one of the big items of tech news in the last year or two has been that Apple has started implementing their own silicon previously all of Apple's computers including the Mac series all used processors that were not created by Apple for example the Apple II used the 6502 first series of Max used 68 000 chips from Motorola then they used power PC chips also again from an alliance of other companies and Motorola then the jump ship to Intel for their mobile devices they've been using chips designed by armor manufactured by Third parties for them but now for their mobile devices and the Mac liner machines Apple have started to design their own processors having licensed just the arm instruction set rather than the hardware design and there has been much fuss about it in the tech press while the tech press has not mentioned though whilst clearing several times this is the first time Apple's ever done anything like this is that apple as well done something like this before or at least they tried yep it's time to talk about project Aquarius before we get stuck into the details of this thing just to avoid any confusion this has absolutely nothing to do with the Mattel Aquarius that's an entirely different machine that did not involve Apple in any way shape or form no what Apple was going for was somewhat more technologically sophisticated than the Mattel Aquarius well to be honest almost everything's somewhat more technologically sophisticated than the metallic areas I mean the Albanian State washing machine company is technologically more advanced than the metallic Aquarius project Aquarius was there to create a new CPU for Apple that's why they were happily still designing machines around the 68 000. the new the 68 000 row map as well coming to an end and also in the world of Unix workstations they've been a huge explosion of risk processors practically every company that made the Unix workstation created their own processor heck even Acorn the people who created the BBC micro they created themselves one heck of a risk processor so Apple kind of figured to itself but why the heck can't we do this if everybody else can we're bigger surely we can manage this now we all know in the end that Apple didn't use their own precious design they went for the power PC which might make you think that maybe they didn't try with this project maybe it was sort of like a half-hearted attempt but absolutely not Apple put some proper resource into this by the time Apple decided to cancel project Aquarius they had 50 people working on the CPU now that may not seem like a lot but let's put that into context with other processors that's one order of magnitude more people than for example Acorn had working on their arm design it's vastly more people than Sun had working on spark it's bigger than the entire research group that created risk in the first place heck it's more people than even the big bad corporate world of deck had working on their Alpha processor in short this was not a half-hearted effort Apple really did put some serious resource and money behind this and so far we've only talked about the Staffing there's a lot of other areas they spent money on in the CPU which we're gonna come on to later but I just want you to keep in mind as we discuss this that this isn't some little side project Apple threw everything into this that they could do speaking of resources being thrown into stuff this is the point where I mentioned this video is sponsored by PCB way find purveyor of printed circuit boards and also pick and Play services CNC Machining 3D printing there's quite a lot of things that they do they have a website you should probably go look at it when project ~ Sponsored Segment Removed ~ Aquarius was first greenlit it was a time of Fairly significant change for Apple Steve Jobs had just been pushed out and John Scully had just become the CEO with Sean Mary glassay being put in charge of the Mac product line so those are the two people that Sam Holland needed to persuade that Apple could design its own processor and now in designer processor but LeapFrog both Intel and Motorola now the initial idea Sam had for the design of the processor was that of a stack processor and it's with this concept that he managed to get the project green lit now stack designs are a bit different to regular CPU designs in that there's no cash in this thing what you have is something surprisingly a stack and values Rider pushed onto the stack or popped from the stack back into memory the processor itself doesn't have any instructions that does anything outside of the stack any values for any of the Opera and choosing any of the instructions are from the stack it doesn't directly go at a memory address for example but this means that software has to take care of loading any values that are going to be needed onto the stack and popping anything that needs to go back into memory from the stack now there's a reason you haven't seen many Hardware Designs that work this way and we'll come on to that in a second where you do see designs like this is in Virtual computers things like the Java virtual machine for example where you have P code or portable code a lot of them do use a essentially a stack design now the reason why you don't see many of these in real Hardware is it's an absolute nightmare for the people who are going to write the compiler and that's what