How the Amiga Worked

[@miasma82]

My first computer was an amiga 600. I was born in 1982 but i can't remember which year i got it, but i know that i used it for a long time. I didn't get many things as a kid but i'm so grateful to my parents that they bought it for me becaue i learned so much about computers by using the amiga.

[@gravious]

A500 was a cost reduced A1000 and was essentially the same machine in a different formfactor, so the amiga being out in 1985 means it was released significantly before the megadrive

[@ronny332]

Cost reduced is always a word that makes me smile in combination with Commodore. They reduced everything to the last screw, but the technical features stayed the same and overall quality was always very good. Reducing the keyboard quality for instance was an option but never done. I don't know any keyboard of the Amigas that was really bad. When you get an SE or Value model today, you can be sure that something is missing.

[@jameswelch2850]

i absolutely love this series. seen them all. you're great at explaining these.

[@sulrich70]

It was originally intended as a games console, and the thing grew into a fully fledged computer, with a multi-tasking OS. A seriously amazing piece of kit for its time, delayed due to initial funding constraints, a deal with Atari and then Commodore took over. It was a passion project for the designers, and amazing it lasted as long as it did.

[@joehoy9242]

As I recall, strictly speaking, the fledgling Amiga Inc. pitched the technology to investors as a next-generation games console (because in 1981/82, that's where the money was, and with Jay Miner being an ex-Atari engineer, that would appear to be a relatively safe bet). According to the Los Gatos team however, as far as they were concerned, once they had secured funding, they'd always always intended to produce a full-fledged computer.

[@johnrickard8512]

The only thing I noticed that was wrong is that "Agnus" is misspelled consistently as "Angus". Otherwise, all is well!

[@sunnohh]

Must have had lunch on his mind!

[@bachaplegic]

Wasn't it called Fat Agnus?

[@epicon6]

Oh daaaaaaymn. This changes EVERYTHING!!

[@johnrickard8512]

@@bachaplegic that chipset was used in the 2nd generation Amigas

[@joehoy9242]

​@@johnrickard8512 - Yup, IIRC "Agnus" was the OG DIP chip used in the A1000; "Fat Agnus" was the PLCC revision used in the A500 and 2000, and "Fatter Agnus" was the later revison which could address 1MB of Chip Ram.

[@gurujoe75]

Agnus. This is probably the most important chip, thanks to which the Amiga is a 2D gaming monster house. Agnus is blitter, copper and DMA management.

[@ronny332]

60 fps on Amiga games was pretty rare. Maybe in the first years, but very quickly the rates were turned down heavily. So, the described time within a frame (1/60th of a second) was in practice for many games a lot longer. I remember games like Testdrive that were more or less a slideshow, but I didn't recognize it as a kid.

[@epicon6]

When i was a kid i was lucky i wasn’t aware of all PAL games running 17% slower than NTSC versions. I would have lost my mind. PAL Tekken 3 was the first game i noriced runs relly really slow

[@SerBallister]

@@epicon6 Depends on the developer, you can adjust speed for PAL or not base timings off of video. You get more CPU cycles per frame in PAL.

[@joehoy9242]

I'd argue it was the other way around - obviously we're talking 60Hz on NTSC machines and 50Hz on PAL (though the "Fatter Agnus" on the Rev 6a A500s allowed for software switching between 50 and 60Hz on PAL machines). Games written specifically for the Amiga hardware would often attempt to achieve a 50/60Hz framerate - Shadow Of The Beast was a famous example. For better or worse, a lot of the earlier games were essentially ST ports which did not take advantage of the hardware. That said, to name another example, Gremlin's "Lotus" series had a 25/30Hz framerate, and on a CRT they were still buttery smooth.

[@TheRealWinsletFan]

5:15 - Original chipset such as in A500 could access only 512K of chip ram. Additional ram would be "fast" ram. ( which may actually have been "slow" ram ). Only ECS chipset ( A500 Plus, A600.. ) could access 2MB of Chip ram.

