It has come to our attention recently that there is a new FX-9860GII series of calculators that got released in mid-2011, according to cfxm here, breaks compatibility with previous versions of these models. The new models have the label "USB Power Graphics 2" on them and the design has changed:



The SD version of this redesign also supports SDHC cards now. The compatibility issues that arises are even worse than the ones the TI-84 Plus series got in 2007 (most notably causing older versions of TI-Boy SE to not run on newer models). Because the new 9860GII models now use the same processor as the PRIZM (SH-4A CPU instead of the previous SH-3), most older add-ins will not run on them. Only add-ins that were made with the official but limited SDK will apparently run. While Casio apparently warned developers to avoid accessing the hardware directly in their programs in the past (unlike TI when they removed the extra RAM pages in the 84+ models then later released MathPrint OSes), this will still make it an hassle for people who get the calculator as well as developers, as several older add-ins will need to be ported to SH-4A models to work again.

Also, the OS that comes with it, 2.02, will not run on older models. Will Casio continue releasing OS updates for SH-3 FX-9860G/II models in the future?

On a better note, could this eventually make it easier to port FX-9860G programs to the PRIZM, despite other hardware differences between both calculators?

Well, I personally wonder why they didn't work on such Prizm instead of an old calc redesign that will be a nuisance for a bunch of people?

Oh, so Casio now makes the same kind of moves on some of its models as TI has been doing many times over the past ~13+ years ?

In the TI-68k series, here are the most important backwards-incompatible changes, but there have been more, minor ones:
* in ~1999, the HW2 TI-68k models grew a RAM execution protection, which was soon disabled by HW2Patch and worked around by program launchers;

* (AMS 2.xx moving around system internal variables and breaking the awful hacks using absolute addresses, which had worked for several years). Later, some people, among which I am, argued that such dirty hacks should never have been used. That said, it was probably hard to avoid entirely on the 92 series, which did not have a jump table, and maybe hard to avoid until there was sufficient knowledge of the jump table on the 89/92+.

* the V200 having 4 MB of Flash memory (instead of 2 MB, as the 89 and 92+), but located at the same address on the V200 as on the 89, which broke model detection (done through the address of Flash memory, for compatibility and efficiency purposes). AMS native programs (those which do not need a "kernel", and were already a majority at that time) made before the advent of the V200 broke, while the abstraction layer provided by "kernels", at some space cost (RAM consumption + Flash consumption + relocations in client programs) shielded "kernel"-based programs from most of the changes.

* AMS 2.07+ support the "clock" on HW2 and later, i.e. an inaccurate gadget, based on a relatively low-priority interrupt (AUTO_INT_3), which a number of programs disabled when they wanted to disable anything below AUTO_INT_5, the programmable timer. A number of such programs were changed in order to use SR=0x0200 instead of SR=0x0400.

* the 89T (HW3) broke even more stuff:
1) the 4 MB Flash memory was located at yet another address, which broke programs again. Said address left open the possibility of 89T models with 8 MB of Flash memory, and the OS (at least - I didn't check the boot code) was prepared to it: upon OS startup, it queried the Flash memory chip for its size, and stored the result (address of the first byte after the end of the Flash memory) in RAM
2) HW3 models (89T) removed some of the ghosts of RAM, which had been used to work around the RAM execution protection. Some ghosts are available at new addresses, but have never been used. On HW3 calculators, HW3Patch or equivalent (tiosmod+amspatch) is almost required (*) for multi-chunk programs whose chunks call into each other and return.
On HW3, a RTC was added, and AUTO_INT_3 was repurposed for USB communication - but this didn't break much code, if at all.

The only upside of that kind of moves by the manufacturer I can see, is that it forces some activity to occur in the communities. But obviously, not in the most efficient or pleasant way...


(*): well, it's possible to do without it, but hardly anybody did this in practice, because it significantly raises complexity and significantly slows down execution.

Thanks Lionel for the summary, didn't know it was that bad
On the other hand I hope CASIO doesn't take examples from TI ...

Quote from: DJ_O on December 13, 2011, 12:34:08 am

Well, I personally wonder why they didn't work on such Prizm instead of an old calc redesign that will be a nuisance for a bunch of people?

They're probably working on said Prizm.  I suspect that this was a cost cutting measure to use similar boards for different lines of calculators.  [/just a guess]

Quote from: ralphdspam on December 13, 2011, 12:50:45 am

Quote from: DJ_O on December 13, 2011, 12:34:08 am

Well, I personally wonder why they didn't work on such Prizm instead of an old calc redesign that will be a nuisance for a bunch of people?

They're probably working on said Prizm.  I suspect that this was a cost cutting measure to use similar boards for different lines of calculators.  [/just a guess]

I actually suspect it had to do with licensing and production costs. Their licenses on the old SH-3 chips are probably running out and they're moving their production lines completely over to the newer SH-4 chips which are likely more efficient anyway. It wouldn't surprise me at all to find out that there were a host of other devices running SH chips that were also being converted.

In any case, this is a perfect [if somewhat ironic] example of one situation where accessing the hardware directly and not using Casio's OS syscalls is probably more portable than the other way

I'd have to agree with this as well. I also noticed that last time I was at Best buy, the older FX 9860 Models were on clearance. I thought it was a bit odd, but I guess I know why now.

Also, I'm not surprised that the switch to the new chip breaks compatibility with old programs. I suspected as much when cfxm mentioned it <a href=http://ourl.ca/14448>here</a> I think it is nice now that it has a newer chipset though, and that their will be a greater compatibility with the Prizm now. =)

Yeah I know about the older FX-9860GII from 2009 and USB Power graphics.

Also I wonder if the FX-9750GII and 7400GII that has the SH-4A can be hacked to be turned into a FX-9860GII like we could do with the older 9750GII?

They still make 7400GII's???!!