It is truly amazing how far we've come in the 30 or so years since this was current. I wonder if people back then could have even conceived of how absolutely ubiquitous computers have become?
to be fair, yes and no. I was lucky to be part of it, I went from being a systems programmer in the mid-1970's on the original IBM 360 systems, to a Senior Distinguished Engineer at Dell 40 years later. I kept these laserdiscs when IBM closed it's formal library at the IBM Pouighkeepsie development and manufacturing plant circa 2002. This size of machine had already become recent history. When I started in the early seventies, the computers were big enough and ran hot enough that for night shift we'd bring our food in and store it in the back of the computer to keep it warm. I was briefly Chief software architect for the development team that built the pre-boot software for PC's. All that stuff that handles the BIOS etc. That system was more powerful than the processor we kept our food warm in. By the 1980's, we knew that things were changing rapidly. I already had a watch with a computer in it instead of a set of cogs and winding mechanism. Moores law mostly predicted the change, it wasn't clear how it would get used, but wouldn't just make big systems bigger. I was lucky enough to work at Chemical Bank in New York with Hank Kee - who was a visionary and he was a big part of the reason I was able to push IBM on some of their changes. I Laugh not not so much on the size of the processors, but the amount of data and size of the operating systems. I get headhunted from the UK to help Chemical Bank start with virtualization in 1983. We ran a production credit card approval system that had some 300 telephone operators waiting to receive calls from merchants, and look up a credit card account. It ran on an 8MB MVS system, we virtualized it and ran it under VM/370 HPO 3.2 on a 12MB system. My Android phone now won't run effectively on 6GB. In the late 1990's one of the big pushes for extended memory on these 3090 systems were the 30-50MB CAAD drawing and modeling files, now we routinely load up files that are hundreds of megabytes.
I'm still here after 40 years. A lot of this heavy metal was in the banking industry where I started. As big as these appear to be, you haven't seen anything if you haven't been in a big rack-mounted server room today. Makes these look a little on the "small" side.
I was an IBM Field Engineer specializing in Large Systems (now known as Enterprise Systems) from 1977 to 2015. As I watched this a lot of memories came back. Thanks for posting.
If you have or see any old IBM training laserdiscs, I'd love to be able to convert them for others to see. I'd really love to find something on the 3850 Mass Storage (cartridge) System.
@@ian_b When I first started in 1973, the mainframe system I worked on, and IBM 360/40, only had 48Kb main memory to load the OS and run programs. A NEST doorbell/Cam has a greater amount of memory and faster processor than the 360/40. A version of the 360 went up in the space shuttle, provided the computing power in 256Kb main memory.
@@MarkCathcart I have found some pdf training and a presentation that runs on Lotus, but no videos. Yes the 70's and 80's were a great time to be in IBM. Started declining for the employees in the 90s and went down hill fast after 2010.
I used to fix this. PST - Product Support Trained level. Our IBM office bought a refurbished 3090. Due to anti-trust laws, we could not get special pricing on any new unit. One tri-lead paddle card was damaged. We identified the signal and it was a sensor line not a high-speed data line. I had good eyesight in those days. I soldered the connection. I don't recall we ever replaced that cable. Then I got posted at a remote site and never came back to the mainframe group. It was a fledgling team that handled Point-Of-Sale and RS/6000 which was a blessing since this introduced me to AIX and I was then trained on the 9076 supercomputer. I ended up being the local GOTO guy for the 3584. I was also trained on the 029 card punch but sucked at it. Lots of war stories. EDIT - it was not a tri-lead but a triaxial cable which was covered in blue foil and about the thickness of a pencil lead.
@@MarkCathcart I saw the writing on the wall and following some advice - left the party early. The moment I was eligible, I went for early retirement. My colleague didn't. He was from IBM Canada and worked here as contract staff. He wanted to reach 50 years working for IBM but passed away on his 49-th year. The prestige of the IBM Customer Engineer had evaporated. We were headed the way of dinosaurs. We never met the client anymore. All we met were the staff of megacentres where the clients had outsourced their servers to. We had descended to the level of janitors. I am glad I never went back to mainframes. Those team members were getting arrogant. Case in point - A newly installed 3584 never finished initializing. We were territorial. 3584s in their account were handled by them. I was asked to assist. I went there with a trainee and within a minute, I pulled off a Richard Feynmann radio repair. I was never trained on the 3584 and learnt everything through reading the Maintenence Information Manual and trying to reverse engineer the machine. I deduced that they used an FPGA to act as the controller. I knew exactly where they went wrong. They installed the machine without referring to the MIM - it was still unwrapped. I told them where they went wrong and even found the page in the MIM but they still refused to admit their mistake. They mounted the crash stops at the wrong position and the robot couldn't find the barcodes and home flags. Reposition those and voila! That really impressed the trainee. Guess who knew shit and who knew his shit. EDIT - I must be losing my memory. What happened was the crash stop contained the home flag and when this is misplaced, the robot would find home, move the preprogrammed distance and shine the barcode reader to read the barcode labels and couldn't find any. This stopped the initialization. Pretty obvious if you did any programming and diligently enforced input conditioning and sanity checks. I published two programs on github. These leave audit trails so I debug from afar. Since I was obedient to the manual for everything, I never encountered this problem but having watched the power on sequence countless times, I realized this instantly even if I had never come across this before. With my knowledge of AIX/unix, I used "strings" to extract the latest HEC and HECQ codes from the microcode file along with their descriptions. Our technical expertise had drastically fallen. NOBODY was interested in this finding. Nobody drilled down to the error codes anymore. Another case in point - a REGATTA was gutted for its parts and when the crisis was past, new and original parts were installed but the machine failed to power on due to a massive mismatch in the VPD - Vital Product Data. Two CEs failed to overcome this. I was the second standby and was called in. I have absolutely no training on this machine. I can't even remember the machine type. But I could go into the error codes and painstakingly identified the serial numbers and error codes and after rearranging the power supplies, it could just power on. We needed to fly in someone from France to sort the rest out. The company hides so many things from us ostensibly to prevent these machines from being serviced by third parties. Indeed my very first manager betrayed us by taking along a curated group of CEs and SEs to offer MA to government clients. It was only at a webinar late in the game that I delved into the documents supplied for the 3584 to find that there was a secret menu that would exercise the robot and cause it to randomly rearrange the cartridges. All along I had to rely on tapeutil commands to perform the same task. Major BUMMER.