actually halted the idea of using the stack design in Project Aquarius the team who is responsible for writing the compiler basically told them which is another thing they were going to be able to do a 17 compiler wasn't the thing that they were going to be able to create so unlike Intel when they were told that when they were designing titanium Holland and team decided that they would change how the processor was going to work rather than just try and force the compiler team to do the impossible in fact the only Hardware implementation that readily Springs to mind for me is that of the transputer and with that particular design it actually started with the language Arcam and then created the compiler and the instruction set for the CPU after that and finally the CPU itself actually the whole transmitter thing is absolutely fascinating at some point I really should do a video about it but anyway we should keep talking about Apple this is where the design switches from being a stack machine to a risk processor and as this proposed risk design is the one that we have all the details for because you may have noticed there's not a lot of stuff out there about project Aquarius and there is a reason project Aquarius was supposed to be secret now as you're about to hear Apple had some problems with the secrecy of this project and that's because it did some incredibly unsubtle things on Apple's campus if you remember earlier I mentioned just how much they were spending in terms of hiring people and having people work on the project and said they also spent some money on some other very significant resources as well well it's now time to discuss the giant purple cray supercomputer now if you're trying to keep a thing secret maybe purchasing an incredibly large supercomputer isn't the best way of hiding that secret people are gonna notice that sort of thing especially if you're the only company in the world that has one that's purple now crazy computer is not exactly the thing you can just pop down to Dixon's in order normally the process of buying a super computer takes years and most of the organizations that buy them tend to be governmental and have long procurement change to go through apple on the other hand may have broken the world's supercomputer purchasing speed record by turning it around in a few months apparently with CEO John Scully being a person who phoned gray sales team and started the whole process rolling well if you think they did quite well purchasing a supercomputer that fast well that's nothing to the speed of which they managed to get the damn thing installed now if you think a crazy supercomputer is Big well they're dwarfed in size by their own power and Cooling Supply systems to the point where Apple had to dedicate and fit out a whole building on site just for the supercomputer and they managed to turn this around in six weeks which is an incredibly short time frame and they did that by having three crews working in shifts that would work seven days a week and 24 hours a day and it did it in six weeks now renovating the entire building on campus in such a short time frame with so much work having to be done and then sticking a giant purple supercomputer in it is the sort of thing that everyone on site notices and this is also you can run your secret CPU project so Apple has first come up with a cover story and they said they bought the giant purple supercomputer installed it high speed so that they could do some interface work for the next generation of Macs now I'm hoping that's a story that almost no one on Apple's campus bought whatsoever and if they did hopefully they didn't work in the bridge purchasing department now there's a few other details about this crate that are also quite interesting you may be able to tell quite like gray computers Not only was it purple and installed and purchased in a record time it also was the first one they shipped with a version of Unix on it prior to this machine crane machines have been all about running Fortran in fact that's basically all it did it ran Fortran programs it just it ran them very quickly because it made use of Grace Fortran compiler and that compiler would take code and optimize it for craze Vector instructions and this is where you essentially give one instruction and a whole bunch of data and it carries out that same instruction on all that data in parallel that's our great supercomputer got its speed by parallel instructions well Vector instructions to be precise now what Unix and the C compiler for this platform didn't do was make good use of vector instructions so you may be wondering why on Earth did Apple choose to have Unix put on this thing well the answer is all their chip design tools were written for Unix so they needed it to run Unix and run it quickly and run units quickly it did not in fact the engineers discovered that the sun machines that they had connected up to the Craig actually ran the chip design application about as quickly as the cray did and although a sun workstation could be quite expensive it wasn't 15 million dollars expensive yeah apple paid 15 million dollars for this crazy supercomputer and if they were the Department of Defense the department of energy that we're doing probably couldn't quite wise spending but to run a chip design application not so much I'm gonna give Apple the benefit of the doubt and hope