CIA 6526 vs 8521 (was: C64C case production molds resurfaced)

Date: Wed, 04 Mar 2015 23:09:53 +0100
Message-ID: <1819950.C1zDhFobHZ@rakete>
(i copied some mails into this one to avoid tons of small replies)

On Wednesday 04 March 2015, 14:41:50 David Wood <> wrote:
> I've heard of (but not seen in the wild) a method to separate the shiftlock
> key from left shift.  As I remember, if you drive the line for shift
> against the key, the resistive element of the left shift key will read
> true, while the shift lock key's switch will not.

yup. see the test program i mentioned in the other mail:

On Wednesday 04 March 2015, 20:47:55 Gerrit Heitsch 
> Probably to keep customers from complaining... Especially since the 8521
> is not quite the same, if I remember right the timer IRQ is one cycle
> late compared to the 6526. Doesn't really matter most of the time,
> unless you use it where the timing is REALLY tight.
> This little tool will detect the CIA (and other chips) you have in your
> box, a 8521 will be listed at '6526A':

note that it should say nmos vs hmos or perhaps even 6526 vs 8521 instead (see 

On Wednesday 04 March 2015, 20:48:09 wrote:
> But is this something fully reproducible? Or a kind of an esoteric
> behaviour? Like that one opcode that doesn't work well on R6502s but we
> still consider the CPU to be fully 6502 compatible.
> If this is somehow fully reproducible then it's a valid point when
> talking about "100%" compatibility.

its fully reproduceable. (in fact, the VIA datasheet even shows the 
differences exactly, and also explains the consequences - its a safe bet that 
CIA is very similar... someone who can read the die shots should verify it 
some day :=))

On Wednesday 04 March 2015, 21:11:40 Spiro Trikaliotis <ml-> wrote:
> > And what if it is a 6526A? Or are all 6526As basically 8521s?
> It seems so. At least, the 6526A all have the difference that their
> timer fires one clock too early w.r.t. the "normal" 6526. Although I am
> currently not exactly sure if the 6526 is too late, or the 6526A is too
> early with respect to the documentation.

NO. thats a misconception that was very popular "in the early days". (probably 
because regular "old" 6526A were very rare "in the wild")

however, 6526A are selected regular 6526 that just happen to work at 2mhz, its 
exactly the same die as regular 6526 (or 8521 for that matter). there are 6526 
labelled chips with 8521 die in them (shows the timer irq difference) and 
there are 6526A with the regular 6526 die in them (does not show the timer irq 
difference). thats why we changed it into "6526 (old)" vs "6526 (new)" in 
VICE, using the "A" suffix to indicate this difference is not only formally 
wrong (it means "2 mhz") but also does not (always) work in practise.


Wenn die Menschen nur über das sprächen, was sie begreifen, dann würde es sehr 
still auf der Welt sein. 
<Albert Einstein>

       Message was sent through the cbm-hackers mailing list
Received on 2015-03-04 23:00:05

Archive generated by hypermail 2.2.0.