Thanks for this! In the late 1980's, I worked in American Express's mainframe center in Phoenix. We had a total of 7 IBM 3090-Js (IBMs largest as I recall). In the same center, there was a vast ocean of DASDs and dozens of smaller mainframes as well.
In 1983, I worked for Chemical Bank in Manhattan, before we got 308x computers, we had 3033MP systems, they were even bigger than these. My experience virtualizing the SWIFT banking system for Chemical Bank lead to me working at IBM Kingston on the firmware/supervisor design for the 308x processors. I have a bunch of these videos on laserdiscs, I've agreed to provide digital copies to the Computer History museum in California.
I wanted to be one the IBM Project Managers on the AE Data Center in Goldsboro in 2011 but it did not work out. I spent years at MCI with the back of my chair up against a 3090-600. They were my first customer to cross the gigabyte threshold on 3390s.
I worked for six weeks in Japan as RSFSG. We once toured a "small" customer. There was a dedicated office for an IBM manager and the CEs. One floor for printers, one floor for DASD, .........
I started in IBM as a part timer in the late 80's. As a young man my job was to pull the Bus & Tag cables out of the 3090 and pull out the cables from under the raised floor. There were a bunch us part timers and what great memories I have. It was non stop work to get the old system out and the new system in. Worked in downtown Brooklyn at Brooklyn Union Gas. The experience IBM's were great to be around. Professional and well respected. They were great educators to all the young people trying to learn.
I’m surprised there is no mention of the vector facility in the video. My company, Amoco Corp, used this system for seismic data processing in the late 80’s. I worked on vector enabled scientific subroutine library for the system (Fortran)
I think that is because there were entire different training programs for dealing with the vector facility. They were not that common at the time. Good spot though.
I worked for IBM from 1987 to 2008, eventually ending up a member of the Academy of Technology and a Distinguished Engineer. I still don't understand hardware, which is why I found this laserdisc so awesome. The best people, and some of the most influential I worked with at IBM were the hardware FE! My first big break came in 1979 when a FE gave me the source code to the IBM 3741 so I can hack in a change for the company I worked at the time. We needed to be able to transmit a single record insurance policy from a remote office to our London office quickly, the insurance record required 140-bytes, the 3741 only supported 128. I patched the code to allow the 3741 to split the record but transmit 140-bytes which avoided a batch program at the HQ to re-assemble the records before adding them to database. Seems trivial now... 6-years later I was working for Chemical Bank in NYC, we used a mod for the IBM Series-1 to do much the same thing, except this time to do online credit card authorization automatically for the first time. Amazing times.
@@mrkkitson I came up with this procedure and it took over the official oscilloscope 50%-50% waterfall technique - but only in my country. I plugged my digital meter to one capstan sensor signal and slowly turned the capstan. I noted the lowest reading and the highest reading. Add the two numbers and divided by two. Then I ran the capstan motor by jumpering some pin or maybe plugging the capstan motor to another socket. I then adjusted the pots until the reading matched my calculation. The meters at those days was average reading instead of true RMS. This spread like wildfire BUT no technical manager nominated this for a PMR to help the rest of the world. I was a science grad not an electrical/electronic grad like the others. I applied my knowledge of class-D audio amplifiers.
I was a part timer In the 80's for IBM and they used the part timers to pull the cables under the raised floors. Talk about heavy Bus & Tag ( I believe it was call ) cables. I still have a processor from a 3090. I also have a memory module from an IBM 1800 system that used Donut memory we called it but was magnetic memory, Great memories from those days.
I worked on an IBM 370/145 circa 1975, the rear part of the main processor was empty. The space was reserved for additional memory which would be present in a 3670/148. So we used to keep our food warm in there when working on night shift.
one thing is the incredible amount of engineering that went into making the computer, the other thing is the equal amount of engineering that went intro trying to squeeze the whole thing into as small a box as possible, even though compared their later creations it's still massive. And these still need a cooled room despite having their own water cooling
I've worked on 3090 and after on ES/9000 IBM and Hitachi HB100 in Banks Data Center in Italy, since 1989 to 2000. It's amazing to see again those big big computers
They were. Even the single frame older 370 Processors had some amazing design, the cables were not just put in where there were spaces, they were designed and sized to fit in exact places. Thats why they were all so expensive...
