the A3000 the last ever BBC branded computer made by Acorn and clearly a perfect fit for my endof thee line series of videos where we talk about the last of a line of computers to be made except this machine's a bit tricky because it's kind of also not the end of the line as well either it's more like the beginning of a new line as many people believe that this is the first of the Archimedes lines of acor machines except it's not the first of the Archimedes machines because there were at least two other lines of machines before this one came out in fact this is just the first one you happen to be able to buy in a shop and the acorn you know marketed to anyone which means that it's the first aredes machine that most people got to try and experience and therefore they think of this one as the first cuz it was their first so welcome to the end of the line for the BBC micro SL or not oh this is going to be a complicated one so before we get into the nitty-gritty and the excessively nerdy details about chipsets I'm going to say that this video is sponsored by PCB a fine Pryor of printed circuit boards yes you Des sign a thing you upload the thing thing comes back in posts as the name would suggest they are primarily focused on pcbs but they also do CNC Machining injection molding 3D printing still no news on knitting patterns though sadly so the big thing everyone knows about the A3000 or certainly did at the time that it's based on the arm processor it's not the first machine Acorn shipped with the arm processor in fact we have the 400300 machines came out first but this is the first time that normal people could lay their hands on the machine with an arm chip in it I mean you can March down to cures and you know buy one of these things and for people outside of the UK cures is a big electrical retailer and the arm processor was something that acor designed inhouse as well as a number of support chips for the processor like MC vidc ioc all of these would become core components of the Archimedes range of machines including our A3000 here now they'd refer to this as the 10x machine a machine that would be 10 times more powerful than the original BBC micro while hitting the same price point and okay they didn't actually get there the AR is not 10 times more powerful than the BBC it is a lot more powerful than the BBC but they also didn't get anywhere near the price of a bpc but there are a lot of very clever design decisions in the arm processor itself and its supporting chips that helped reduce the cost of this thing down significantly what really kept the A3000 price above its contemporaries like the Amiga and the St was just simply the likes of commodore and Atari had better economies of scale because they were shipping more of these things Acorn and the BBC's biggest presence was in the UK cuz that was their home Market but there were not enough machines being sold there to help drive down their costs now one of the things everyone seems to know about the arm is that it was a much faster processor than other contemporary processors in that space at the time and I'm going to give you an idea as to how fast that is cuz sometimes I think people exaggerate it a bit too much so for example compared to a 68k clocked up the sort of speeds it was in the Omega 500 it's about 7 to eight times faster than that we're about three or four times faster than say F 386 clocked at the typical frequencies it was at that time and we're about say quarter of the speed of a spark processor from this period like a sun 4 so compared to a lot of machines particularly those in a similar price point we were substantially faster than those CPU wise but we're not quite up to what say a spark workstation is which given those things started at like $20,000 that's not you know super surprising now M the the likes of the Amiga had specialist graphics iips to kind of help make everything run at a decent speed and the Archimedes range very much does not we just have a straight frame buffer basically and a CPU to do all the work kind of like an Atari without the midi ports but you know a lot faster ooh that's going to generate some comments before we get into the really nerdy detail of how this machine actually works and all the chips that are in it we should probably talk a little bit about the BBC micro itself now for those of us of a certain age based in the UK this computer is basically the ubiquitous machine that we used in all old schools and it ended up being so because of a competition that the BBC ran amongst computer manufacturers to get a computer that would match a particular spec that they were going to use as the basis of a program to educate school children about the use of computers and to go along with that program there was also quite a lot of government funding for schools to be able to purchase computers where the government would cover half and the school would cover the other half it wasn't just acorns machine that had access to this funding in fact schools could buy a variety of machines it's just the BBC micro was the one that the BBC was using in their programming and educational materials so with all that material behind it it kind of became the de facto standard for computing in schools and by looking at it you can see why this is a fairly rugged solid machine that can survive contact with a 5-year-old whereas the machine that they were planning on using for this prior to it being an open competition the new brain well that's not going to survive 30 seconds of contact with a child who's one sure but dib dab away from committing a war crime now it's while dominating this particular education Market the acorn starts work on their new machine well first the arm processor which they made available to software developers as a second CPU Adon for the BBC micro so they could prep stuff ready for their new machine and then next they develop the chips that would support the arm processor and make their actual computer however while acorn's developing all of this the world is not standing still the IBM PC has come out clone manufacturer an hour thing and the PC's basically taken over the business world at least in terms of micros Commodore get the Amiga out in 9 1986 and Atari managed to get their St out as well Apple have finally started to move away from the Apple 