As promised, here's the pinned errata/clarification - I'll add info here about anything I didn't get right in the video - apologies! Quite a few people have pointed out that the screen clearing behaviour in KERNAL v2 was actually to help hide the "screen sparkle" problem that former Commodore engineer Bil Herd describes in this video around 15:45: kzfaq.info/get/bejne/Y8CgqZmmmtynemg.html Rob Clarke says: Your experiments with the 4064 kernal would not work that way on the real machine. There is no color RAM, so you cannot change the text colour, the board is hard-wired to only return white. Also, the Educator 64 and 4064 are fundamentally different machines. The Educator is a normal 64 with a SID and a monitor that can display different luma levels, ie. "shades of green". The 4064 is a standard PET monitor, so the video signal is either on or off, so true monochrome, no shades at all.

That was really interesting, Robin! I learned a few new things!

If I ever run into you at a conference, I'd love a selfie of me and just your KZfaq hand. :D

The original Kernal was set to white to work similar to a PET in that it was just no color. What was found though is that the character ROM was a bit too slow and when graphics and sprites were being drawn the screen would get white snow in the image. This would often cause the sprite collision register to fire a collision detect and made many early games unplayable. At first Commodore had us techs adding a resistor and small capacitor to the video chip to try to stop the snow which only marginally helped. They then started sending us Kernal V2 which changed the default character color to the current background color. This eliminated the snow by making it the same color as the background. I actually figured out it was the character ROM being too slow when I got sick of having to fix the C64's we were selling and I started changing chips. When I found the character ROM had to be 150ns or faster I purchased a bunch of 65866 (I think??) 150NS eproms, these are the ones that fit the pinout of the C64, and burned the character set on them. These fixed all of the 64's that came in..no snow at all and no oddball sprite collision detections. I reported my findings to Commodore in Shamburgh IL and shortly after that the character ROMs were improved and Kernal 3 was released, changing the default character color to the current character color. Those were fun times.

There are still many nervous ex Commodore employees worried that Robin will uncover their secret message they hid 40 years ago.

8:15 Collaboration with Adrian's Digital Basement maybe? :)

12:00 My guess: it jumped into the tape routine and crashed out of it, while already having disabled interrupts. Triggering the tape routine again and finalising it correctly will re-enable interrupts. I don't know how many hours of C64 I watched on this channel, I'm still not bored.

As a Turkish citizen who lived the 80s playing with my and many of my friends' C64s, I can tell you that I had never seen a Turkish kernel before. I can recognize the company name shown there "Teleteknik A.Ş.". "A.Ş." is the abbreviation for "incorporated company" (Anonim Şirketi) in Turkish. That company was the sole distributor for 8bit and Amiga Commodore computers in Turkey until the end of 1992. I believe the reason characters seem wrong is that it appears you need the Turkish character ROM too, which you can find on the internet. But this is the first time I'm seeing the Turkish character set ROM being used :-) English alphabet is 26 characters, Turkish is 29. Missing (or unnecessary) is x and w, and additionally, there are "ı, ü, ğ , ş, ö, ç". Also, ı I and i İ are separate characters. But you can substitute English characters for the missing Turkish characters. So it seems to be a real unnecessary feat to have a Turkish character ROM which you would need only for Kernal anyway since for a program you can do whatever you want. However, I really don't remember seeing this kernel neither at a friend's house nor at a fair on a commodore back in the day and ever :-) Probably this was made as a proof of concept or maybe as a show-off?

That kernal rom bug that would cause the C64 to lock up after scrolling to the bottom of the screen and scrolling over two lines of text was so frustrating. I cannot count how many times that happened while I was typing in long BASIC program listings and losing my work because I had not stopped to save my progress to a data cassette. It instilled in me the notion of saving often. Apparently, the workaround for this bug was to simply change the default colors (background blue, foreground light blue) to something else (e.g. background black, foreground grey) before typing a BASIC program listing. I was never aware you could use the data cassette to unfreeze the Commodore 64 after it hung from that kernal rom bug.

