪😁

[@David-lb4te]

In the late 1980s I managed the installation and commissioning of a DEC installation for image processing, that used 3 VAX 11/785 mainframes working across a star coupler, each equipped with a FP parallel processor and an APTEC IOC2400. We also used a PDP-11/23 and PDP-11/73 for additional dedicated tasks. The FP certainly speeded up numerical operands, but the APTEC was a loadable processor for real number crunching and Fourier transforms. Regarding Intermittent operations of your system; I would look at backplane continuity. It plagued us.

[@wmrieker]

PDP-11 does not need to load into registers to do arithmetic! The instructions have flexible operands, registers, indirect, auto increment/decrement, indexed. You can add one memory location to another without using a register (except program counter).

[@captainchaos3667]

Always gratifying to see at the end that your projects are properly supervised.

[@David-h5r1x]

"If your computer isn't heavy enough to kill you is it really computing?" LOL

[@RodHartzell]

Again with an inspirational video. I know that you're not showing the off camera frustration that you've been through, but I think we can all feel it. The inspiration is the tenacity and fortitude that you demonstrate to keep looking for answers. Well done Mr. UE.

[@clangerbasher]

I remember when my small college was to have a Vax. We wondered if it was to be installed over summer break in the tiny room where the PDP-11 lived. What we students didn't know was the Vax was already living in the corner of the tech support office and had been since Easter. The techs were just testing it and learning stuff. Back then I don't think we quite grasped the speed of component miniaturization.

[@NateEngle]

And these days the horsepower of about 4,000 MicroVAXes fits in a Raspberry Pi. One of the things that originally attracted me to computers was the fact that they tended to live in rooms with extra air conditioning, and during sweltering summers in southern Indiana that seemed like a nice perk. So much has changed since then.

[@clangerbasher]

@@NateEngle Yes! I used to say never work with a computer you can pick up and throw out of a window. I spent 10 years working on boxes as big as wardrobes in rooms with no windows.......

[@crossbow1203]

I worked on VAC BASIC when I was in tech school. I remember spending hours trying to figure out problems on huge tractor feed print outs just to find out I named the same thing 2 different names in 2 different places! Duh!

[@clangerbasher]

@@crossbow1203 I think we have all earned that badge at one time. You can take make variable names more descriptive too far. The silliest I ever did was writing something in Pascal and using an apostrophe in a string of text and constantly getting errors and just not seeing it. You need to use two of them so the compiling can work and one is displayed.

[@NateEngle]

@@clangerbasher So you were basically stating the opposite of what Woz said, though for what it's worth I don't recall that Woz was ever actually reported to have thrown a computer out the window.

[@jeremylindemann5117]

When the camera shifted position at 28:16 and 28:23 I got a little freaked out then laughed at how comical it looked. It's like a scene from naked gun where there's a bunch of cameras looking over someones shoulder.

[@sjfriedl]

@29:30 - "This PDP-11 is fully DEC'd out" - I see what you did there 🙂

[@johnleclair663]

Keep up the pdp-11 videos coming ! I have a running rack mounted pdp-11/23+ with RL02 and RX02 drives as well as a vax 11/730 which might be still running. It was last time I checked … thanks for the video

[@PaxtonSanders]

That wire-wrapped card at 8:10 gives me the shivers. Try debugging that joker...

[@horusfalcon]

Lovely lagomorph at the end there. This kinda stuff, you working with PDP-11 and its various oddware peripheral systems, is what makes this channel worth watching, man. Every Success!

[@reneboeije3257]

As a side note: It is possible to add a 2nd CPU to the PDP, I remember that the RSX11M manuals had commands in blue or red that were only for multiprocessor systems. It was never officially released/supported, but there were systems in the field. So if you want to extend the capabilities of the 11/44, maybe a simpeler job...

[@PlayerClarinet]

"ran when parked" Story of my life right there.

[@PapasDino]

"Thou shall always check power rails first!" ;-) Great work, thanks for sharing!

[@stefanobiaggini]

A beautiful work of research and study, congratulations!

[@AjinkyaMahajan]

Imagine, creating all of that on a soft core on an FPGA!! Excited to watch how this project turns out.

[@hedgeberg]

The BeagleBone isn't an FPGA. Beagle makes an FPGA board, but the board pictured in this video is a BeagleBone Black, which is an SBC built around the TI AM335x, which is just a 4 core (iirc? But might be 2-core) 32-bit ARM SoC with some nice peripherals.