2 and have released the Mac a home computer that costs somewhat less than the family car unlike the Lisa which is their first go round at this so by the time we get the first two Archimedes machines and then onto the A3000 the world's somewhat more competitive than when Acorn started this thing out so with the A3000 again acorn's turning to the BBC branding to help push and Market this machine as it's already an established machine Mach the BBC micro so they're hoping to build on top of that particularly in the educational Market with schools that already have a large software liby of BBC applications and add-ons for the BBC so how well does the A3000 do as a successor to the BBC micro a follow on if you were because on the surfice of it you look at the spec and say these two machines have very little in common apart from their manufacturer the BBC micro is an 8bit machine with a 6502 processor and somewhere between 32 and 128k of RAM and a whole bunch of propri ports for adding various add-ons to the bbcn the A3000 on the other hand is a 32-bit microprocessor with 26-bit memory addressing and ships with 1 Megabyte of memory it also has its own proprietary podule expansion system but it doesn't come with any of the ports that the BBC micro came with well that's not entirely true we're going to cycle back to that one the operating system on the two machines is also pretty much completely different Acorn had their text based Moss for the BBC machines and we have a GUI based oper system or at least that's how it appears to most people known as riskos for the Archimedes line starting from the A3000 on on before that it was called ARA o but Acorn didn't want to call it ARA 2 because there was a film called that coming out at the same time and they didn't fancy getting sued although as we'll learn later on the video there's a lot more similarities between acor mosque and risos than you might first assume by just looking at its UI so how does this machine Fair as a BBC micro the answer is surprisingly well unlike every other manufacturer of 8bit machines out there like Commodore and Atari Etc Acorn included an emulator for the earlier system with their A3000 and while this emulator's not perfect it is actually really good you can run most software targeted at the BBC's Hardware on this thing there are some games that stepped around the BBC's operating system that did not behave overly well on the emulator but for educational titles and applications yeah they pretty much all just ran fine in fact the emulator would also emulate some of the rarer hardware for the BBC like a second processor so assuming your BBC was using a 3 and 1/2 in disc you can just take that from your BBC slap it in the A3000 start the emulator and run software off the disc the emulator also adds something known as host FS which let you access the risos machine's filing system inside the BBC emulator if you had stuff written in BBC basic that you wanted to run well you didn't even have to run the emulator cuz BBC basic came built-in CH Archimedes it had been extended somewhat so you could write GUI applications in it but all your original BBC basic code would run just fine in fact the archimedia supports all the screen modes that the BBC did so stuff that assumes that the screen mode works in a particular way yeah no it it just works the emulator is really just there for machine code stuff other very important areas of compatibility that the A3000 had with the BBC is it supported echonet acorn's networking system now at the time this was a pretty big thing for the Archimedes cuz eanet was the most widely deployed land technology in the UK at that point and was in pretty much every school that had more than a handful of bbcs so being able to plug your A3000 straight into that Network and access files off the file server or in fact use your A3000 as a file server to bbcs well that was extremely useful although for a 32-bit machine that can Addle up to 4 mega ram echonet would be relatively slow at least for loading software off not terrible slow but software would noticeably take a little while to load whereas for a BBC micro has a usable 32k of RAM for programs loading stuff of fonet seemed basically instantaneous furthermore there are add-ons available to the A3000 that let it have some of the ports that the BBC micro had like the user port or the 1 MHz bus and those ports were usable in the emulator so if the school had invested into lab equipment for example that made use of that port on the BBC well you add the podule you connect it to the A3000 use the emulator and you're all good when you're not wanting to run all the 8bit software well you have all the nice 32-bit software you can use now although there's not a lot of it around the release of the A3000 there is certainly some and its software Library will get a lot better but compared to the PC for example there is a relatively low count of software titles although if you're a scul awful lot of them are geared up to you also for schools that heavily invested in teaching programs for BBC basic well that updated improve version the basic is just there ready for you to use built into the machine so schools can still make use of the same teaching and lesson plans they had previously only now they can introduce Concepts like the GUI into it now all of this is a good example of acorn Levering its position in the UK education sector that means there is a machine that's a good fit for what's already being done however it is pretty proprietary and acorn doesn't have much of a presence outside of the sko market so for some skulls why it leted them keep using the software things that they already had it did mean that they couldn't teach new things that were more PC based that schools were starting to fill would prepare children more for the wider world of business so we do see the previous main presence in the UK education Market RM stage a bit of a comeback with their IBM compatible is line of PCS the RM Nimbus and in scores that is the main competition for acorn's Archimedes space machines the Home Market has diverged somewhat or at least in the UK as species are far too expensive for home use and while the Spectrum may be effectively dead now Commodore has come in with its Amiga computer which in 1989 the same launch year as the A3000 