Now that we've seen so many versions of the ROM I'm really looking forward even more to your in-depth review of Compute's Toolkit Kernal book from 1985

24:09 The text says "Teleteknik A.Ş", which was the distributor of Commodore 64s (and Amiga's) in Turkey. They published Commodore magazines and scheduled events about Commodore and Amiga also. Went bankrupt in '92.

One of the first rules of computing that I learned: Save when it's more trouble to recreate what you've done than it is to save.

Longstanding mystery solved! When I was a kid, I borrowed a Cosmi game called 'Navy Seal' from a friend. The briefing screen before the mission would print text on the screen on my friend's C64C machine, but on my C64 breadbin the text was not visible even though the game worked otherwise. I must have had an -02 kernal in there.

0:19 Ah, yes, the era before ubiquitous spellchecking. 4:12 The VIC-20 had a screen-fill character color of White. We'll assume they just didn't change this from the VIC, though the VIC had a default screen-background color of white, rendering naked screen POKEs invisible. 4:42 There's also lots of color-fringing around the letters. I don't know how people could stand that on the Apple II or the TI-99/4A - or with color-fringed sub-pixel fonts today in Windows! 6:00 Interesting - that's $FF80, one location before the Kernal JMP table. 10:50 What if you change the background color to white, clear the screen, and change the background back to blue? 12:44 The text is a lot more clear now. 13:22 I remember reading that they used the background color because using any contrasting color increased the "screen sparkle". 14:35 $FFF6, the last four bytes before the interrupt-vector table. 21:56 Tautology. 22:12 Seems like an endless source of incompatibility. 22:40 It'd be a spot to put a Japanese character set, though if the Kernal didn't set this up itself, they could just assume that software would take care of it in a way that lets the base machine be compatible with the rest of the C64 world. Did the Japanese C64s have a special character ROM?

This was an extremely informative and fascinating video. I wish you were around when I was a kid getting into all this stuff, I would've learned a hell of a lot from you!

Interesting that the Turkish kernal uses mixed upper and lower case characters in the banner.

Great stuff here Robin, thanks for sharing! Keep being awesome, and merry christmas!

Enjoyed the video, now I want to check what Kernals I have in all my various C64's. Quite enjoyed your ending Song with the words being based on the different Kernal version quirks!!!

Another great video! Love seeing all the changes the made

Awesome as usual Robin. Never loose that thoroughness!! :D Merry Shrismas!!!

The screen clearing routine ($E544) works differently in the second revision because there was a problem with the VIC where errant pixels of foreground color would flicker randomly on blank characters. There wasn't enough time to fix the hardware, so for demonstration purposes, the screen clearing routine was temporarily changed to fill the color RAM with the background color.

Thanks to that "KERNAL" spelling business, I used to spell it that way for all OS kernels. At the time, I thought it was like bite vs. byte...like it was computer-specific jargon. Thanks Robert Russel! :/

That's cool that on the SX version of the 64, the default drive from direct mode is the built-in 1541. Makes perfect sense!

Robin, I love your videos. What an awesome video. Merry Christmas, brother.

Thanks for this video. Happy Christmas!

Super captivating and informative. Thanks!

I took my SX-64, a black-and-white tv, and a 2400 Baud VICModem to University with me and wrote, uploaded, compiled, executed, and debugged Pascal and COBOL programs on the University mainframes from my dorm room. I had a question about one of my Pascal assignments and sent a VMSMail message to my Professor. After a brief mail exchange, he commented that it looked like I was using a 40 column screen editor. I told him what setup I was using, and, after a brief pause, he said, "Why are you in my class?"

You should cover the C64DTV + easter eggs! Especially the hidden messages!

Very interesting! Did not know that at all. Merry Xmas to all 8 Bits! :-) Greetings, Doc64!