The engineering is amazing-not just the design of the logic, but the physical arrangement of the components. Where do you put the power supplies? Where do the CPUs go? How to you route the cables to connect everything
Yes, it had to be, these machines were far from perfect but the complexity of the design, engineering and operating system software support was suitably impressive for the time. I'm looking for some more interesting videos/VHS/Laserdiscs to show off the technology of that era. Thanks for watching.
there were a lot, and I mean a lot of cables under the floor interconnecting the components. the 3090/400 was two 3090/200 interconnected with only a few common components. The machine itself had firmware called "Logical Partitioning" which would allow you to virtualize the hardware to a point, and share hardware components between a number of partitions by allocating dedicated resources to the partition. The entire machine, or one side, or one or more partitions could run the Virtual Machine operating system, which could further share resources.
to be fair, these three boxes are just the processor part of the computer. There would be other boxes that contain the "modem" communication processor, disc storage boxes(lots of them) and other things found inside modern PC's and laptops.
@@MarkCathcart Why put the computer in a whole room? Just build the room into the computer already hahahah it's amazing how small and efficient things got.
@@michrain5872Keep in mind, this was 1985. Microprocessors had been around for 10 years, and were slowly catching up to high-end minicomputers. 16-bit machines were making headway amoung the cheaper and more widespread 8-bit micros, and they were starting to talk in MB, not KB. 32 bit processors were only a few years away, and close the gap in terms of capability, if not raw performance.
I worked on both a 3033MP and a 370/165. I wasn't around for the installation of either but the deinstallation of the 3033MP was something to behold. The 3033MP also had a dual 3850 MSS. Those were the days, I still have some of the cartridges as souvenirs. 100MB per cartridge.
Surelly this system could have been built in a way it would occupy less space. It could even be more faster and have more capacity. But I suspect that IBM systems engineering had other goals on mind. Like reliability, availability, fail-safe features. One key factor to achieve these goals is temperature control. And power supply quality and reliability. To achieve goals like those, you have to avoid crowding modules and componentes (beyond some limit), you need space for airflow. Those were some of the challenges on the late 80's hardware engineering and frankly the customer targeted by this kind of system did not care about some dozen floor square feet more.
You are ignoring or don't know this was a replacement for the IBM 3033 processor. The 3090 required 50% less space in total for more than double the performance per sq-inch. These systems, as the 3033 systems did, required water cooling.
Remember these machines would often require a large installation no matter how small the actual computer was. Between electrical, cooling, archives of storage media, terminals, etc etc. It probably could have been smaller as you say, but shrinking it wouldn't have been high on either the customers or engineers priority list.
For about $3 Million in 1988 you got a processor that ran at 66 MHz, delivered 67 MIPS, and 130 MFLOPS. In MIPS this compares to a 486 @100 MHz, in FLOPS it compares to a Pentium 133.
It's a fun data point, but as someone who worked on these machines in real commercial environments as well as later graduating through IBM as a Systems Engineer, to Distinguished Engineer, we all knew it was never about MIPS or FLOPS, it was always about data, I/O, concurrency and virtualization, and yeah, reliability, almost none of which you get could from x86 architectures until the early 2000's.
@@MarkCathcart I've always been a workstation guy. I was doing some orbital mechanics simulations in the late 90's and using a Linux work station as a terminal for one of the schools mainframes. As a joke I ran the sim on my "terminal" and it only ran half as fast because I had to cache to disk and could not fit all the data into main memory. I finished the project on my workstation and saved a lot of charged CPU time on the mainframe.
@@AndrewTubbiolo Yup. I was a developer on the mainframe side of the worlds first home banking system PRONTO, the pre-cursor for .zip files was developed in part for us, .arc files by Thom Henderson, and I wrote grid-like programmers compilation system that allowed programmers to submit batch compilations on the mainframe, which were offloaded to PC's running MicroFocus COBOL.... heady days all between 1983-1986. One of the managers I worked for was Hank Keyes, who won a PC Tech Journal gold medal the first year they did them, other winners that year, Bill Gates and Phillipe Kahn of Borland... I went on to join IBM, and wrote the first Linux mainframe strategy paper in 1997/8. Hence my interest in this stuff... still.
@@MarkCathcart a friend of mine worked at a bank and was lucky enough to grow from helpdesk to first mainframe and then unix groups. i remember there was one time in his early years when he got curious and ran seti@home on a linux partition - but only for a short time till IBM called them up since he created a messy overload on the whole system. we both laughed since it was such a sad, unprofessional first impression to create of linux as a citizen in the system he created. naturally it was without bad intentions, just wanting to see how amazingly big that machine was - like when you'd stand in the largest church of the world and you look up and say "woahhhh". and it was big - but the scheduler was not amused nonetheless.
Not only well built but excellent support in the field right back to the factory. System down situations were rare but Montpellier could send engineers if the field CE's couldn't fix it.
Wow, amazing video, thanks very much!!!! From what I've seen from the cover, this laserdisc has two other contents, would you also upload those? I would LOVE to watch it too!!!
@@raphaelmsx I don't know yet, I have the videos done, I have to complete a submission form and once approved submit. I have some lserdiscs on early IBM PC training that are all part of the same submission, so it's taking longer than I expected.
Actually a lot simpler than these. So much now is on board the actual processor and SOIC which improves speed and removes latency and wiring both on and off the "system board'.
Laserdiscs had digital sound but an analog picture. I have many music laserdiscs films in jazz that I've published on another channel @ctproduced - the sound is as good as, and sometimes the same as CD quality.
I haven't found any info but the 3090 was announced in 1985 and the hairstyle looks late eighties so I'm guessing circa 1986 until I can find out more.
This would have been made in 1985. It was internal processor support education, so would have been used during that year along with "Fist Customer Ship" (FCS)
If you have the source code, it can certainly be ported and would work. That said we were playing Adventure on three generation earlier mainframes in 1976....