has the most successful sales period for their cost reduced machine the A500 and while Atari has slipped from being the most popular of the 16-bit Platforms in Europe to number two behind the Amiga it's still doing very well and you will find it in the music Departments of most UK schools though in inside scores the A3000 has some proper competition outside of scores it also has some proper competition now price is actually a pretty big deal here so by 1989 the price of omiga 500 had dropped to £399 the A3000 on the other hand at least at its discounted price to schools was selling for £529 at retail that was £ 649 so even if you're a kid that really really wanted the A3000 you were getting an Amiga 500 for Christmas I mean I was that child and admitt I'm not that disappointed that I got an Amiga let's put it that way I should probably also mention that £ 649 that's xv8 as well so let's get this machine open and look at the nerdy detail inside now there have been plenty of videos around that talk about the arm and arm architecture and how great that is and how it came to be and completely ignore all the other chips that Acorn had to develop to make this thing go so not only are we going to look at the arm process so we're also going to look at vidc MC and ioc but we'll start with the CPU which is an arm 2 clocked at 8 MHz now this doesn't contain a lot of differences from the original arm one that Acorn chip doesn't add on to the BBC micro for developing software for the arm and it's a pretty simple design because they wanted to keep the thing with sufficiently low power drawer that they could package the thing in a plastic package rather than a ceramic one because that would substantially reduce the cost of the individual chip that would also Faithfully lead to Arms great advantages that would allow it to become the thing that it is now now there is no onboard cash or cash management inside this CPU which was a reasonably wise decision cuz when you're designing your first chip doing a CPU cache is well pretty challenging even now cash coherency in the processor is not the easiest thing in the world to get right and test for and at this point in time memory is still available in similar speeds clockwise to a CPU so an 8 MHz CPU U could be paired with 8 MHz memory so you could still get excellent performance out of your CPU with no gash the other thing that they didn't particularly want to do with the arm processor and all its subsystems was dma direct memory access so they made the CPU particularly good at handling an interrupt as they knew the CPU would be doing all the work of dealing with stuff from the io devices this would also be where this 26-bit memory addressing thing would come from because 26 bits is a really odd size for the pointer to memory so you knew there had to be a good reason for it right now you might be wondering why not just do dma other systems like the PC did dma and that comes down to this trying to keep it at a reasonable price thing cuz dma controllers were expensive I mean a lot of computers the dma controller cost the same as the CPU so if you can have decent performance without doing the ma and shove all that functionality into the CPU well that's going to significantly keep your costs down another reason for not going for the cost of doing dma is when your CPU has no cash if you remember I mentioned earlier that arm 2 there's no onboard cach on that CPU so if you have to lock the CPU out of memory so Hardware can do dma it's not like your CPU is going to be doing very much because it doesn't have a cash so it'll just become stalled waiting on getting access to memory again at which point you may as well just asso the whole dma thing and get the processor to do iio instead so that's what they did and knowing that that was the plan when they designed the processor well they could design the processor to do IO in software as fast as it possibly could so in order to aim with that the arm processor had a fairly neat trick referred to as the fast interrupt request now when handling one of these fast interrupt requests known as an fiq for short registers 8 to 14 of the CPU were banked and when we're handling on these requests we switch which bank is in use so this means that if you write your inter handling code to only use registers 8 to 14 then your code can run without having to save anything in or out of those registers and if you're not having to call either the load or store instructions very often that means when running this code your CPU doesn't have to go in or out of RAM and wait for Ram and so our code can run very quickly but this is also where the bit with the 26 bit addressing comes in because register 15 is our program counter that starts the pointer to where we are executing stuff from RAM and in any inup Handler you're going to have to save that value somewhere so you can write it back in and jump the code back to where it was supposed to be once you finish doing your interrupt handling but those aren't the only bits of processor State you need to squirrel away and put back again there are all sorts of flags that need to get stored as well so here acor came up with a fairly neat trick what they do is they cram the program counter and all the flags into one register so by the time we've used up all the bits to St the flags and CPU mode interrupt mask Etc we're left with 24 bits left for addressing but I said before it was a 26-bit addressing system didn't I well that program counter is 32-bit word aligned so that effectively gives us a 26-bit addressing system now with the PC and the entire CPU stay all in one register that means you have one register to save at the cost of limiting yourself to 64 Megs of address space which in 1986 was an awful lot of address space so no one felt like this was a major restriction on the platform and if we'd ended up having a full 32-bit program counter and another register for the status Flags well every time you handle an interrupt you got twice the work to do two registers to write out and pull back instead of one so this is an incredibly sensible tradeoff for this point in time I mean eventually arm would move to using a 32-bit program counter because 64 Megs of addressable memory becomes somewhat of an issue later on but at this point in time 64 me of memory is way