Great video. I always found it strange that Sweden got a localized keyboard fof the C64, but we in Norway had to get by with the standard english keyboard layout. Oh well. It didn't keep me from loving my C64 :)

Great Video, I was watching it early in the morning and had to get out of bed to see what KERNAL's I owned. My original C64 as a kid was a 2. My other one is a 3. Good Stuff!

Always important to stay watching till the very end for that little piece of magic. ☺️

Re the KERNAL version byte: the Apple //c had weirdness with the byte indicating the ROM version as well; it started at 255, then 0, 1, 3, and 4. I wonder if in both of these cases, they just picked an unused ROM address and gave it meaning starting with the first revision after the original.

another great video, thanks for sharing. Merry Xmas

BTW - “Greyscale” is a kind of “Monochrome”. I may have misunderstood you, but it seemed like you were implying that monochrome was a binary color space consisting only of Black or White. Monochrome just means what it says on the tin mono + chroma, “one hue” or any color system that consists of only Black and White or a single hue and any of its shades at various degrees of saturation and brightness. That’s not to say that monochrome color can’t be digital or even just binary black and white; it’s meaning is just more broad. If you already knew that, I apologize. Love your videos, I’ve learned a lot from them.

ehy Robin, the final song is really good! just to relisten to it I watch again the whole thing! cheers

Merry Christmas & Happy New Year!!! :-)

Thank You very much for Yours videos👍 Greetings from Poland!

Just have seen your video. Excellent explications as always. I noticed how old my C64 really is. And that Commodore really used what they accidentally found in their shelves. I have a 1983 250407 mainboard with a Version 3 Kernal 0584, a CPU from 5083 and a VIC-II from 0685. Think that one was replaced by service.

Great video. We always used the poke trick to see which kernel we had. I had a tape turbo that was probably developed on v2, because before showing any graphics you would think it was loading on a blank screen with border color changing. But the same program would spit out bunch of garbage starting $0400 on my Commodore before showing loading graphics :)

Another great video, thanks!

It's cool that you still own your first C64. I no longer have my very first computer, a ZX Spectrum, nor do I have the Amiga 500 and Amiga 1200, both of which I loved very much. But I do have a model of the very first computer I've ever used, a ZX-81.

THANK YOU for not saying "RUNSTOP RESTORE". I'll bet you never say "instant delete" either. I have a '64 from '82 which doesn't have the silver label. Oddly it's early enough to have the buggy VIC II (with snow). Sadly the clock generator went haywire and IIRC most if not all custom chips died. I bought by first '64 when the 64C first appeared in stores. When I bought it, the original 64 and the 64C were both on the store shelf, but there was a promotional discount and the 64C was cheaper! It had a board in it that was meant for an original '64, not a 64C board. Sadly that one gave up the ghost in 1994 and all that remains are the keyboard springs (I have no idea why). For years the metal Commodore 64 label from it was stuck to the dash in my car. Thank you for this video. I recall reading about the different ROM versions in a magazine, but completely forgot the differences.

Commodore could probably have made an additional small fortune if they'd socketed all the ROM chips and sold updates for advanced users.

excellent video, learned alot i didn’t know 🙂

Very cool video !!!

I was thinking that it would be cool to start reviewing programs that came in compute's gazette. Two really cool things that I can recall off the top of my head were something called CHR$ Graphics (Compute Gazette - Issue 73 -July 1989) and The Great Arcade Machine (Compute Gazette - Issue 68 - February 1989). Really cool stuff to support basic etc. I used CHR$ Graphics to make a pool or radiance/The Bard's Tale clone to adapt the Super Endless Quest #1 Prisoners of Pax Tharkas into a game. I actually got very far, intro, character building first part where you make it to Solstice. Unfourtunately my Dad had me throw all my stuff out in the mid 90's so I lost all my stuff. I also used the Great arcade machine to make a ninja shooter where you could throw ninja stars to kill samurai running around on the screen....really fun memories hours spent working on that stuff...