@@MarkCathcart your a dreamer silly little dreamer oh you got your head in your hands oh no oh no I say dreamer Doom needs more processor power then that old beast can muster it can only do a a few hundred thousand calculations per second and Doom needed at least 25 to 33 million calculations per second on a 486 to at least run Doom and faster 486s were better for frame rate from there so nope because it would take minutes per frame and you would get killed before you could see it coming sadly
@@ctproduced yeah and that's the funny part the old IBM is way to slow to handle Doom as it only does a few hundred thousand calculations per second the first personal computers made that system look like an ancient slow pos computer and they were a fraction of the price which was the best part to because the c16 could do over a million calculations per second on it's CPU it goes to show how they were obsoleted by commodore and apple back then can you imagine buying a commodore pc and realizing you paid so little and got a computer that kicks the crap out of that huge massive IBM pc you would have been on cloud nine back then
Almost certainly so. One of my good friends from my time at IBM Poughkeepsie worked in information development, she often did voiceovers for marketing and training videos. I still have an MP3 from a video for object oriented technology in 1994... she wrote many of the manuals and definitely understood the technology.
imagine buying one of those then getting a commodore 64 only to figure out how much more powerful the commodore 64 is over that beast you spend million of dollar to buy🤣🤣🤣 that would be a kick in the teeth🤣🤣🤣
@@deepsleep7822 I think he means his opinion is "junk". At the time these machines were state of the art, the biggest fastest processor in the industry. One man's opinion won't change that anymore than all those people who said mainframes were dead in the 1990's. 30-years later they are still here and running many of the world's core business apps...
This computer was over-big and over-bulky even in its day. Compare to any of the Unix vendors of the time and you can understand why IBM declined so quickly.
That's simply not true. When the 3090-200 and 3090-400 were introduced in 1985/6 Unix systems were nowhere. Sun Sparc wasn't even released until 1986/7. Cray and OEM IBM systems were the only effective alternatives. Cray pretty much only targeted engineering and scientific, and the OEM systems were for the most part running the same IBM software. I agree that UNIX systems and CMOS processors made rapid progress during the life of the 3090 processors and that's why this was effectively why this was the last major range of water-cooled multi-frame IBM mainframe systems.
You are wrong, of course. As an old IT hand who worked on both kinds of systems, I saw it first hand. Look at those things. They're the size of small city buses. In terms of computing power per cubic foot, IBM wouldn't catch up for another 20 years.
@@igfoobar as a former Distinguished Engineer in IBMs Systems Group, and a member of the IBM Academy of Technology, I make mistakes from time to time. So, go ahead and tell me what UNIX systems there were in 1985/6 that could even run transactional workloads let alone scale to a similar capacity at that time? FYI in 1984, I was part of the ad-tech group at Chemical Bank in New York City, we were literally desperate for faster systems than IBM Could provide, we'd gone from a 3033MP system which was even physically larger than the 3090-400, through the 308x series and we'd just rolled out PRONTO the worlds first home banking system. If UNIX systems were as good as you say, we'd have been all over them. So, which ones specifically? I'll wait.
@@igfoobar Really, go ahead... you are off by a decade at best. Unix Clusters couldn't do what the 3090 mainframes could do until at least 1996. My guess is you are confusing the IBM 4300 mainframes. So, what Unix vendors?
It is truly amazing how far we've come in the 30 or so years since this was current. I wonder if people back then could have even conceived of how absolutely ubiquitous computers have become?
to be fair, yes and no. I was lucky to be part of it, I went from being a systems programmer in the mid-1970's on the original IBM 360 systems, to a Senior Distinguished Engineer at Dell 40 years later. I kept these laserdiscs when IBM closed it's formal library at the IBM Pouighkeepsie development and manufacturing plant circa 2002. This size of machine had already become recent history.
When I started in the early seventies, the computers were big enough and ran hot enough that for night shift we'd bring our food in and store it in the back of the computer to keep it warm. I was briefly Chief software architect for the development team that built the pre-boot software for PC's. All that stuff that handles the BIOS etc. That system was more powerful than the processor we kept our food warm in.
By the 1980's, we knew that things were changing rapidly. I already had a watch with a computer in it instead of a set of cogs and winding mechanism. Moores law mostly predicted the change, it wasn't clear how it would get used, but wouldn't just make big systems bigger. I was lucky enough to work at Chemical Bank in New York with Hank Kee - who was a visionary and he was a big part of the reason I was able to push IBM on some of their changes.
I Laugh not not so much on the size of the processors, but the amount of data and size of the operating systems. I get headhunted from the UK to help Chemical Bank start with virtualization in 1983. We ran a production credit card approval system that had some 300 telephone operators waiting to receive calls from merchants, and look up a credit card account. It ran on an 8MB MVS system, we virtualized it and ran it under VM/370 HPO 3.2 on a 12MB system. My Android phone now won't run effectively on 6GB. In the late 1990's one of the big pushes for extended memory on these 3090 systems were the 30-50MB CAAD drawing and modeling files, now we routinely load up files that are hundreds of megabytes.
FYI - If you don't know about Hank Kee he still has a radio show on WBAI. He has a website with show archives here hankkee.net/
I'm still here after 40 years. A lot of this heavy metal was in the banking industry where I started. As big as these appear to be, you haven't seen anything if you haven't been in a big rack-mounted server room today. Makes these look a little on the "small" side.