more memory than anyone is going to stick in one of these machines mean remember acorn's got this thing initially strapped up to a 32k BBC micro while they're developing this stuff so at 64 Mega dress space seems pretty flipping huge and from a risk point of view it's also really nice cuz you don't have to have a separate instruction to return from interrupt you can do a subtract into PC to jump back to the right point now we've talked about how fiq works with the CPU let's have a look at ioc the chip that's designed to handle the io now we've already established that it does not do dma so what does it do well for a start it handles all the cycle timings and glue Logic for all the various IO inputs that come into the system those 16bit Pils and that 8bit ecinet slot it also has a little serial uart in there and microcontroller that handles all the keyboard and mouse input so our processor does not have to handle all the keyboard scanning stuff with row and column our little microcontroller in ioc just gets on with looking after the keyboard we also have a series of 4 2 MHz timer SL counters that various software applications can make use of it also handles all our gpio and i2c stuff as well it also stores all our interrupt masks Etc and the interrupts from all our devices end up getting Marshal there before it flags up and interrupt to the processor the next chip we're going to talk about is meming now there's a lot going on in meming but it's basically responsible for controlling and managing all the memory in the system and getting the CPU in and out of it it also has a pretty special function related to the video chip as well which we'll come to later now one thing that might surprise you about a chip that's designed to manage what's happening with memory is that it only lives on the address bus and not on the data bus now you might think that's a pretty odd choice that acor made there but they did have their reasons if you need to put your chip in a package that had enough pins to be able to do both the data and the address bus well that was a lot more expensive package so if Aon can get away with putting this thing on one bus well that make this individual chip a lot cheaper to make they also repeat this but it's just on one bus trick with a number of the other chips as well always conscious about how packaging can really push up the price of a chip so this chip provides some features that you might at first thought not occur to you it's going to be here so clock generation is one of the things that this chip does and that allows it to control when the CPU is getting in and out of memory by essentially stalling the processor for a little while with the clock signal now you might wonder in a system with no dma why on Earth you'd want to stall the CPU from talking to memory and that's because I sort of lied to you the vast majority of our Hardware has no dma access the video chip however vidy sort of does dma as I mentioned before this chip has some special features to do stuff with the video and I will come back to this later on but this is around this whole dma sort of idea so that's why it's useful that clock generation is done inside memi also we have you have to stick it somewhere right unsurprisingly this thing does the refresh of the dams so stuff keeps being in memory and it also does the translation from the address to the row and column system that dram chips use for addressing now one of the interesting bits by the time we've got round to doing this we've hit the point where accessing every bite in RAM does not happen at the same speed if you're doing a read or write to memory waking up the first cell in the block takes longer than all the subsequent linear reads or wres that happen after that and this is why the arm chip breaks the Risk rules a little bit and has store and read multiple where it all read or write multiple registers at once to memory rather than just calling the one load or store instruction multiple times it does have a single register read or write instruction but the more we make use of the read or WR multiple the higher utilization of the memory bandwidth we make this is actually a big part of why the arm processor is quite so fast as other competing CPU architectures at the time didn't really make use of memory bandwidth in quite the same way that the arm did now the other thing M's got and it takes up nearly half the die in this chip that's more or less the same size as the arm CPU itself is it's the mmu now even around this time you're getting CPUs that have the mmu built in if they've got one but again this goes to cost if you tried to get the mmu inside the arm chip well it would have needed about 50% extra die space and as each time you increase the size of the die the yield falls off the arm chip already had a die that was the size that was the most economically viable at the time so it was definitely going to be cheaper to not have the mmu on the armch itself but have it in meming now m is also where one of the big restrictions in the whole Archimedes platform comes from the whole 4 megabyte physical RAM limit or at least 4 megab per MC chip you could have more than one MC chip in the machine but the likes of the A3000 being a more budget conscious machine at least for acon only had one MC chip on the board that's why the A3000 maxes out at 4 megabytes of memory the bigger pizza box and desktop machines some of those would have a MC chip on the ram upgrade card they made use of and for each one of those MC chips they could have 4 Mega physical RAM all the way up to the limit of 16 mbes as physical memory can only use up half the address space cuz virtual memory takes the other half and a bunch of that address spaces taken up with ROMs and also the memory address space for Io devices leaving us with 16 anyway back to the mmu now this particular mmu Works backwards from the way that most mm used do or at least I'm pretty sure that it does there's a lot of documentation that says it goes from physical address to flanking which page in virtual memory that physical address belongs to rather than us trying to map a virtual address to a physical address it works backwards effectively at least that's what all the documentation seems to say unless I've horrifically misunderstood it now I honestly can't work out