that they were thinking well for chip simulation it'll be really good just by the looks of it they never really got Ram to doing that part in fact this poor crane machine ended up spending most of its life being used as a system to help design cases for various Apple machines a task that in no way you needed to create super confusion to carry out so with its collection of a crazy supercomputer a number of sun workstations and a few apples and of course a couple of ax machines and it was the 1980s project Aquarius continued to come up with what a quite sophisticated risk design now the reason we know a lot of details about what was a secret project I'll be its secret project with its own building and the giant purple supercomputer is because of elite specification documents well leaked or it was put it onto the internet archive many years later and no one cared that specification document was for a CPU code named Scorpius which was aquarius's risk CPU and it is a very ambitious CPU because architecturally this thing has what it refers to in the document as independent Processing Unit or pu processing units which is an idea we know today as cause yep in the 1980s Apple was working on or designing a multi-core risk CPU and that don't just mean that these things are a bit like modern course now this really is exactly the the same idea each processor core had 16 general purpose registers the same number of general purpose registers armed us each core also has seven local registers and there are eight Global registers that are shared by old CPU cores there are also instructions that allow these cars to talk to each other and then there are other instructions that allow CPU packages to touch each other so you could have created a multi-processor system with this as well and these instructions are necessary so you get cash coherency across the system because if you didn't have that what ends up getting written back into main system memory would kind of depend on which core or processor wrote to memory last as changes that have been written out by one CPU core to memory might get completely dropped because the other CPU core doesn't have that change in its cache when it makes an update to memory in fact guaranteeing cash coherency is one of the main testing tasks that go on when modern CPU designs are being tested the other feature they had in the specification was something that at this point we've not seem inside micro purchases yet although we had seen them in supercomputers although we probably shouldn't be too surprised about this because Sam Holland's background was in Super Computing the processor was intended to feature single instruction multiple data these are these Vector instructions we were talking about with the cray earlier and there's something we'd not see introduced into the world of PC processors for example until the Pentium MMX came out and then much more usefully the SSE extension oh that's going to get me comments from the Intel Fanboys or they are slightly preemptive and go yeah okay the MMX was better than nothing just all these features if they'd actually got the ship out would have put the CPU way ahead of its competitors as it contains a lot of architectural features that are now standard in modern chips but we wouldn't see for nearly 20 years if nothing else did a pretty good job of predicting what the future of a CPU is going to be like but as you're not all running apples based on the Scorpius CPU yeah no the project didn't end well one of the reasons since all these computer companies have been able to create risk processes and run them at pretty rapid clock speeds whilst the design Simplicity everyone else's designs were pretty simple from the first ever risk processor up until the likes of spark and armor mips Scorpius was very far from simple this is a really complicated design with core sharing things like the memory management unit cash you got synchronization instructions Global registers there's a lot going on in this chip which means there is a lot to design there is a lot to get wrong and then fix and then redesign and all of this just burns through resource resource that Apple was maybe struggling a little bit with in fact Apple was about to go into a decade which you see it almost go bankrupt but it seemed a big turning point for the project it was when Sam Holland left apple and apple brought in Al Alcorn to replace him and yes that is a name that will be familiar to some people who are watching this and it is the same Alcorn who created bong but by this time in his career left tari behind and become a fellow at Apple so Al was put in charge of the project and they must have felt a bit like being handed a poison chalice because you handed the project that people are beginning to think of as failed the previous guy has left and you don't really have anything to prove so it goes well it's not exactly gonna bolster your reputation so much because you're the dude who invented Punk now it seems alcorn's view was that despite the very very bright people that they had working on this project it was probably not something that he could rescue but he did bring in Hugh Martin from bridge computers Ridge computers have created a very early 32-bit risk CPU so bringing in someone who'd already successfully designed a risk processor and got it into production and use would seem like a pretty