How did they get Princess Di to demonstrate this system?😀
I was an IBM Field Engineer specializing in Large Systems (now known as Enterprise Systems) from 1977 to 2015. As I watched this a lot of memories came back. Thanks for posting.
Wow, you must have seen a lot of change!
If you have or see any old IBM training laserdiscs, I'd love to be able to convert them for others to see. I'd really love to find something on the 3850 Mass Storage (cartridge) System.
@@ian_b When I first started in 1973, the mainframe system I worked on, and IBM 360/40, only had 48Kb main memory to load the OS and run programs. A NEST doorbell/Cam has a greater amount of memory and faster processor than the 360/40. A version of the 360 went up in the space shuttle, provided the computing power in 256Kb main memory.
Sounds like we have something in common. I was there 1979 to 1998. Great times!
@@MarkCathcart I have found some pdf training and a presentation that runs on Lotus, but no videos. Yes the 70's and 80's were a great time to be in IBM. Started declining for the employees in the 90s and went down hill fast after 2010.
I used to fix this. PST - Product Support Trained level. Our IBM office bought a refurbished 3090. Due to anti-trust laws, we could not get special pricing on any new unit. One tri-lead paddle card was damaged. We identified the signal and it was a sensor line not a high-speed data line. I had good eyesight in those days. I soldered the connection. I don't recall we ever replaced that cable. Then I got posted at a remote site and never came back to the mainframe group. It was a fledgling team that handled Point-Of-Sale and RS/6000 which was a blessing since this introduced me to AIX and I was then trained on the 9076 supercomputer. I ended up being the local GOTO guy for the 3584. I was also trained on the 029 card punch but sucked at it. Lots of war stories. EDIT - it was not a tri-lead but a triaxial cable which was covered in blue foil and about the thickness of a pencil lead.
Thats awesome, I had so much respect for real hardware engineers, it was a world I could get totally intimidated by!
@@MarkCathcart I saw the writing on the wall and following some advice - left the party early. The moment I was eligible, I went for early retirement. My colleague didn't. He was from IBM Canada and worked here as contract staff. He wanted to reach 50 years working for IBM but passed away on his 49-th year. The prestige of the IBM Customer Engineer had evaporated. We were headed the way of dinosaurs. We never met the client anymore. All we met were the staff of megacentres where the clients had outsourced their servers to. We had descended to the level of janitors. I am glad I never went back to mainframes. Those team members were getting arrogant. Case in point - A newly installed 3584 never finished initializing. We were territorial. 3584s in their account were handled by them. I was asked to assist. I went there with a trainee and within a minute, I pulled off a Richard Feynmann radio repair. I was never trained on the 3584 and learnt everything through reading the Maintenence Information Manual and trying to reverse engineer the machine. I deduced that they used an FPGA to act as the controller. I knew exactly where they went wrong. They installed the machine without referring to the MIM - it was still unwrapped. I told them where they went wrong and even found the page in the MIM but they still refused to admit their mistake. They mounted the crash stops at the wrong position and the robot couldn't find the barcodes and home flags. Reposition those and voila! That really impressed the trainee. Guess who knew shit and who knew his shit. EDIT - I must be losing my memory. What happened was the crash stop contained the home flag and when this is misplaced, the robot would find home, move the preprogrammed distance and shine the barcode reader to read the barcode labels and couldn't find any. This stopped the initialization. Pretty obvious if you did any programming and diligently enforced input conditioning and sanity checks. I published two programs on github. These leave audit trails so I debug from afar. Since I was obedient to the manual for everything, I never encountered this problem but having watched the power on sequence countless times, I realized this instantly even if I had never come across this before. With my knowledge of AIX/unix, I used "strings" to extract the latest HEC and HECQ codes from the microcode file along with their descriptions. Our technical expertise had drastically fallen. NOBODY was interested in this finding. Nobody drilled down to the error codes anymore. Another case in point - a REGATTA was gutted for its parts and when the crisis was past, new and original parts were installed but the machine failed to power on due to a massive mismatch in the VPD - Vital Product Data. Two CEs failed to overcome this. I was the second standby and was called in. I have absolutely no training on this machine. I can't even remember the machine type. But I could go into the error codes and painstakingly identified the serial numbers and error codes and after rearranging the power supplies, it could just power on. We needed to fly in someone from France to sort the rest out. The company hides so many things from us ostensibly to prevent these machines from being serviced by third parties. Indeed my very first manager betrayed us by taking along a curated group of CEs and SEs to offer MA to government clients. It was only at a webinar late in the game that I delved into the documents supplied for the 3584 to find that there was a secret menu that would exercise the robot and cause it to randomly rearrange the cartridges. All along I had to rely on tapeutil commands to perform the same task. Major BUMMER.
Thanks for this!
In the late 1980's, I worked in American Express's mainframe center in Phoenix. We had a total of 7 IBM 3090-Js (IBMs largest as I recall). In the same center, there was a vast ocean of DASDs and dozens of smaller mainframes as well.
In 1983, I worked for Chemical Bank in Manhattan, before we got 308x computers, we had 3033MP systems, they were even bigger than these. My experience virtualizing the SWIFT banking system for Chemical Bank lead to me working at IBM Kingston on the firmware/supervisor design for the 308x processors. I have a bunch of these videos on laserdiscs, I've agreed to provide digital copies to the Computer History museum in California.
I wanted to be one the IBM Project Managers on the AE Data Center in Goldsboro in 2011 but it did not work out. I spent years at MCI with the back of my chair up against a 3090-600. They were my first customer to cross the gigabyte threshold on 3390s.