Thanks for sharing!

Very informative and interesting. I'm now pretty certain that my old 64 is a revision 2. It has a smaller manual that mentions the compatibility with Max cartridges.

Interesting video, thank you!

Excellent video =D

as always, a great interesting video. mhh are there also different kernel versions on the c128/c128D?

I'm thinking about changing my middle name to "I'd like to have one just because". Very cool, I think I find the Japanese version the most strange. That 2k is pretty important.

Amazing minutia!

Unique video!

Several hundred thousand units were made in Canada (I own 4), and all of mine don't have the "made in Canada" sticker covering up a "made in USA" sticker - they just have a single original sticker saying made in Canada. Many Commodore fans forget that Commodore was founded in Canada and was originally a Canadian company (and hey, if that blows American viewer's minds, then they should look up the origin of IBM - also Canadian). Great trick with the PEEK(65408) -> The C64 Memory Map doesn't even show that! FYI - Vice shows "3" as it uses version 3 rom. Now you've made me want to check if I can run Jiffy DOS on Vice. :)

Commodore's German manual referred to Kernel 2 for years. To every POKE into the character RAM there was a POKE into the color RAM. I don't know if they ever changed it. As Bob was missing in the 4064 Kernel: I think these machines were sold without SID, as they shouldn't be machines to play with. I also think this Kernel disables the sprites every time it sets the background color to black. So it would be logical removing the BY.

The 4064 ROM - I just looked and it has a loop, not only zeroing the background and border colour but also the remaining VIC-2 colour registers. So any attempt at playing most C64 games could see you with all black sprites if the default IRQ routine was left in place. The attribute clear-screen thing - I got bitten by that, wrote a game which relied on bars of the sides of the screen using special characters to force a collision to prevent your ship going offscreen, but I had no idea at the time of kernal revisions that didn't set the cell attribute to background colour. C128 - I do suspect by that time they'd have gotten things right and settled on the one kernal version.

Its always nice to have your original of something even if it doesn't work I own 6 NES's but none matters as much as my original one (which I use in my living room) I also own 9 Gameboys, and though my original doesn't work I keep it because its my original.

IIRC according to some video I watched ages ago the C64GS was more expensive to produce than the regular C64 too. Commodore did some weird things sometimes on the business side of things.

Hey Robin. I have a memory of a bug that I think is relevant. I thought maybe you could test for. The dates are best guesses, as this was a long time ago. I got my first computer in 4th grade, 1984. It was a Commodore 64 with the 1701 monitor and 1541 DD. One of the habits I picked up fairly quickly was to poke both the background and border to black. Around 7th grade, 1987 (I think), I started having issues with my C64. My uncle said he could fix it (he was good with electronics). But instead of fixing it, he swapped it out with one he had. The one he gave me, would not accept a poke 53280 to change the border color. If I recall correctly, there was a work around, but I cannot remember what it was.

Very interesting the bug in the first versions. Terrible the computer crash... after a long programming session it was an immense joy to see the update saved on the disk😓😀👍

Really weird, in a local electronics-store there was once some empty C64GS, so everything was missing, just the empty plastic case... bought one but never did anything with it... was just some bucks back then...

Okay, a bit of an update regarding the "sparkle" in early Commodore 64's. Bil Herd said that some of the original 6567R56A chips (the VIC-II chip) caused one type of sparkle after the chip heated up a little bit. However, Albert Charpentier (who is one of the _designers_ of the VIC-II chip) said that the problem was with the pre-charging circuit in the 901225-01 ROM chips that held the character set. It turns out _they're BOTH right._ There were two different kinds of "sparkle", and fixing one would not necessarily fix the other. If you only had the 901225-01 ROM, the sparkle could be fixed by installing several capacitors to change the bus timing. However, Jim Drew (who worked in the Commodore Service department) told me that because they were getting hundreds of machines per week for repair, they just started doing motherboard swaps with the Rev.B or Rev.C motherboard. They were literally getting palettes of replacement motherboards every month. They were instructed to pull all of the socketed chips from the Rev.A motherboards and discard the PCB itself.