[@off1k]

A500 had slow ram through the trap door and fast ram with the zorro slot side expansion, later revision OCS A500s (rev 6a) could have 1MB chip ram because of the 8372a fatter agnus chip. You could also do a mod on the board and cut traces to covert 512kb slow ram into chip ram giving you 1mb chip.

[@Stevie-Steele]

I got the Amiga 500 Plus "Cartoon Classics" bundle as a 7/8 yr old kid - I enjoyed it a lot but it was a bit harder to use that the Sega Mega Drive that my friends had. In retrospect - I didn't appreciate how cool this system really was. Felt like a PC/Console hybrid.

[@davidretrogamesplayedbadly3533]

Such an iconic machine, enjoyed this mate.

[@doctorsocrates4413]

I have 4 amiga500s .i use one daily...love the machine ...thank you for the video.

[@Saudade720]

I still feel excited when I hear that Turrican theme

[@JohnnyWednesday]

Will have to listen to the Turrican music before I start - bare with me

[@mcjdubpower]

My heart sank when it stopped, really like fire and ice theme too ❤🎉😅

[@rwestvang]

Oh the days when you just waited for a vb for a smooth scroll. True multitasking years ahead of the PC. So granular control over everything even with disk access etc with the possibility to create your own trackloader as well. Everything could be made custom if you wanted. If not, you had lot's of libraries to draw from as well.

[@themeangene]

The sprite hit detection honestly is better than anything else I've seen on 2D architecture with the exception of the Sega Saturn but it's not fair to compare mid 80s hardware to a 2D mid 90s powerhouse

[@jalipeno97]

im actually glad you arent an audio guy, since im an audio gal! you are correct in the assumption i like learning about how graphics work more to answer that part, paula has 4 signed 8bit pcm sampling channels which use dma as with most of the ocs. two channels are panned left and two are panned right, and afaik those are set in hardware. there was a low pass filter applied to paulas output to prevent aliasing, enabled by default. it can be disabled however there is a feature where you can use one channel to modulate another one, which as you mentioned CAN be used to achieve rudimentary fm synthesis and other effects, but it was rarely used

[@off1k]

There are two low pass filters on OCS and ECS Amiga's, One that is always on and one that could be controlled (on or off). AGA Amiga's had the always on LPF removed resulting in a slightly different sound. AM modulation and Ring modulation but like you say, very very rarely used, PCM was the new kid on the block :)

[@johnellis3383]

Don't really know much about the Amiga, this is gonna be good!

[@ArneChristianRosenfeldt]

Amiga has Tiles? That’s news to me. Just like in the Atari Jaguar the blitter is used to copy the tiles into the playfield as they enter the screen while scrolling. I wonder how playfields wrap around at memory boundaries. The register in Agnus have so ancient names. 9:17

[@sulrich70]

Correct, no tiles. Blitter was used to build playfields with tiles.

[@SerBallister]

Playfields didn't properly wrap around at boundaries like (for example) a Nintendo, they were scanned out from memory in linear fashion, which would create something that looked like it was wrapped, but was offset by 1 scanline (as at the end of the first scanline, the next scanline starts in RAM)

[@ArneChristianRosenfeldt]

@ I meant “vertical” wrap around in order to not crash into wavetable or sprites. The other wrap around would be important for texture mapping in the AtariJaguar, which was designed by Amiga FanBoys.

[@DaveByrdUK]

you're going to get some shit for this buddy!

[@dan_perry]

Yep, errors all over the place.

[@fredsmith1970]

I had Amigas back in the 80s and 90s, and whilst this is a nostalgia trip, there are a lot of errors in this video.

[@GenerationPixel]

All you've achieved for me is deliver the need to dig out my A500 mini and spend an afternoon with Cannon Fodder 😂

[@random007nadir]

Angus? Oh dear.

[@rednight2476]

In the US PC clones where both cheaper and more readily available then an Amiga. I think Commodores refusal to make the A500 cheap and available is the number one reason for their demise. Yes they literally did refuse to sell the A500 in American chain stores.

[@fredsmith1970]

The A600 was meant to be a cost reduced A500 (it was to be called the A300), and ultimately the entry level Amiga to replace the C64 in that position. But apparently it cost more to manufacture, so they sold it as a replacement to the A500.