[@axelBr1]

Wonderful to see you working on some DEC hardware, may be you'll end up with a VAX cluster! Connecting your PDP 11 to the Floating Point Systems hardware will be amazing. Good luck!

[@bobdan9856]

I am not going to shakes my head at turning a pdp-11 into a supercomputer. You are the one person who could actually pull it off. Now a difficult but awesome project would be building a pdp-11 cpu card out of vacuum tubes.

[@VishalDudhani]

You're my favourite tech guy out there... Every night you're that kind friend that sings me tech lullaby to sleep.... You calm me... Help me ground myself....

[@PlaywithJunk]

There was a more modern CPU module for the PDP11/44. It came from a third party manufacturer but I dont remember which. That CPU was only 2 boards with LSI chips and also a higher clock speed...

[@RicoD5]

More of these large Mini’s please! Makes me go back to the MAI Basic Four MPx supermini’s I specialized in during the 80’s and early 90’s.

[@jaybrooks1098]

love this fork of your journey.. pdp 8 was my first mini computer..

[@iggysfriend4431]

It' bonkers how the CPU spans 8 cards. Mind you, I remember the DEC Vax 11/780 having its CPU on something like 20+ PCBs that made up its CPU.

[@rigues]

About the weight, a famous computer once said: "never trust a computer you can't lift". 😂

[@HessTruckBOOKS]

Honestly I didn’t understand most of this episode but stayed for the positive vibes!

[@SuzuranMajere]

The Unibone does grant forwarding in software. Until the Unibone software is running, the Unibone is an opening in the grant chains. The 11/44 console will not be happy about it.

[@jakeehrlich8113]

regarding keeping it fed, as long as the coprocessors have their at least some of their own memory there will be an arithmetic intensity (I'll define that later) cross over. Let's say a coprocessor can perform 10 MFlops per second and you have a 40 KB/s link between them. That means that the coprocessor needs to be able to do 10 MFlops for ever 4KB or about 250 operations for every byte sent. This number, 250, is the arithmetic intensity your algorithm needs to reach in order to make full use of the system. That's actually extremely close to what modern GPUs and AI accelerators need. Its perfectly reasonable to keep them fed under these constraints given the number crunching you're doing has sufficiently high arithmetic intensity. Matrix multiplication can have very high arithmetic intensity. Let's say you want to multiply two 512x512 matrices in single precision floating point. That requires 2*512^3 flops and 3*4*512^2 bytes which gives an intensity of 2*512/12 = 85.33 flop/byte so the intensity isn't quite high enough to hit our theoretical 250 operations per second. If we go upto 1024x1024 and we give ourselves a crutch leaving the result on the coprocessor (not valid for all use cases) the math becomes 2*1024^3 and 2*4*1024^2 which gives an intensity of 1024/4 = 256! A 1024x1024 matmul bound only by compute at 10 MFlops would take ~100 seconds. This seems very reasonable. I think this sadly requires on the order of ~1MB of memory even with a lot of careful implementation. There are lots of tricks possible if you know more architectural details though. If the co processor only has about 1kb of local memory its going to be extremely hard to keep it fed sadly no matter what tricks you use. Other possible options: * Instead of matrix-matrix multipication, do matrix to a power multiplied by a vector. This decreases memory requirements and increases arithmetic intensity. * Mandelbrot set calculations naturally can have a high intensity (and they look cool if you can print them out!). The super nice thing here is that you can achieve high intensity at very small memory footprints (like 1 pixel at a time).

[@timothyp8947]

When I started my first job, back in the '80s, my first team leader had previously written programs for ICL's DAP - an array processor for ICL mainframes. Think he said the programs for it were written in Fortran, but honestly it’s so long ago now I can’t be sure. He was certainly very enthusiastic about it. I believe the claim was that for certain workloads like weather forecasting, this thing was right up there with a Cray but at a fraction of the price. I think at one point the Computer Conservation Society (CCS) in the UK were trying to resurrect an ICL mini-DAP, a version of the DAP that was attached to ICL PERQ workstations, although I think they hit a brick wall on getting that going. The PERQ was a workstation from 3 Rivers in the US that ICL packaged up for the UK market.

[@makomk]

Most of the larger British computers from that era seem to be pretty badly documented and conserved, unfortunately.

[@timothyp8947]

@ there is some information around. I found electronic copies of the ICL Technical Journal online (maybe on the Fujitsu website?) a number of years ago which has technical articles on all kinds of ICL gear & s/w. There’s a book called 'The ICL 2900 Series' by JK Buckle about the design of those mainframes which I’ve seen online, and an in-depth on the design of the VME operating system by Nic Holt called 'The Architecture of OpenVME'. I’ve also found bits and pieces of Atlas, KDF-9 and other machines courtesy of Google, and the CCS has a handy online quarterly magazine with articles on these various beasties.