Great video! Honestly, it's never crossed my mind that there are multiple kernal versions. Obviously there is, just strange I never thought about it before. :) Speaking of kernals - did the source code ever leak out from Commodore, before or after their demise? That'd be a fascinating read. Yes, I know I can just disassemble it, but that's not the point; was curious if there'd be commentary from the programmers and such.

Thanks!

Yep! At around 12:12 you showed a massive portable reset switch. Nice.

That lock up bug pissed me off sooo many times.

Great video :) One little question - what version of kernal(s) did the various types of C128's C64 mode have?

I love your videos

While re-watching this video, I just noticed that the power LED on your 64C is red, while mine is green.

I'd guess that the "C-64T" on that screen is shorthand for "Commodore 64 for Turkey."

I used to run into the bug with the lines at the bottom of the screen all the time, except I'm almost positive that mine used to automatically print the "PRESS PLAY ON TAPE" message when it happened. I made myself a reset switch and then would use a small Unnew program to get my BASIC program back.

Thanks for showing that invisible 1024 poke. I've been working on a c# emulator, and one of my first tests was 1024,81.. When it showed up blue, I was sure I had a bug in my color ram/VICII code. I certainly didnt remember my original commie doing that. You just explained why, so thank you for saving me head bashing time!!

I thought that business of the character colors defaulting to the background was unique to the VIC-20, and didn't happen with the C-64 at all. I guess I was wrong. Anyway, the clock frequency of an NTSC C-64 turns out to be exactly 45/44 MHz. But no literature seems to mention this. The two values given in the Programmer's Reference Manual are just decimal approximations of that value to different numbers of decimal places.

12:30 I would love an analysis of that corner delete freeze bug. There must be something rather weird going on.

The rev2 initialize to dark blue/dark blue was specifically to address the problem of then contemporary VIC chips which had color clock synchronization issues, compounded with marginal RAM timings, which would ultimately result in characters "sparkling" when printed. It was literally a "get this done by end of today or you're fired." solution.

You mentioned that the EasyFlash can't replace the CHARROM. It's running in ULTIMAX mode this is why the HIGH BANKS appear over the KERNAL ROM in the 1st place. If you POKE 56834,4 when using the KERNAL replacement, it will revert to the real ROM value.

Great video as always! :) 9:29 Wasn't this the fix for the sparkle bug? Check out Bil Herds talk "VCFMW 11 - Bil Herd: Tales From Inside Commodore", 15:51.

I have a theory why Bob Yannes initials are missing in the 4064 Kernal ROM. First the 4064 and the Educator 64 are not the same and they have not the same Kernal ROM! The Educator 64 was a standard C64 board with standard ROMs in a CBM case (with an extra speaker added inside). The CBM4064 had this special Kernal and the boards had no SID Chip (and no speaker). The chip is not just missing - it was never populated on these boards. The missing SID chip could be the reason for removing the initials of Bob Yannes, right? BTW: All Educator 64 were NTSC and all CBM4064 were PAL but there was an other C64 in a CBM case. The PET64 was the NTSC version of the CBM4064 with the 4064 Kernal and a board with no SID chip (and no speaker). If you have any questions you may contact me via my website: www.c-64.org Thank you for your awesome videos! I watch all of them!

Maybe I'm reading into too much here, but the KERNAL responding with 170 reminds me of the early VIC-1540's when asked to report their version. They say V170. I wonder if that's related...

Robin, I would love to see your JiffyDos video ...

There are a lot of differences between the various revisions. Multiple breadbin designs, different long and short board revisions, and at least 2 64C style cases. Prefer the 250466 board.

It's interesting that with the SX-64 the VIC-20 color scheme is back (but not "back again," since the idea of a return is built into "back").