why this was their design decision I assume it was cost based it's just I haven't figured out what makes this design massively cheaper than doing it the usual Ray round but this design is also where we get our 4 megabyte limit from because the chip is essentially cutting the physical memory up into virtual Pages now the mmu provides 128 physical pages and that is limited by essentially the size on die you have for this page table now it doesn't matter how much memory it's controlling there's always 128 pages so the size of a page varies with the amount of memory not the number of pages that also means there can only be one virtual mapping to each physical page of memory and with memy only being on the address bus means it only does address translation it doesn't do any IO remapping now with a fixed number of pages and variable page size it turns out there really is a maximum size you want your page to be for it to still be useful and that's round about 32k so 128 32k Pages gives us this 4 Mega physical memory limit per memy obviously if we added a second memy well then we have 256 pages of memory and 256 pages of memory at 32k maximum size well that gives us 8 Mega memory and so on and so forth with each MC that we add to the system and four megas of ram now you might be wondering to yourself with it only being on the address BS how on Earth do we set up the page entry in the mmu because we can't write data to it well here the design of memp's Fairly smart how it works is there are number of addresses that are reserved for controlling memi or a memi and we get the CPU to set that particular address and then it writes any value out it doesn't matter what the value is only that it writes the MC lives on the address bus so it gets to see the address that's been selected and it gets to see that the right line has been hit with memi not being on the data bus it doesn't actually get to see the value that the CPU writes out but that doesn't matter it's just the coding this address data and it uses that address data to work out what entry you intended to put in its little page table now having a block of address bace that can only be written to would be a little bit of a waste but if we look at the Archimedes memory map we can see right at the top of that address space in the section that's only accessible when the Chip's in supervisor mode you can see we have the ROMs for the operating system risos and the control for bsei has an 8 mgab area that overlaps with the 12 mgab of ROM space if you read from something in that 8 mgab range you get what's stored in the ROM if you write to it then you write to the control registers of meming this is a pretty smart way of not wasting a bunch of address space because ROMs can only be read there's nothing you can write to them cuz they're ROMs and the MC will only take values that are written to it there's nothing to read so this system of overlapping saves us some valuable address space after all we only have 64 mgab of it you can also see in this memory map that we only have 16 MB allocated for physical RAM and that's because we've got to have a mirror of the physical address space in the virtual address space So Physical gets 32 megabytes of address space and virtual gets 32 megab of address space we're then using our top 12 megabytes for our ROM space we're then using the next 4 megab for our IO space after all your Hardware has to map somewhere with its buffers little bits of memory and the ROMs that are on there and that 4 Meg range is where it happens and after that well that leaves us 16 megab of address space hence the 16 megabyte limit in this particular architecture with later Archimedes machines like the risk PC for example assuming you can that as part of the Archimedes range well then you've got 32 bits addressing here we can have a much bigger address space and memory map but thus we can have a lot more physical RAM in a machine this 60 Meg limit might feel like it's quite a big problem but it's really not in 1989 as then 16 MB of memory is still a huge amount of RAM for example the entire Mega 52000 Etc that can only physically address 16 megab of RAM and barely any of those machines ever held 16 megab of ram some machines for example like the original Mac why it may have had a 16 mbte addressing limit in terms of what the processor could do you still couldn't s 16 MB of ram in it in fact 16 mbes of ram would have been about four times as much as a pretty wellspect PC had at the time in fact plenty of PCS at the time didn't have the physical slots to install more memory than four mag or in sometimes in the nicer ones eight mag so there's one major chip left vinzy which unsurprisingly is the chip that handles video and also audio most people seem to forget to mention the audio part for the time period vids is a surprisingly capable Graphics chip you can have color depths of 1 bit 2 bit 4bit or impressively 8 bits per pixel now that gives us a maximum of 256 colors on the screen at once and if we compare this to other machines around the same price point well more or less like the Omega for example which is a machine very much thought of as graphically capable without doing clever tricks that's essentially limited to doing 64 colors on the screen at once and even that's an extra half bright mode you have 32 colors and then the other 32 colors are just half bright versions of those original 32 colors like the Amiga though vidy also pulls from a 12bit color palette which means of the however many colors you're currently displaying or want that draws from a pette of 4,096 colors however unlike the amiga's display this is a essentially a process driven frame buffer almost will come on to how that actually works later on but you don't have the blitter you don't really have Hardware Sprites either well you do you have one which is used for the mouse pointer there's also no fancy planer modes or anything like that so this is all done as chunky pixels where essentially you write the color you want to be in a particular pixel into a location in memory but this is a lot better for doing 3D stuff than the plan mode that's all about essentially fast moving Sprites which we don't have any well we have one it does also have the concept of transparency built in so this makes gen locking the output of the arc pretty