[@steveunderwood3683]

We had an FPS AP-120B with 2 of the 18MFLOP cards based on a Weitek floating point chipset in the mid 80s. We used a VAX11/750 to drive it. I originally specified an 11/730, as it was just a controller. However, everyone said nooooo. The I/O is so slow on the 11/730 you really need to get an 11/750 instead. So, that's what we bought. We also had a graphics and memory box from a UK company called Bear Systems. That had 192M bytes of RAM, which in the mid 80s made it a >$100k box. The Fujitsu Eagle was quite a ground breaking product. It really brought the price of fast reliable storage down. It gave you 470MB of storage for about $10k, which sounds eyewateringly expensive now, but seemed like a bargain in the mid 80s.

[@eak125]

I love how the instant the pateron list disappeared, the screen brightened just as you said "hallelujah!". It was perfect.

[@der.Schtefan]

You're a star, and if you can program tube computers, and the quite obtuse Centurion, then I have no doubt you will familiarize yourself with Unix, the operating system that was created on PDPs (many say PDP11, but aaaaactually it was the PDP8 first). Zero doubt with you :)

[@linxbit8984]

oh the title alone tells me this is gonna be rad

[@antronargaiv3283]

You also (don't know if you simply didn't show it) want to clean the face of the flat flex that mates with the busbar itself (and the busbar, obv)...that's where the main current should flow. Cleaning the clamping plate and the top of the flat flex helps the current flow to the threaded studs.

[@johncloar1692]

Nice work I would of never though the power bus would of been the problem. Thanks for the video!!

[@noth606]

"have" not "of" like "I would have never" even though that isn't optimal either. "I would never have... would _have_ been the problem". I would invite you to _not_ take my word for it but instead look it up, since the explanations properly written up provide proper detail and I'd rather not type it all out. And no, it's not grammar, it's the wrong word entirely.

[@Aeduo]

The computer was at least pretty helpfully pointing to a power issue but yeah with these old things it's really easy to overthink what might be wrong, because there's so much which could be wrong, that one could miss the obvious pretty much "is it plugged in?". He's probably used to these sort of diagnostics though and will know to check voltages and where voltages stop appearing or where things aren't connecting. It's lucky it was in such a nicely accessibly spot.

[@RikkiCattermole]

I had a look through UseNet archives, and it appears Oracle owns FPS now. So doesn't look like there is a possibility of going that route to find schematics. There wasn't anything obvious although I didn't deep dive it enough to say if there were schematics available on it. But I kinda doubt it, FPS hardware didn't have much technical details regarding using it on it.

[@paulmoffat9306]

Back in the early 80's I designed and built a microcomputer prototype for future production at the company I worked for, using a complex architecture on the main board, and I chose Intel's Multibus interface for the auxiliary I/O functions (disk drive etc) and for primary functions during development. It took me weeks to get it to work at all, and finally discovered that the Intel handbook for the Multibus priority selection was INCORRECT, and had one control line inverted. (verified by using their boards as 'gold standards'). I still have that computer on my shelf, almost 50 years after the fact, and all my board layouts and manuals. Chip sets: 8086, 8087, 8089, NEC 7220.

[@stuartrobb673]

The reason the cray 1 was semi circular as, apparently, to shorten the signal paths as much as possible to maximise data transfer speeds….

[@Peter_S_]

More than being short, the C shape allows all the interconnect lines to be the same length. The clock was 12.5 nS and the architecture is fully pipelined superscalar so any internal latency essentially disappears once the pipe is full but the pipe needs to be fed consistently, thus the need for consistent delivery of data from any part of the ALU or registers.

[@teklastm405]

A piece of nostalgia. Thanks.

[@William-t5v]

I just love this channel! The tech information is outstanding

[@augustuscrocker9328]

Well done again ! It's very satisfying to see old code running again. These were very capable machines in their day, doing important work. Would be great if you could find some "period" application software to run on it -- compare it to some of today's bloatware. Back in the day we knew how to conserve space and make things run quickly.

[@mattilindstrom]

I have no doubt the CT reconstruction greatly benefits from an array processor's help. The (inverse) Radon transform is a demanding computational load, even if done with the help of FFT. And oh boy, the wire wrapped backplanes are glorious. The practice is a bit clunky for production, but as a rapid prototyping technique it's surprisingly useful. Had the pleasure in the mid 90's of modifying some ancient shop-made lab equipment with a good wrapping gun and very nice wire. Ripping out old wires in the Nth basement floor of the loom was painful, but adding new ones was lightning fast.