[@TheRealWinsletFan]

Paula did not do FM, not natively anyway, you could have done it in software I guess.

[@Nebulous6]

Actually, it does do FM and AM in hardware by modulating one channel with another. It's in the Amiga Hardware Reference Manual. Of course, if you want finer control over it then, yet, you'd use software like Sonix. You can place any shape of single-cycle waveform into each of the channels and it'll auto-loop them as a state-machine (i.e. no CPU required to keep the loops going). "The audio channels may be linked to provide amplitude or frequency modulation or both forms of modulation simultaneously." p. 4.Details on page 149. You can even combine channels to raise it to 14-bit audio (2 channels instead of 4). And, finally, wavetable synthesis can be achieved using software like Octamed (again, easier on the CPU than a lot of other computer sound architectures due to audio DMA and the state-machine implementation. Joe Decuir covers some of this in his talks as well.

[@juu226]

Yeah, first Amiga 1000 was revelation when it came out, nothing could compare to that, but it was so expensive that I got mine A500 years later and it was used one. If those 2000, 3000, 4000 series would be cheaper, today PC market might be looking different than it is now. Mid and second half on 90s was time where hardware development was so fast that it could not compete with slow improvements like late AGA, absolete on start. Mac barely survived, PC started by IBM was so open that it had best chance to compete, even if IBM itself collapsed.

[@joehoy9242]

IBM never intended the PC architecture to be open - the only reason the platform became open to cloning was because when Compaq reverse-engineered the BIOS to create the first compatible machines, IBM tried to sue them and - infamously - lost the court case. The A1000's price point was not actually that expensive - it was cheaper than IBM's PCs and the OG Macintosh, but more expensive than the market sector encompassing the Atari ST and the 8-bit home machines - where CBM had traditionally performed the strongest. As I mentioned in another comment, CEO Tom Rattigan drove through an important strategic course correction for 1987, splitting the platform into the A500 aimed at the home market and the A/B2000 aimed at the professional market. With 20/20 hindsight, that's really what they should have done in the first place - even if it delayed the initial release of the platform. While this strategy was put into effect somewhat belatedly, even in 1987 the potential should have been there for the Amiga platform to make serious inroads in the US as well as Europe and Australasia - but Irv Gould couldn't see past the end of his nose and canned Rattigan as CEO before he could follow through on his business strategy. He ended up hiring notorious (and somewhat clueless) "yes-man" Medhi Ali to replace Rattigan as CEO, and CBM International swiftly became a bit of a basket case.

[@juu226]

@@joehoy9242 True, same was with intel-amd, my point was that open architecture helped to grow and survive in dynamic moore law enviroment.

[@joehoy9242]

@@juu226 - Possibly, though I suspect it's probably a bit more complicated than that beyond a high-level view. What allowed the first- and second-generation Amigas to punch well above their weight in terms of perceived performance into the early '90s was the way that the chipset was a collection of co-processors which sat on a well-engineered DMA implementation. Unlike the contemporary PC compatibles, Macintosh and Atari ST the CPU didn't need to drive everything directly - it could point Agnus, Denise and Paula at a memory address and they could process data themselves to some extent with little or no subsequent intervention. You didn't really start seeing anything like that in terms of the competition until the early '90s, and the machines that could take advantage of it (e.g VESA) tended to be on the expensive side until prices started to come down in the middle of the decade. The problem was that the chipset which underpinned the Amiga's early advantages was never properly updated until it was too late because CBM as a parent company couldn't find their arse with both hands after about 1988, and that chipset - designed around the state-of-the-art in 1982-83 - ended up holding things back when the competition started catching up.

[@TheRealWinsletFan]

Not sure I'd class it as an advantage to have to use Bobs rather than hardware sprites such as on Megadrive. Megadrive was a monster with sprites and tilemaps, it mainly lost out to Amiga when doing other stuff.