easy actually there are also some cheaty display tricks like there are in the Amiga so although you can't quite do hold and modifi the way it did to get 4,096 colors on the screen at once you kind of can with those programmable 20 MHz timers we talked about in ioc you also had a reasonable range of screen resolutions as well you can get vid2 output so you could do 1024x 1024 in monochrome all the way down to 640x 265 in 256 colors now I should also cycle back to the audio cuz I did say that a lot of videos ignore the audio when it comes to vidy and this very much is the sound chip as well and it supports eight stereo channels in 8 bit but that's a logarithmic 8bit so this actually sounds better than you would think in fact if you should compare it to non-logarithmic systems it sounds more like it's a 12-bit audio stream if you're familiar with PCM audio it's basically like that now interestingly putting the audio and video in the same chip at least for vidc the original version did actually cause a little bit of interference there was noise on the video output signal created by the audio signal at least in vid C1 in vid c1a the they did actually fix this in quite an interesting way they took the audio signal inverted it and fed it back into the video signal so the inverted audio signal would cancel out the noise from the audio signal thus giving you a clean video signal again I must admit I love the beauty of this as a solution it's basically noise cancelling headphones via video signal now vidy kind of like memi only sits on one bus it only sits on the data bus not the address bus and again this keeps the pin count down on the package that vidc is packaged in thus keeping its cost down now you might be wondering so how does vidc select which bit of memory to read so it can output its video signal and the answer is it doesn't memi does this is where we come back to that whole sort of dmma thing I talked about with memi and vidy so vidy itself just has a fifo in it and memy essentially streams data along the data bus without actually being on it into that fifo and VY basically takes what ends up in his fifo squirts it out through its digital to analog conversion process and there's your video signal okay the practicalities of that are a tinely bit more complicated but that gives you a good overview of the actual architecture that's underlying this so MC holds a number of registers like where in memory does the video frame buffer start where are we currently in the video frame buffer and how long is the frame buffer now remember M's only on the address bus so it can hold the addresses for us and vidy is only on the data bus so there's no point in it holding addresses cuz it doesn't have access to the address bus so there are a couple of lines that run between vidy and MSI to do a bit of a handshake so when vid's ESS she knows it's fifo starting to get towards empty it exchanges a message with vidy to say send me some more data MSY because it does the clock generation for the CPU can now stall out the CPU to stop it from Reading from the data bus and because it has a register for where the frame buffer starts how long the frame buffer is and where the current position in the frame buffer is can calculate out the address in memory that vidy needs to start receiving data from and it passes those addresses into the dam which outputs that content onto the data bus at the same time it sends an acknowledge signal to vidy to tell it to start reading and shifting whatever's in the datab bus into its fifo once it's moved out the little chunk memi can stop inhibiting the clock to the CPU and start sending it its normal clock cycle and the CPU can just get on with its life until the next time vidy happens to need more data at which point we go this little cycle again sound basically works the same way as well the only difference being there's a double buffering scheme going on here so when half the buffer has been got through it generates that little interrupt that gets it to shove and fill that sound buffer back up again now this is a really neat system for getting most the advantage of dma for your audio and video without doing all the sorts of things that other systems that do dma have to do and for a machine designed to be essentially a personal workstation SL desktop computer this is a perfectly decent Arrangement we're not being being a server we don't care how fast we Shuffle things from the hard disc to the network card for example or at least in terms of compared to keeping video alive so programmed IO through the CPU for anything that isn't graphics and sound is still pretty performant just without pushing up the cost of the Machine by having to acquire a lot of complicated dma components while still getting us a lot of the advantage of regular dma for our video and audio or whilst keeping the pin counts down on all the different packages to make them all much more cost effective to man manufacturer so how does our A3000 stack up to the competition well it kind of depends on which Market you're talking about in the educational school space it has the advantage that is a very good follow on to the BBC micro compatible with a lot of BBC micro based stuff that schools already have however in the UK at least the curriculum is changing somewhat it's becoming much less based around what you think of as computer science you know doing software development Etc and learning how computers work and much more about let's learn to use a spreadsheet yes it got very business it focused and they wanted to train kids to be able to use a spreadsheet and a word processing program and ideally the ones that businesses were using you know on PCS so that got RM and their Nimbus line solidly into schools now previously RM had been a big supplier of machines for educational use until the BBC micro became a thing at which point their share of the market kind of plummeted somewhat but they have a line of sort of IBM compatible is machines I mean they're not really true PC but they can supply you a modified version of Doss that will work on it and a modified version of Windows that will work on it and some wrappers that will allow other applications that run under Doss to work I mean once you had the modified Windows applications those tended to work just straight out of the