[@baronvonschnellenstein2811]

Yay! David has brought the PDP-11 back out to play :D But wait, there's more! A super-computer add-on, PLUS, a co-processor for the supercomputer that's 2U or 3U high in itself, 🤣 Completely insane ... and I love it! Overshadows maxing out Amigas just a tad 😂 - Ah yes, I get it now - A T-shirt-worthy quote starting at ~20:15: "Remember folks: If your computer isn't heavy enough to murder you, _can you really call it computing_ ?!" - GOLD!!

[@chuckperry3313]

I used to work for DEC, FPS, and CRAY. There are things that need to be done on the Unibus when you have boards in or out of standard slot you have to install or remove jumpers OR little bus grant cards on some of the slots. The KKAAxxx stuff is the code for the type of computer you are on. I'd really have to talk through what is going on. It has been a long time. I forget how to set switches on the 11/44. you can make tests loop and do quick or extended testing. Have fun with this. Last I knew there are still a lot of old engineers from FPS that live in the Portland OR area. What a walk down memory lane. Have fun.

[@Spiker985Studios]

It's *wild* to me that the power bus's relatively tarnished surface was enough to cause that voltage drop - but the specification for that rail must be a super tight variance I was also surprised to see it has ECC memory!

[@somedudeonlineidk]

New year, new project... nice 👍

[@VK2FVAX]

Those FP units used to go through Ross's disposal auctions in Perth back in the 90's. Perth was back then and now a mining and resources town so this stuff was churn gear from the hundreds of exploration companies. Maybe some of it was preserved by people or even better ended up at the ACMS WA Museum. I can remember Prep talking about them a fair bit.

[@georgegonzalez2476]

Truly a labor of love. Not terribly practical, though. Your basic iPhone has much more number crunching power than that ton of special equipment. Good luck.

[@jfbeam]

iPhone??? Many modern _TV remotes_ have more computing grunt. We've come a very long way from those days. Looking around my office, some of the oldest and weakest tech is likely the two ancient HH SCSI (1) drives. The print server in the decades old HP laser printer is a close second - and likely faster than the PDP 11/44. (technically, the SH8 on the LOM in the SunFire T2000... but that's more microcontroller than CPU/computer. the HD and printer run vxWorks.)

[@cozmothemagician7243]

I can't wait to see what kinda use you put your super computer to. Someone below mentioned the mandelbrot set . Maybe software exists. We know there were some decent ways to show graphics even back in that beast's day. Whatever your supercomputer does, I am sure it will be fun to watch you put it through its paces. Much fun ahead!

[@halitimes2]

Mark is a legend!

[@vmisev]

Thanks for the video; great project! It's always joy watching DEC machines 👍

[@GothGuy885]

Awesome ! I had the chance to get the entire enterprise system out of our bank of America that had been there since the 70's, and they were moving to a new building, and doing a total upgrade of their system. unfortunately, I was not able to take advantage of it due to a loved one being on hospice. but he did recover after a year and two weeks. and he came first, so I imagine that the system went to E-waste, but my loved ones always come first

[@paulforester6996]

I love the intro music.

[@t.w.3]

I remember scrapping a lot of old PDP and other stuff back in the mid-90's when the craze with Pentium servers and UltraSparc was the thing.. Didn't have space to save them, but grabbed a few of the smaller under-desk units... Sadly had to get rid of those due to moving in 2003.. None wanted them, so they went to the recyclers.. Had lots of old stuff that went to the recyclers, Norsk Data Nord5500, Digital MicroVax both with their terminals, keyboards, books and disks, and 4 different HP-UX 9000 workstations, and old IBM AS/390 Processor module and disk-unit, and lots of mid to late 80's computers, among them a "Corona 8086 "portable"".. Sad to look back now, but that's life. Can't save them all.

[@denawiltsie4412]

While I have had no DEC experience, you likely have another problem. I spent a good deal of time with a General Automation 18/30 which did all the floating point in software. A company offered a floating point unit for it that you would install in the I/O cage and away you would go. The hitch was you had a different set of software routines that would sent your calculation to the processor instead of doing them in software. If you don't have it already, you should be looking for the software support package for that hardware.