I actually did some research and teleteknik AŞ(distributor of commodore computers) released a Turkish f keyboard c64 named Commdore Türkçeyazar(writes-in-turkish) which I guess is the Turkish kernal you showed in the video. I have never seen one here in Turkey though. Must be super rare EDIT: Welp, now that I have delved deeper into the comment section other people wrote the same things that I wrote above. Better luck next time I guess.

I'm curious about the Japanese C64. Was the 2k removed from Basic used by the system in any way? An extra 2k at that location would have been very useful for assembly programming.

Ah, ok, I thought my C64 was broken because when I poked into the screen a character, it would not show up until I changed the color value

22:25 Moving Start of BASIC ram up by 2K might improve Vic-20 BASIC program compatibility, but it would harm the ability to run C64 programs in 2 common cases: 1) BASIC programs that needed more than 36K 2) Mixed BASIC/Assembly programs. For example programs the jumped into ML like: 10 SYS2070

AC/DC reference checks out.

Here in the UK i had heard about a C64 Rom cart only version i never bought one due to the lack of the Keyboard and thought what was the point. I know it was instant play time from a cart and you can get more memory for better games . I had Robocop on both tape and cart and the cart version was so much better. Thank`s for the flash back can you believe i sold my breadbin c64 and all my tapes , carts for £40 trust me i had a massive box of games 🤦‍♂️

One ROM fix that Commodore should have made early on, in both the C64 and 1541, was to speed up the floppy drive data transfer speed. The reason this was slow to begin with has a potentially long explanation that involves something very dumb that someone did to break the C64's board design (in a nutshell, it was supposed to have had the serial "burst mode" that the C128 had from the start, but some critical lines were removed to make room for a new mounting hole 🤬), but the problem still could have been fixed in the KERNAL and 1541 ROMs after release. Had this been done in a timely manner, then compatibility issues would have been very minor, if there would have been any at all, but apparently Commodore management didn't care. Obviously, third-party solutions were developed, but this really ought to have been fixed early on by Commodore and rendered a non-issue. By the way, on an odd historical note, it appears that Commodore did make their own fast disk transfer protocol for the C64 at some point, so perhaps they were thinking about fixing this issue. I know (or think I know) this because of the game _Mail Order Monsters_ , which states "FAST DISK LOGIC COURTESY COMMODORE BUSINESS MACHINES INC." on the initial loading screen. I'm not sure how else to interpret the meaning of this statement, other than Commodore themselves providing a fastloader for this game. Now, anyone who has played this game on the C64 wouldn't think that it loads very fast, but for reference I've had the C64 version running side-by-side against the Atari version, and there isn't much of a difference in the loading times. The Atari has a slow disk transfer rate out of the box, as well, although it is significantly faster than the C64's, so this fastloader of Commodore's does make a big difference, even though it is much slower than third-party fastloaders. I don't know when this fastloader was made, but this game was released in 1985, so maybe it was too late and too slow to make a real difference, and Commodore shelved it. I still say they should gave gotten on this immediately and provided updated ROMs, even if they would have had to source the faster protocol from a third party.

Everyone, take your time to listen to the lyrics of the song at the end.

that swe C64 is "normal" C64 (for me atleast ;-) )

Funny how the Japanese C64 colors are the same colors the VIC-20 used

I heard that kernal version 2 changed the color to match the background intentionally because it hid defects in the early VIC chips. After the defect got corrected the change wasn't necessary anymore. But I remember Bill Herd saying it was like glittery snow flakes. After it got fixed people were stealing vic chips at commodore. lol

It sure looks like the version 2 LOAD bug uses SEI so only NMI can get through.

I find it odd that the colors in the Commodore logo on the Games System boot screen are incorrect.

How do you do your C64 screen captures for the videos?

Does the NEW comand erase basic memory? I always ignored it when actually writing a new basic program after turning it on but the gmagazines always put the new command in there at start

Does adjusting the output gain trimpot for the video output improve the quality at all on rev 1?