box they also provided some compatibility layers to try and get some stuff that wouldn't just work in DOS working and while the RM nimus graphically is a little bit better than most PCS of the time it's still nowhere near as capable as the Archimedes line is particularly the A3000 processor wise they shipped it with a 186 so again the A3000 kind of Tres that fairly easily I mean by really massive margin actually and when it comes to sound the Aron numers doesn't have any networking wise the Archimedes can go onto an econet Network that your score probably already has with all its bbcs whereas the arous it does come with zet which is its own fairly cheap system for networking the nimbuses together and actually data rate wise they're very similar to each other they even have a very similar addressing scheme of an 8bit station number although I'm not sure if zet also had a network number as well so I don't think you could link multiple zet networks together I may be slightly wrong about that but all in all if you compare the A3000 versus the Nimbus 186 the 3000 is a better machine it's just not a PC and with the changing curriculum to let's do word processing in Excel it was starting to need to be a PC now you can probably tell that I don't think that that was a particularly good J Chang in the curriculum and well it turns out these days most schools and the people who set the curriculum agree with me and it shifted a lot more back to a more computer sciency kind of a curriculum the assumption is kids can just pick up learning to use a word processor in an Excel spreadsheet as they need to in other subject areas and you know what it turned out they could in the Home Market the leading system it was stood against was that of the Omega the omiga 500 in particular sorry Atari uses I I know the St was out there but there's a limit to the number of machines I can pair this thing to at once so we'll pick the leading system in the UK Market which is where the A3000 was mostly sold so the Omega 500 was a lot cheaper than the A3000 there's a couple of hundred in it but at least in terms of processor it is significantly more powerful compared to a 68k clocked at say 7 mahz which is what the A500 has it's about 7even times faster but the igga does have a lot of custom graphics chips so for example it's able to to do a lot of Hardware Sprites you can move around bits of the screen purely in the graphics Hardware without the CPU really having to get involved apart from laying some instructions down for the graphics chip you also have the Blitzer which again writes a lot of screen memory all at once and can do all sorts of interesting little effects and we've got Hardware scrolling as well for the screen so you might think that that adds up to games on the Amiga running a lot better I mean for those games that reported to both the Amiga and the Atari where the Atari didn't have the same level of graphics Hardware the Amiga version would typically have more things on screen scroll more smoothly particularly the backgrounds and that's down to all that custom graphics Hardware that's in the Amiga so when we get games ported to the arc where we have an atarian amga version we probably expect them to behave a bit more like the Atari but no they tend to look and move a lot more like the Amiga version because that processor is seven times faster that gives us more than enough CPU resource to move the same amount of stuff around on screen in the same sort of way or at least visually as the Amiga does we also tend to sound a lot more like the Amiga version because it t tended to use sort of sound based on Sample playback even for the music and again our A3000 it can do exactly the same thing it can actually do it in something that sounds ever so slightly better and has more channels but most game software houses are not re-recording and doing all their audio again for the Archimedes where the art gets to really shine over both the Amiga and in fact the St is with 3D graph not having to do everything in pler modes makes writing stuff into memory from the kind of raycasting calculations you use a lot faster the Amiga also doesn't have much in the way of special Graphics Hardware to really help with the whole 3D thing so it has to just fall back to what the 68k can do and the 68k can do about seven times less stuff than the Archimedes processor can so hence you get some 3D titles that start life on the Archimedes and the A3000 that look well really good so you have David braen Zar which starts off as the Lander demo and eventually we do end up with a version of that on the Amiga called virus it is one of the standout titles though for the Archimedes platform as well as its version of elite AR Elite which for many years was basically the best version of elite out there also a pretty much standout game for the platform was conqueror a 3D tank battle game created by Jonathan Griffith the guy who did Snapper for the BBC and acorn electron he apparently used part of the landscape drawing engine that David braen had created for Zar or virus used that to do the terrain and created essentially a really good tank battle game this is what we tended to run on our schools Archimedes machines when the CDT teachers weren't looking and it was one of my favorite games on the whole Archimedes platform now that's not to say there weren't a ton of really good games ported to The Arc in fact most of the ones you think of is wow that's a really good game on the Amiga or the St yeah there's an Archimedes version of it it may have come out like a year or two later than the version on the Amiga or the St but it did event get them so there was no shortage of games for the platform and sometimes the Archimedes version of the game would actually be better than the Amiga or the St version but not very often a lot of it was kind of shovel Weare from other platforms onto the ark which is a shame cuz with all that extra processing power and 2 Mega of ram as standard you could have really made some significant improvements to a number of these games at least graphically and soundwise and also it could just display more colors at screen at once than the A600 and 500 it wouldn't be until the Amega 1200 that the Amiga finally got a 256 color mode so how well did the last ever model to Bear the name BBC on