[@janbrittenson210]

Funny, I haven't touched a pdp-11 (other than tinkering very briefly with simh to boot rsx-11m+) in 35 years. Yet, when I see 012700 I immediately recognize it as the instruction "mov (pc)+, r0", so loads a constant into r0 (with the constant in the next word). And 112700 would be the same except "movb" for a byte instruction (still loads a word as the PC can't be on an odd addr). Fun video and fun project! These days though I'd probably take the easy route and whip up a board in KiCad, with a simple FPGA (like a MachX02) doing all the logic, bus arbitration, DMA, etc. Admittedly not very period accurate! :) Unibus is very simple.

[@peterw8430]

This takes me back to my first internship in the 90s, coding in MUMPS on a pdp/11

[@tomteiter7192]

Damn, seeing it again, I really have to do something nice with my DC-J11 Chips... It"s totally fascinating the these things are essentially a PDP-11-on-a-chip. Add an UART chip and BAM Debug console. Add a few static RAMs and you can execute code...

[@bradnelson3595]

Thanks, Mark.

[@jamesszabo6105]

Not sure what your talking about half the time, but really enjoy your video anyhow.

[@douglasmayherjr.5733]

Don’t know much about it but seems cool. Thanks for the videos.

[@FryAndLeelaAndBender]

This KZbin channel rocks!! And also the PDP11-related content...

[@TD_YT066]

DeOxIt for all your connectors, there is a Gold version for gold fingers. It lubricates and protects metal-metal connections. They also make a version for potentiameters.

[@frankowalker4662]

Ambitious project. It should be fun.

[@markramsell454]

New sport: Computer Power Lifting, pumping you up!

[@pianoman4Jesus]

At @4:28 FINALLY! May you share the source image you inserted for that? My father trained on repairing the GE 8800 CT scanner. So I spent countless hours tagging along on service calls to that unit. Beautiful MSI 850 X-Ray generator at the far left of the photo... I remember the front panel being off of it, I holding a hand held exposure control utilized while seasoning a brand new X-Ray tube. My hand would get tired from holding the back rotor button in for soooooooooo long. (I thought of placing a rubber band around the hand control to at least keep the Rotor button held in. Then I could be left to focus on pushing the X-Ray exposure button each time the beep tone was emitted.... or as best as I can remember after all these years.) My favorite era of GE X-Ray generator... when the MSI series was being produced. The Data General looks like Nova 820 minicomputer came with it. I do now recall there being some co-processor my father mentioned in the rack! The medical center also had a Tecnhicare 2060 in the Nero institute area.... which he was also trained to repair. That was controlled by a DEC PDP-11 of some sort. He got trained on that unit by a third-party expert based in the Detroit area that had a reputation for being able to troubleshoot down to the offending logic chip.... I recall one story of that person being called to a situation where the scanner would put a light streak across the image, always at the same angle. First evening he reviewed the scan, took measurements of the angle the streak was appearing on the image. Then went to bed at the hotel. Next morning, he arrived back on-site, instructed THE EXACT IC to swap out. That being done, the streak was GONE! He knew the Technicare 2060 architecture as a third party expect THAT WELL. At @8:16 So sorry..... this request comes about 33 years late. That GE 8800 CT scanner got sold and sent to China back then. My access to full schematics of the machine left with it. But yes, full schematics of the GE X-Ray unit, likely Data General and that co-processor schematics were also included.

[@HfLuo]

I'm in China and recently rescued a FPS AP-120B from scrapper. I'm amazed to know Usagi is also working on FPS systems. My AP-120B came exactly from a 8800 CT scanner, but rest of the system were all scrapped, including the DG Eclipse minicomputer. Such a pity! It seems China has imported a lot of 8800 CT systems during the 80s. A member in Discord group also worked on this system and one of his job was to modify the AP-120B so that it can't be used for other purposes. They are just too powerful for its time!

[@pianoman4Jesus]

@@HfLuo Small world indeed! I am guessing the one (GE 8800 CT) I know of going over went either in 1992 or 1993.

[@bluerizlagirl]

The oldest problem in the entire history of electronics: Poor connections! Nice to see this old beast up and running ..... It should be pretty capable once it's fully pimped-out!

[@leoedate]

Absurd undertaking to think about

[@NateEngle]

Dave seems to be motivated by the whimsical nature of the idea, but the PDP/11 was such an elegant machine just as a minicomputer of its time. This project reminds me of a Taoist parable where a carpenter called an old gnarled tree useless because he wanted to make it into something else. But for what it's worth the task as he outlined it does seem like it could work, and I'm sure it will be an interesting series of videos which is the real point.

[@nathantron]

I wonder if it would be good to apply some corrosion preventative conductive grease to those rails for long term preservation.