[@off1k]

For me it really depends on the game, biggest difference is sprite flickering and sizes but HW sprites have the performance. Many MD games have sprite flicker which is a pet peeve of mine, I mean it's not like the 8bit consoles flickering but it's there for the action packed games. MD sprite limit was 80 total with 20 per scanline as long as they were no bigger than 16x16 pixels. MD "normal" max size sprite 32x32 but at 5 per scanline and they took priority. MD exploited max size was 32x64, not sure of per scanline, assuming 5. I think between 16 and 32 pixel size were limited to 13 per scanline. MD sprites 16 colours or 15 and transparency. Amiga Bobs cons: Use more CPU cycles compared to Amiga and MD Sprites

[@joehoy9242]

@@off1k - Also Gremlin's "Lotus" series, which makes very clever use of the Copper and Blitter to effectively triple-buffer the display - the MD version may have had more colours on screen in-game, but it wasn't remotely as smooth in terms of the road drawing routine.

[@off1k]

@@joehoy9242 I'm not too sure if the MD version did have more colours, maybe but maybe not, it's real close I think. The big difference is the resolution, the MD version is in it's lower res mode of 256x224, it's why the cars looked stretched. Also less objects on screen.

[@joehoy9242]

@@off1k - You know what? I just double-checked and I think I must have mis-remembered; I may have got it mixed up with "Top Gear" on the Super Nintendo. On the Amiga version, it's using 4-bitplane mode and all the BOBs are limited to four colours each, because the draw routine is using Shaun Southern's entire bag of coding tricks in order to wring every last cycle out of the blitter and copper it possibly can - it's drawing two frames ahead, with the visible frame in the viewport and a second, already completed frame waiting to be displayed while a third frame is being drawn. The end result being that Lotus 1 and 2 remain unmatched in terms of speed and smoothness on the platform, to the extent that you don't really notice the somewhat limited palette. Apparently the first two games also exploited a quirk of the OCS/ECS chipset to eke out a little extra redraw performance. With the release of the full ECS chipset in the A3000 and the AGA chipset a couple of years later, that trick no longer worked, which is why Lotus 1 and 2 will not work on those machines without a patch. Closing this loophole was apparently the reason why Lotus 3's draw routine was ever so slightly slower.

[@off1k]

@@joehoy9242 Interesting about Lotus 3 and difference to the first 2. They were a clever team and showed what many peeps thought to be impossible and make a Sonic style game with Kid Chaos. Another game with plenty of tricks thanks to the copper.

[@ArneChristianRosenfeldt]

Why don’t coders use sprite collisions much? 12:34 With the two playfields on Amiga even collision with the background should work. But what if you want to slide along? Penetration must not be visible on screen.

[@remykamermans5544]

I had a a1200, loved it outside the fact that the system aga chips barley used. Most games i played where all playable on a A500. Best compared with a sega megadrive and pc of that time had a child.

[@ren7a8ero]

I love seeing these systems that I never had the chance to play, engineers and programmers were wizards. It really puts on perspective the beefy hardware we have nowadays, to keep our current machines as optimized and efficient as they could be, because back then everything was put cleverly to use.

[@ArneChristianRosenfeldt]

Isn’t this the competition between Unreal and Unity3d: who has the better Wizard programmers? Bitcoin. Scaling websites on as cheap as possible hardware (horizontal scaling).

[@StuartQuinn]

I could listen to that Turrican music all day long. And in fact, I plan to!

[@ArneChristianRosenfeldt]

The 8 sprites only use one DMA channel in the side borders of the screen. Ah, yeah, weird, do they somehow map to the DMA channels of the bitplanes? The numbers don’t add up. Popular modes use 4,5,or 6 bitplanes. 10:46 Unfortunately, 16 color sprites are made by overlapping two sprites. So you lose one sprite. I have not found out if partial overlap works. Then use copper to move the parts left and right each scanline to encompass weird shapes. In reality I Wonder why PC engine can have 16 sprites of this quality and still overscan the background. Is chipmem so slow in OCS?

[@joehoy9242]

Don't forget you can use the Copper to multiplex the hardware sprites.

