CIA 6526/8521/8520 differences

[slingshot]

I searched, but still didn't found a comprehensive list about the differences of the models. What's known by me:
- TOD is binary on 8520 / BCD on others
- Just found this: writing to the timer hi registers when the cra/crb[3] is set (oneshot mode) starts the timer on 8520/don't start on others. This is used during early boot, maybe the reason why a CIA from C64 cannot replace CIAA? Someone else can confirm this?

[ross]

Originally Posted by slingshot

- TOD is binary on 8520 / BCD on others

Not only is binary (and not BCD) but also with only 3 registers available.
On Amiga, Kickstart use it.

- Just found this: writing to the timer hi registers when the cra/crb[3] is set (oneshot mode) starts the timer on 8520/don't start on others. This is used during early boot, maybe the reason why a CIA from C64 cannot replace CIAA? Someone else can confirm this?

Yes, true for 8520.


So 6526/8521 cannot be used on Amiga.
Vice versa 8520 can be used by some extent in C64
(for what I remember the C64 Kernel does not uses the TOD, but I could be wrong because I haven't programmed the C64 for a long time ).

[slingshot]

Then I wonder if a 8520 in C64 passes the Lorenz CIA tests (except oneshot). Or the VICE tests (is it has the same timings as the 6526/8521)?

[ross]

Originally Posted by slingshot

Then I wonder if a 8520 in C64 passes the Lorenz CIA tests (except oneshot). Or the VICE tests (is it has the same timings as the 6526/8521)?

Sure it fail the TOD tests

6526 have also a TODIN bit in CRA not existing in 8520, for accurate time in 50/60Hz.
8520 simply count external event (24-bits): on Amiga CIA-A count 50/60 Hz, CIA-B H-Sync.

For timings you mean operating frequency? I suppose 8520 is rated for 1MHz, underclocked on Amiga at 7xxKhz.

[slingshot]

No, I mean internal timings. E.g. after a register write, when the timer actually starts to tick, when it's reloaded exactly and so on. There are a lot of tests for cycle-accuracy of the CIAs in C64, but I'm not aware of anything on Amiga.

[ross]

Originally Posted by slingshot

No, I mean internal timings. E.g. after a register write, when the timer actually starts to tick, when it's reloaded exactly and so on. There are a lot of tests for cycle-accuracy of the CIAs in C64, but I'm not aware of anything on Amiga.

Well, I can only assume this, but I think the internal timings are the same.
Too few differences to justify something substantial...

However you have to try it

[nogginthenog]

Originally Posted by ross

On Amiga, Kickstart use it.

What if you used a 6526 for odd and a 8520 for even CIA (or vice versa). Is TOD used on both chips?

[ross]

Originally Posted by nogginthenog

What if you used a 6526 for odd and a 8520 for even CIA (or vice versa). Is TOD used on both chips?

Unfortunately both CIAs TODs are initialized by the KS.

[slingshot]

6526 doesn't work as CIAA because of the Oneshot mode difference (when Kickstart writes to Timer A HI byte, the timer don't start, and it waits forever). But it wasn't tried with real chips, just with FPGA replicas only (however the 6526 model is a very accurate replica).
Reportedly Amiga's 8520 works better in C64. I wonder if anybody tried the VICE tests on it (without the TOD ones).