good step for helping move everything along but it would seem Martin's view of the project wasn't exactly positive to the point where he apparently described the scorpius's designed to both Alcorn and John Scully is ridiculous and suggested that maybe Apple should concentrate on its straps which at that point in time seem to be creating ever more beige boxes than try and compete with giants like intelin Motorola now it seems whatever Martin said seemed to either chime with what both Scully and Alcorn also thought or his opinion carried such great stock that they decided to cancel project aquarius so this left Apple still in need of something to replace the Motorola 68 000 as well then you processor it turned out to be a dead end so Hugh Martin was put in charge of Designing a Mac based around an existing risk CPU and the one he chose was Motorola's 88k yeah you thought this was gonna go power PC but no not yet and this new machine was codenamed jaguar and like all pieces of electronic equipment codenamed Jaguar you know it's going to be a failure in fact the ataria Jaguar is now considered the SI unit of failure one of the problems with the CPU they chose despite also coming from Motorola it was in no way compatible with the 68 000 no existing Mac application software would run on it and the operating system wasn't going to either so they'd have to come up with a whole new operating system for the Mac which they started to work on it was codenamed pink however a different group of Apple Engineers decided that they would tackle creating an emulator for the new hardware that would run the old Mac OS and all its software and it turns out that emulator was a lot faster than most existing Mac computers but Apple decided in the end they weren't going to use the 88k and as Motorola's only other customer for the 88k was Ford and Ford then decided rapidly that they weren't going to use it well the 88k ceased to exist now at this point Apple explored two other processor options for their machine the arm the mips practices now at the time Apple actually owned a share in arm and they really could have shortcuted quite a lot of their Journey if they just jumped to arm at this point but no that was not to be and that's because IBM turned up they'd heard Apple had tried porting their operating systems to Intel at one point the Star Trek project and wanted to know would they put Mac OS to their new power PC 601 the same chip they've been using in their RS 6000 workstations IBM even offered to help finish off this new pink coal base which is going to be the base of the operating system just as long as Apple would adopt power PC and it seems that offer was very tempting to Apple because they took it and formed aim the group that contained Apple IBM Motorola and as they say the rest is history so I'm gonna briefly engage in the think you probably shouldn't engage with what ifs if Apple had continued this project and actually managed to deliver it we would have got what is essentially the common CPU designed today more than a decade earlier than we did and why undoubtedly he would have been regarded as a very very good processor it may also have been the thing that killed Apple at this point in time the problem for apple is they just about to survive this period I mean really by the Skin of Their Teeth the 90s was really not a good time to be apple in power PC came out at just about the right time that they needed it if it had been much later that could have well been the end of the Mac line of computers and the Scorpio CPU why it may have been great they couldn't have got that finished in the same time frame that they managed to get a power PC based Mac out there and also the cost of r d and taping out manufacturing a CPU is huge and Apple after all its capital reserves and then started selling off assets to keep itself afloat starting in around this time so if those assets and capital reserves have not been there because the whole CPU designing thing and apple could have easily just tipped into Oblivion now admittedly if they'd had a better CPU than power PC they might have been there to sell a few more Macs as well so we will never know how this would have worked out but it's not as simple to say that as if they'd finished this they were doing a better Mac therefore Apple would have done much better we don't know that and the capital requirements to build that chip well that may have left Apple without the reserves to make it through there let's say Basket Case years I'm not going to talk anymore here about that period of Apple's history in fact one day I might well do a video on it but that day is not today well hopefully you've enjoyed today's video and for some of you it's answered that question of what was that weird Gap in their history between the 68 000 and moving to power PC because it always seemed like there was something missing from that story and it turns out there was a secret project involving a purple cray supercomputer as ever if you've made it to this point I'd like to say thank you very much for watching and if you did enjoy the video please use that thumbs up button that YouTube's created explicitly for the purpose of indicating that and if you'd like to help the channel out why not hit the Subscribe button because that makes YouTube actually tell people that this Channel and its videos exist thank you Back To Top