[@ArneChristianRosenfeldt]

@ I don’t know why Copper is not just Turing complete and counts as software. Then we know what is possible and concentrate on real hardware constraints. I don’t get why copper can reposition a sprite multiple times per scanline, but coders lose a line when multiplexing. There are only 200 lines on most games. So one lost line is annoying.

[@Kig_Ama]

Wowwww just wrote a comment a few hours ago and here it is, great work!

[@ArneChristianRosenfeldt]

14:58 Commodore was the master of R2R DACs Integrated into digital ICs. And then they feed this signal through (integrated) MOSFET gates to mix analog signals. Why don’t they go full PWM on the 28 MHz crystal clock? After all, 640px are also far above 7 MHz. So Commodore Chips are faster than 68k fabricated by Motorola? I wonder if it is possible to use a ring oscillator to create 4 7 MHz signals with a phase shift. Commodore tends to use wired-Or gates to lock these to a single clock: the clock signal on high floods the pull-up transistor and prevents a low wave from passing. The result is a potential very fast circuit still locked to the fast crystal. Additional phases allow finer pulse width control.

[@TheGeekPunkGamer]

What the names of the games featured in the video?

[@Great-Documentaries]

Dude doesn't even know why the Amiga was clocked at 7.16Mhz, but he's going to tell the rest of us "how it worked." Yeah, right. You also don't seem to get that its killer app was desktop video, and THAT is what it was designed for. No one was paying the modern day equivalent of $3800 for game PC in 1985 (the Amiga 1000). Yes, the A500 was for that home market, but it was an afterthought and a response to the Atari ST.

[@joehoy9242]

Yes, the A500 was pitched to compete with the Atari ST, but it most certainly wasn't an afterthought! The problem with CBM was always that after Irv Gould ousted Jack Tramiel, the former became utterly paranoid about the potential for a CEO to threaten his position the way Tramiel did. CBM pulled off a bit of a coup in effectively swiping Amiga Inc. lock, stock and barrel from right under the nose of Tramiel's Atari Corp, but as a company CBM was somewhat rudderless. The A1000 was a brilliant piece of kit technically, but its price point (pitched between IBM PCs and the new Apple Macintosh at the higher end and the Atari ST and 8-bit machines at the lower end where CBM had traditionally been strongest) plus its limited expandability meant that it was neither fish nor fowl in terms of market position. When Tom Rattigan became CEO, he very quickly put his finger on the issue and led the drive to separate the Amiga product line into the A500 - cost-reduced but effectively as powerful and as expandable as the A1000 - pitched at the home market, and the A/B2000 with far greater room for expansion aimed at the professional market. Far from being an afterthought, I'd argue that move probably saved the platform, at least in the short-to-medium term. Of course, for his efforts and success, Rattigan was defenestrated from CBM by Gould around the time the new machines were launched, which was arguably the beginning of the end.

[@ArneChristianRosenfeldt]

TV never bothered with a palette. Real video systems always had true color. NTSC has a low horizontal resolution. I dunno why commodore hates fast page mode. Buffer could be filled in the side borders. 16 bpp 256px could have been . Possible. CPU runs in fast RAM .

[@mcjdubpower]

Gud vid 💯💥

[@ebanite]

sprites can have 16 colors, only by reducing the number from 8 to 4, innfact to have 16-color sprites (4 bitplane), you have to set the sprite register bit, (attached sprite) to on, sprites were divided into pairs, each pair shared 4 colors, and could be joined to have a 16-color sprite.

[@cpu_UP]

Paula (sound chip of Amiga) was not able to do FM, only PCM.

[@cpu_UP]

Adding some extra (I made a few trackers for Amiga back in the days): 4 channels but 2 left and 2 right on the stereo channels. 8 bits samples. Max sample rate was 29k, so max frequency possible was around 15k (not top but not bad), min was 20 Hrz, so also quite low frequencies were possible. As "depth" was possible (using the CPU) to use the volume together with the Paula sounds channels to extends the bits fo the sound up to 14 bits (but was hard to do it stereo, cause the volume control was not stereo.) Later many trackers were just mixing multiple channels together using the cpu (or also the blitter) to mix together way more than 4 channels, and then playing it on 2 or 4 channels. The best I remember (but using 100% of A500 CPU) was 10 channels mixed to 2 stereo channels. On A4000 You were able to mix up to 36 channels.