I worked for six weeks in Japan as RSFSG. We once toured a "small" customer. There was a dedicated office for an IBM manager and the CEs. One floor for printers, one floor for DASD, .........
I started in IBM as a part timer in the late 80's. As a young man my job was to pull the Bus & Tag cables out of the 3090 and pull out the cables from under the raised floor. There were a bunch us part timers and what great memories I have. It was non stop work to get the old system out and the new system in. Worked in downtown Brooklyn at Brooklyn Union Gas. The experience IBM's were great to be around. Professional and well respected. They were great educators to all the young people trying to learn.
Raised floor, heavy duty air conditioning, blue/white cabinets…. Those were the days!
Well, we still have 1 & 2, right? :-)
Raised floor is to hide bodies.
I’m surprised there is no mention of the vector facility in the video. My company, Amoco Corp, used this system for seismic data processing in the late 80’s. I worked on vector enabled scientific subroutine library for the system (Fortran)
I think that is because there were entire different training programs for dealing with the vector facility. They were not that common at the time. Good spot though.
@@MarkCathcart Thx. I’m not sure which frame would have housed the VF anyway, I was never in the room with the mainframes 🙂
@@zrlw05 there is a copy of the ibm redbook "IBM 3090 VECTOR FACILITY" if you to look it up...
TECHNICA REFERENCE
GG on bitsavers
Hi Mark. I was a FE and mainframe specialist for IBM (hardware) appreciate seeing this video. Brings back memories!
I worked for IBM from 1987 to 2008, eventually ending up a member of the Academy of Technology and a Distinguished Engineer. I still don't understand hardware, which is why I found this laserdisc so awesome. The best people, and some of the most influential I worked with at IBM were the hardware FE! My first big break came in 1979 when a FE gave me the source code to the IBM 3741 so I can hack in a change for the company I worked at the time. We needed to be able to transmit a single record insurance policy from a remote office to our London office quickly, the insurance record required 140-bytes, the 3741 only supported 128. I patched the code to allow the 3741 to split the record but transmit 140-bytes which avoided a batch program at the HQ to re-assemble the records before adding them to database. Seems trivial now... 6-years later I was working for Chemical Bank in NYC, we used a mod for the IBM Series-1 to do much the same thing, except this time to do online credit card authorization automatically for the first time. Amazing times.
Did you ever fix 3420s? If so, how did you you do the capstan squaring check/adjustment?
@@elye3701 oh I did. Oscilloscope, test tape and lots of time... I don't recall the exact process. Many years ago
@@mrkkitson I came up with this procedure and it took over the official oscilloscope 50%-50% waterfall technique - but only in my country. I plugged my digital meter to one capstan sensor signal and slowly turned the capstan. I noted the lowest reading and the highest reading. Add the two numbers and divided by two. Then I ran the capstan motor by jumpering some pin or maybe plugging the capstan motor to another socket. I then adjusted the pots until the reading matched my calculation. The meters at those days was average reading instead of true RMS. This spread like wildfire BUT no technical manager nominated this for a PMR to help the rest of the world. I was a science grad not an electrical/electronic grad like the others. I applied my knowledge of class-D audio amplifiers.
I was a part timer In the 80's for IBM and they used the part timers to pull the cables under the raised floors. Talk about heavy Bus & Tag ( I believe it was call ) cables. I still have a processor from a 3090. I also have a memory module from an IBM 1800 system that used Donut memory we called it but was magnetic memory, Great memories from those days.
Talk about a blast of the past. I was a computer operator for several years.
My working environment from 1969 until 2000.
Okay. I need to continue the course....
Omg, completely different from what computers are now. I have an interest in computer history but this is whole new field.
I like a CPU you can walk in.
I worked on an IBM 370/145 circa 1975, the rear part of the main processor was empty. The space was reserved for additional memory which would be present in a 3670/148. So we used to keep our food warm in there when working on night shift.
one thing is the incredible amount of engineering that went into making the computer, the other thing is the equal amount of engineering that went intro trying to squeeze the whole thing into as small a box as possible, even though compared their later creations it's still massive. And these still need a cooled room despite having their own water cooling
I've worked on 3090 and after on ES/9000 IBM and Hitachi HB100 in Banks Data Center in Italy, since 1989 to 2000. It's amazing to see again those big big computers
The interiors and hardware of these old machines look so complex.
They were. Even the single frame older 370 Processors had some amazing design, the cables were not just put in where there were spaces, they were designed and sized to fit in exact places. Thats why they were all so expensive...
Thank you for sharing this nostalgic stuff. I remember viewing this training in a computer room.
The engineering is amazing-not just the design of the logic, but the physical arrangement of the components. Where do you put the power supplies? Where do the CPUs go? How to you route the cables to connect everything
Yes, it had to be, these machines were far from perfect but the complexity of the design, engineering and operating system software support was suitably impressive for the time. I'm looking for some more interesting videos/VHS/Laserdiscs to show off the technology of that era. Thanks for watching.
there were a lot, and I mean a lot of cables under the floor interconnecting the components. the 3090/400 was two 3090/200 interconnected with only a few common components.
The machine itself had firmware called "Logical Partitioning" which would allow you to virtualize the hardware to a point, and share hardware components between a number of partitions by allocating dedicated resources to the partition. The entire machine, or one side, or one or more partitions could run the Virtual Machine operating system, which could further share resources.
We still have a IBM 3090 running, it has 4 CPUs boss refuses to move on.
damn, where in the world is this?