itself while being squeezed by cheaper machines at home and the PC coming into scores it's still sold reasonably well well enough for acor to make a number of SQL machines in the wedge form factor that the A3000 is in we see that acor releases the a3020 which is the top end more school focused version and the A310 which is aimed much more at the home user and there's a bit more budget now these machines were somewhat more performant than the A3000 about 50% faster and that's a couple of things working together firstly there's faster Ram in there they've also managed to clock the processor up to 12 MHz but also this contains an arm 250 which is the armf free instruction set but with no armboard cache but all the other chips that make up the Archimedes ioc vidc MC Etc all in the same package this is effectively a system on chip now again this is very cost effective but also all that tight on die coupling means that things can clock up faster at the higher end we see aor releasing the a5000 to replace some of their more expensive machines and then we'll also see a laptop version of this in the form of the A4 after that we would then see the risk PC line of machines and the a7000 now you will hear people referring to the risk PC and the a7000 as aredes machines and formerly actually they're not heor referred today RIS PC line as a replacement for the Archimedes line and they are actually architectured a bit differently there is still an arm processor at the core of these things there's a chip known as vid C20 that's the display controller effectively but you won't find a separate MC chip or ioc chip because those functionality have sort of been taken into the arm CPU now and they they just don't work the same way you don't have the same memory restrictions in fact this risk PC you can see here it's got 256 Mega memory on board which was a lot for the time in fact I doubt many people at the time had that much memory it's just you can stick this much memory in and that's quite cheap to do now so I I I did we sadly also no longer have something on board to let us put an echonet interface in so I've had to have a podule that does that and also I have an IDE podu as well so I can set these little disc modules on and use them instead of spinning glass cuz who need spinning glass hard drives in their life anymore for their retro machines particularly 30-year-old spinning glass hard drives this thing also has a strong arm processor in it which is a processor not designed by arm but by deck the digital Equipment Corporation who licensed the instruction set off arm but did all the chip design themselves and at the time gave us the most powerful arm chip available unfortunately can't really be used to its full potential in the risk PC cuz there's not quite enough bandwidth in and out of ram but it was still a hair chunk faster than the process that the machine originally shipped with the risk PC line runs risk OS just like the original Archimedes machines but we can't necessarily run all the software we ran before particularly if you happen to have a strong arm as it doesn't support the then Legacy 26-bit addressing however some clever people did create a software layer that allows it to do or at least sort out the 26 bit addressing issue so older software particularly games can run sometimes a bit too fast though as well as the special podil slots we do actually have a special Network slot for an EET card to go in Yes we finally moved on to the world of eonnet because an EET card now doesn't cost more than the computer it's going in but other networking systems like sj's Nexus which is cheaper to purchase also could make use of that slot but as a sign of the times you'll see in its second processor slot I essentially have a PC on a card well or at least a PC processor on a card and with that I could boot up an entire PC in here and run dos or Windows 95 on it this did mean that risk PC owners could run Doss or even Windows 95 and some even ran Linux on this thing although Linux would get a direct Port that you could just run on the risk PC itself and while this did broaden up the software that a risk PC owner could use I do fear that it maybe hasten the migration off the risk PC platform for a number of people and this is where acorn's line of computers actually genuinely comes to an end after this there are no more computers from acor they did start prototyping out the risk PC2 of which many of us have seen the case for because those all got sold off there is apparently still at least one prototype kicking around at the center for computer history but sadly the risk PC would be well the end of the line for all things Acorn and Computing it's process of the arm though that would go on to have well Commercial Success beyond the wildest dreams that acor could have ever imagined in fact based on the metadata I get from viewers more of you watch these videos on an arm-based device than anything else because they power your phones your TVs your tablets Chromebooks even Macs now yes someone is making an arm-based computer these days and it's Apple well if you got to this point I'd like to say thank you very much for watching I would also like to thank Dudley of yeser for the Canon father footage a phrase that turned out to be slightly more of a tongue twister than I was expecting Canon father you don't know how many takes I did of that but I'm also going to mention him in a new section further viewing because in an entirely unplanned move despite the fact that I actually did a little bit of voiceover work for it this month's yerene is acorn and is about the right period that we've been discussing here so if you want to see some of the magazine coverage of these archimedian machines from the period Then go have a look at that you can also see some of the stuff that would have come on its cover floppy disc too if you've not already followed the link in the description to go watch yeser scene and you enjoyed this video YouTube have provided a button you can use to indicate that fact if you didn't that there is also another button then we'll just say thank you for the engagement if you really like the video well then there's a subscribe button that will subscribe you to the channel that really helps the the algorithm tell people that these videos exist [Music]
2024-12-21