[@SerBallister]

IIRC Paula had a modulation mode where one channel could modulate the frequency of the next, it was rarely used.

[@Nebulous6]

Paula does support AM and FM. See the info on the audio channel Attach bit and how it switches modes so that the other channel changes from handling individual bytes to instead using modulator words on p. 149 of the Amiga Hardware Reference Manual.

[@nickolasgaspar9660]

The workings of Amiga are pretty close to how Jay Miner's previous 8bit computer worked! The Atari 8bit had Antic for Agnus and Pokey for Paula. Antic had DMA with its own set of instructions taking care of scrolling and scalling like Copper and Blitter did in Agnus. Hardware sprites were similar in dimensions (A8 had 8pixels x full vertical screen, 4 players 4 missles, 1 color per sprite while the A500 had 16pixels x full vertical screen, 8 sprites, 16 colors). Paula had 4 pcm/8bit sound channels while the Pokey chip had 4 channels, 2 pcm/ 4bit channel that could merge and reproduce 8bit quality samples. Color wise A500 had a huge palette compared to her smaller brother but the A8 color capabilities were way ahead of its time. From 256 color palette a user could use 5-9 colors on screen and with DLI(Antic) and color sprites one could display hundred of colors with limitations in their location on screen. That resulted in having games with more than 60 colors on screen and with multiple resolutions along the height of the screen. Both system elevate their performance by adding memory expansions. It was the only way to take max advantage of their chipsets. Both systems built a great name in video processing while the A8 won an EMMY Award as the first computerized teleprompter( and the only 8bit machine that can play 256 color video/50fps with PCM sound). Looking at their similarities one can understand why Atari sued Commodore when A1000 was released but we were fortunate enough to have Atati loose that battle, as it should in my opinion, and the rest is history.

[@joehoy9242]

By 1985, Atari had effectively been split into two companies - Atari Games (which was effectively the arcade division) and Atari Corporation, servicing the home and business market, which was to all intents and purposes an entirely new firm headed up by former CBM CEO Jack Tramiel and his sons after the former was ousted from CBM. Jay Miner quit Atari long before that happened - the reason Tramiel tried to sue CBM was because Amiga Inc were in financial trouble - he wasn't interested in the team, just the chipset - and he'd tried to take advantage of their money worries by low-balling his offer to buy the technology to an almost insulting level. CBM had been haemorrhaging engineers since the C64 project was completed, and Tramiel took a fair chunk of those that remained to Atari Corp. with him - CBM needed to address that issue with alacrity, and buying Amiga Inc. outright (and for a much fairer price than Tramiel was offering) provided a rather neat solution. That said, I'm also fairly sure Irv Gould was motivated to spend more to acquire Amiga Inc. at least in part because he knew it would drive Tramiel absolutely out of his mind.

[@tabsntoot]

ganja farmer was one I remember and cannon fodder good times

[@MrMarianoamigo]

in the perfect way!

[@atomicskull6405]

Please do one on the Sharp X68000

[@jordanwhite352]

Honey, cancel all plans, the latest Technically Correct just dropped!

[@itxptube]

A Christmas Classic!

[@Galahadfairlight]

Chip names were female themed so Agnus not Angus, Denise not Denis. Lowres could be a lot larger than 320x256, Lowres overscan was the entire screen and outside screen borders. 4096 colour's not 496. Tiles are a specific console hardware feature, Amiga used blocks. Sprites could be multiplexed to cover lots of the screen. Difference between sprites is the background didnt need to be preserved and restored, with BOB's you would need to.

[@ArneChristianRosenfeldt]

Sprites have color keying (like the playfield) instead of an alpha channel. Sprites don’t pay as heavy for alignment problems due to the bitplane nonsense. Sprites have their own palette.

[@kirishima638]

Too many errors and mis pronunciations. Disappointing.