Your boss (or his replacement) will rue the day.
Thanks for the upload! That was interesting.
Damn another 3090 that is not being sold anywhere
When you want the job done, run a mainframe.
We had a 3090 in the Uni, I was a user from 1992, but at some point it was mainly replaced by a number of IBM AIX servers.
3090 was an amazing machine. So fast and so reliable..
fast? u wot m8
It was amazing in it's day and the Model J's were bulletproof
@@mrkkitson well back in the day, lol. 69 mhz isnt exactly speedy today
@@ttuori8592 of course! The mainframes real strength was it's parallel processing capability
@@mrkkitson yeah, the amount of ram and storage for the era is nuts
Wow 2 very cool things.. LD and an IBM
thanks
LOL, I remember watching this video/optic disc when learning the 3090
私は日本人ですが、昔IBM３０９０システムの保守を担当し、日本全国の機械を直してきました、とてもなつかしいです。
秀夫。 3420を修正しましたか？ もしそうなら、どのようにキャプスタンの正方形をチェック/調整しましたか？
なにをいているのか分かりません私をただIBM機械を直し歩いただけです。
You mean these three huge boxes comprised a single computer? This is unreal lol
to be fair, these three boxes are just the processor part of the computer. There would be other boxes that contain the "modem" communication processor, disc storage boxes(lots of them) and other things found inside modern PC's and laptops.
@@MarkCathcart Why put the computer in a whole room? Just build the room into the computer already hahahah it's amazing how small and efficient things got.
@@michrain5872Keep in mind, this was 1985. Microprocessors had been around for 10 years, and were slowly catching up to high-end minicomputers. 16-bit machines were making headway amoung the cheaper and more widespread 8-bit micros, and they were starting to talk in MB, not KB. 32 bit processors were only a few years away, and close the gap in terms of capability, if not raw performance.
I remember installing the first one in Ontario Canada. A lot easier that it's predecessor's 3033 or 3168.
I worked on both a 3033MP and a 370/165. I wasn't around for the installation of either but the deinstallation of the 3033MP was something to behold. The 3033MP also had a dual 3850 MSS. Those were the days, I still have some of the cartridges as souvenirs. 100MB per cartridge.
Did you ever fix 3420s? If so, how did you you do the capstan squaring check/adjustment?
Surelly this system could have been built in a way it would occupy less space. It could even be more faster and have more capacity. But I suspect that IBM systems engineering had other goals on mind. Like reliability, availability, fail-safe features. One key factor to achieve these goals is temperature control. And power supply quality and reliability. To achieve goals like those, you have to avoid crowding modules and componentes (beyond some limit), you need space for airflow. Those were some of the challenges on the late 80's hardware engineering and frankly the customer targeted by this kind of system did not care about some dozen floor square feet more.
You are ignoring or don't know this was a replacement for the IBM 3033 processor. The 3090 required 50% less space in total for more than double the performance per sq-inch. These systems, as the 3033 systems did, required water cooling.
Remember these machines would often require a large installation no matter how small the actual computer was. Between electrical, cooling, archives of storage media, terminals, etc etc. It probably could have been smaller as you say, but shrinking it wouldn't have been high on either the customers or engineers priority list.
For about $3 Million in 1988 you got a processor that ran at 66 MHz, delivered 67 MIPS, and 130 MFLOPS. In MIPS this compares to a 486 @100 MHz, in FLOPS it compares to a Pentium 133.
It's a fun data point, but as someone who worked on these machines in real commercial environments as well as later graduating through IBM as a Systems Engineer, to Distinguished Engineer, we all knew it was never about MIPS or FLOPS, it was always about data, I/O, concurrency and virtualization, and yeah, reliability, almost none of which you get could from x86 architectures until the early 2000's.
@@MarkCathcart I've always been a workstation guy. I was doing some orbital mechanics simulations in the late 90's and using a Linux work station as a terminal for one of the schools mainframes. As a joke I ran the sim on my "terminal" and it only ran half as fast because I had to cache to disk and could not fit all the data into main memory. I finished the project on my workstation and saved a lot of charged CPU time on the mainframe.
@@AndrewTubbiolo Yup. I was a developer on the mainframe side of the worlds first home banking system PRONTO, the pre-cursor for .zip files was developed in part for us, .arc files by Thom Henderson, and I wrote grid-like programmers compilation system that allowed programmers to submit batch compilations on the mainframe, which were offloaded to PC's running MicroFocus COBOL.... heady days all between 1983-1986. One of the managers I worked for was Hank Keyes, who won a PC Tech Journal gold medal the first year they did them, other winners that year, Bill Gates and Phillipe Kahn of Borland... I went on to join IBM, and wrote the first Linux mainframe strategy paper in 1997/8. Hence my interest in this stuff... still.
@@MarkCathcart Thanks for the stories. Great stuff.
@@MarkCathcart a friend of mine worked at a bank and was lucky enough to grow from helpdesk to first mainframe and then unix groups. i remember there was one time in his early years when he got curious and ran seti@home on a linux partition - but only for a short time till IBM called them up since he created a messy overload on the whole system. we both laughed since it was such a sad, unprofessional first impression to create of linux as a citizen in the system he created. naturally it was without bad intentions, just wanting to see how amazingly big that machine was - like when you'd stand in the largest church of the world and you look up and say "woahhhh". and it was big - but the scheduler was not amused nonetheless.
IBM's build-quality is visible on the video ...
IBM built things well
Not only well built but excellent support in the field right back to the factory. System down situations were rare but Montpellier could send engineers if the field CE's couldn't fix it.
Seeing this brings back fond memories.
Makes you appreciate a laptop!
Wow, amazing video, thanks very much!!!! From what I've seen from the cover, this laserdisc has two other contents, would you also upload those? I would LOVE to watch it too!!!
As soon as I have time, my plan is to donate all the contents to the Internet Archive. archive.org
@@MarkCathcart alright!!! yesss!!!! please let me know here, or is there a page that I can follow?
@@raphaelmsx I don't know yet, I have the videos done, I have to complete a submission form and once approved submit. I have some lserdiscs on early IBM PC training that are all part of the same submission, so it's taking longer than I expected.
Those cables looked like Bus and Tag cables.
I try to imagine what the internals look like in a Z-15.
Actually a lot simpler than these. So much now is on board the actual processor and SOIC which improves speed and removes latency and wiring both on and off the "system board'.
Why does audio sound like recent recording?
Laserdiscs had digital sound but an analog picture. I have many music laserdiscs films in jazz that I've published on another channel @ctproduced - the sound is as good as, and sometimes the same as CD quality.
Hey I want to get on the internet, yeah you need one of these.
Wow when roughly was this footage likely recorded?
I haven't found any info but the 3090 was announced in 1985 and the hairstyle looks late eighties so I'm guessing circa 1986 until I can find out more.
This would have been made in 1985. It was internal processor support education, so would have been used during that year along with "Fist Customer Ship" (FCS)
Wow interesting video, do you have more you could upload.
To operate the IBM computer back then you had to be a computer yourself
hey, I have one of those in my PC
Interesting.
can it play doom?
If you have the source code, it can certainly be ported and would work. That said we were playing Adventure on three generation earlier mainframes in 1976....
@@MarkCathcart your a dreamer silly little dreamer oh you got your head in your hands oh no oh no I say dreamer Doom needs more processor power then that old beast can muster it can only do a a few hundred thousand calculations per second and Doom needed at least 25 to 33 million calculations per second on a 486 to at least run Doom and faster 486s were better for frame rate from there so nope because it would take minutes per frame and you would get killed before you could see it coming sadly
@@raven4k998 But to be fair, you couldn't play doom on anything else at that time. So good news, no one gets killed!
@@ctproduced yeah and that's the funny part the old IBM is way to slow to handle Doom as it only does a few hundred thousand calculations per second the first personal computers made that system look like an ancient slow pos computer and they were a fraction of the price which was the best part to because the c16 could do over a million calculations per second on it's CPU it goes to show how they were obsoleted by commodore and apple back then can you imagine buying a commodore pc and realizing you paid so little and got a computer that kicks the crap out of that huge massive IBM pc you would have been on cloud nine back then
i wonder if the presenter was also the voice & and knew the system - it appears so.
Almost certainly so. One of my good friends from my time at IBM Poughkeepsie worked in information development, she often did voiceovers for marketing and training videos. I still have an MP3 from a video for object oriented technology in 1994... she wrote many of the manuals and definitely understood the technology.
👍
imagine buying one of those then getting a commodore 64 only to figure out how much more powerful the commodore 64 is over that beast you spend million of dollar to buy🤣🤣🤣
that would be a kick in the teeth🤣🤣🤣
junk
@mehmet: what are you basing your comment on?
@@deepsleep7822 I think he means his opinion is "junk". At the time these machines were state of the art, the biggest fastest processor in the industry. One man's opinion won't change that anymore than all those people who said mainframes were dead in the 1990's. 30-years later they are still here and running many of the world's core business apps...
@@MarkCathcart : actually Cray’s were around at the time. They were faster, but weren’t general purpose. But, I get your point.
This computer was over-big and over-bulky even in its day. Compare to any of the Unix vendors of the time and you can understand why IBM declined so quickly.
That's simply not true. When the 3090-200 and 3090-400 were introduced in 1985/6 Unix systems were nowhere. Sun Sparc wasn't even released until 1986/7. Cray and OEM IBM systems were the only effective alternatives. Cray pretty much only targeted engineering and scientific, and the OEM systems were for the most part running the same IBM software. I agree that UNIX systems and CMOS processors made rapid progress during the life of the 3090 processors and that's why this was effectively why this was the last major range of water-cooled multi-frame IBM mainframe systems.
You are wrong, of course. As an old IT hand who worked on both kinds of systems, I saw it first hand. Look at those things. They're the size of small city buses. In terms of computing power per cubic foot, IBM wouldn't catch up for another 20 years.
@@igfoobar as a former Distinguished Engineer in IBMs Systems Group, and a member of the IBM Academy of Technology, I make mistakes from time to time. So, go ahead and tell me what UNIX systems there were in 1985/6 that could even run transactional workloads let alone scale to a similar capacity at that time?
FYI in 1984, I was part of the ad-tech group at Chemical Bank in New York City, we were literally desperate for faster systems than IBM Could provide, we'd gone from a 3033MP system which was even physically larger than the 3090-400, through the 308x series and we'd just rolled out PRONTO the worlds first home banking system. If UNIX systems were as good as you say, we'd have been all over them.
So, which ones specifically? I'll wait.
@@MarkCathcart "I'll wait" is a tell for cognitive dissonance. Enjoy your altered reality if it makes you happy.
@@igfoobar Really, go ahead... you are off by a decade at best. Unix Clusters couldn't do what the 3090 mainframes could do until at least 1996. My guess is you are confusing the IBM 4300 mainframes. So, what Unix vendors?