new Casio’s fx-CG50

[EdDereDdE]

Here the continuation of that article series:

https://translate.google.com/translate?h...&sandbox=1

this time about the featured and promoted 3D-Capabilities.
It's pretty limited though...

[Luigi Vampa]

I think I will stick to GeoGebra 3D: https://www.youtube.com/watch?v=L8A4FnASNpA

[EdDereDdE]

Maybe the CG-500 is an alternative for you if you like Casio calculators.
http://www.hpmuseum.org/forum/thread-7849.html

[jebem]

(01-31-2017 03:19 PM)jebem Wrote:  

(01-30-2017 08:58 PM)rprosperi Wrote:  According to the Casio Europe website, this is the Specification for memory:
"◦Available RAM/Flash ROM memory: 61 kB/16 MB"
This MUST be typo!? Above was copy/pasted, to be sure I didn't introduce an error.

I believe those are the correct values.

This question from Bob is a very good one, and my reply contains misleading information.
My answer was given based on the CASIO officially available information, but I found it suspicious as well. Then I forgot about it, until a recent CASIO announcement for the GRAPH 90+E equivalent model for French market.

Yet Casio keeps on supplying confusing information (May-2018) on the so called "fx-CG50 Hardware User's Guide":
- Programming capacity: 61,000 bytes (max.)
- Storage memory capacity: fx-CG50: 16MB (max.) fx-CG50: 4.5MB (max.)

I now realize that those values are referred to the end user available usable spaces, not to the actual RAM/ROM configured sizes.
So, the user would have around 61Kbyte of programming area, and a maximum of 16MByte of storage area on a fx-CG50.

Meanwhile I found a few pictures of the internal PCB at TI-Planet and a lot of investigation going on there as well, and that allowed me to find out the exact memory configurations:

- RAM: ESMT M12L64164A-7TI 8MByte SDRAM (versus 2MByte SRAM on CG10/20)
- Storage/ROM: SPANSION S99-50272 16M 16-bit words Flash ROM

A lot of people comments that there is no information available on the SPANSION S99-50272 chip, and some of them comments that it is a 16MByte chip.
Well, it is not. As correctly stated on Casiopeia Forum, this is a 32MByte chip.

And it makes sense to me, for two reasons:
- The Hitachi/Renesas custom made SH7305 processor for Casio fx-CG machines, uses a 16-bit data path.
- The Firmware OS would take a substantial amount of storage, increased if additional applications are installed (the latest 3.11 update fx-CG50 Series OS Ver.3.11 Update.exe file takes 19.7MByte, so I expect the base firmware to take close to 16MByte).
Therefore, a 16M 16-bit words (32MByte) storage is required to accommodate the firmware, additional apps, and leave the referred 16MByte of storage to the suer.

And I was lucky. I found the official Cypress (SPANSION's owners) product notification information containg the equivalence table from the PANSION S99-50272 to the
Cypress S29GL256S90TFI020.

Concerning the processor, there is a consensus on the SH7305, despite CASIO not disclosing information about it.
It seems to be similar to the standard SH7724 (which is documented), according to someone that should know what it is talking about, since he successfully built several overclock tools for those Casio calculators.

[rprosperi]

(05-09-2018 06:04 PM)jebem Wrote:  

(01-31-2017 03:19 PM)jebem Wrote:  I believe those are the correct values.

This question from Bob is a very good one, and my reply contains misleading information.
My answer was given based on the CASIO officially available information, but I found it suspicious as well. Then I forgot about it, until a recent CASIO announcement for the GRAPH 90+E equivalent model for French market.

Yet Casio keeps on supplying confusing information (May-2018) on the so called "fx-CG50 Hardware User's Guide":
- Programming capacity: 61,000 bytes (max.)
- Storage memory capacity: fx-CG50: 16MB (max.) fx-CG50: 4.5MB (max.)

I now realize that those values are referred to the end user available usable spaces, not to the actual RAM/ROM configured sizes.
So, the user would have around 61Kbyte of programming area, and a maximum of 16MByte of storage area on a fx-CG50.

Thanks for the usual good detective work Jose, it's good to know the published values are indeed correct, even in misleading.

It's hard to imagine they could only leave less then 1% of the DRAM space for user programs to run (seems they can be stored in the 16MB Flash area) but without knowing something about the machine, I can't say if that's a problem or not. Are programs copied from flash to ram to execute, or are they XIP and only use RAM for stacks, variable storage, etc.?

[jebem]

(05-09-2018 06:56 PM)rprosperi Wrote:  It's hard to imagine they could only leave less then 1% of the DRAM space for user programs to run (seems they can be stored in the 16MB Flash area) but without knowing something about the machine, I can't say if that's a problem or not. Are programs copied from flash to ram to execute, or are they XIP and only use RAM for stacks, variable storage, etc.?

Good point. It makes no sense at all.
I will have to register myself in one of the Casio/Texas specialized forums to ask about this subject there.

My pure guess is that the 8MByte SDRAM are not used to run the Basic like programs, as the processor itself would have internal very fast SRAM big enough to accommodate the 61KByte user programming area.
However all the other builtin and added apps would need to be loaded to and also use the medium fast 8MByte SDRAM as a scratch workarea.

Again, this is total guessing work, as the more I read on Casiopeia and TI Planet forums the more confused I get (not because I can't read and understand French, because I do. Many folks there have deep understanding of the Casio CPU architecture and don't bother to document the basics such as this subject).

[pier4r]

If I would throw a guess I would say: they compiled the basic/sdk for a certain previous version (say a 9860) and they are not investing in building different versions for different products, as it would raise the maintenance costs.

I know that the 9860G and 9860GII have also this 61kb limit, that I strongly believe being "false" at least on the latest 9860GII.

If most buyers don't care about the 61kb of addressed memory instead of 2/8MB (really a pity) there is no economical incentive to work on it.

Could also be a CPU limitation, that has already the storage addresses full and has little addresses free for the RAM.

For example on the 50g the ram was split in port0 and port1 as the Saturn emulation couldn't address more than 512 Kb of ram.

[Chasfield]

I just bought a Casio fx-CG50 for my birthday treat. I have resuscitated this thread to save starting another.

The Basic interpreter is surprisingly slow, despite the 116 MHz system clock. I ran a couple of small Basic program fragments to benchmark it against my ageing fx-9860G Slim monochrome unit.

1) Iterate 10,000 addition operations:

fx-9860G Slim, 14 seconds
fx-CG50, 21 seconds

2) Plot 2500 pixels in a 50x50 square:

fx-9860G Slim, 107 seconds
fx-CG50, 230 seconds

The results are disappointing with respect to basic number crunching and a deal breaker for my plan to code a Moonlander program in full colour!

I guess the graphics performance reflects the need to update 16 bits per pixel rather than just one. However, I was surprised that the calculation performance was so low.

The calculator has some good points:

The familiar workman-like Casio OS, which I have spent many hours messing around with; the display is great, with a nice large screen font; the unit pops up as a USB mass storage device when it is connected to a PC.

The last point is important because I think you now have to pay a subscription to Casio for their fancy calculator management software.

I quite like the thing for what it is but I am further convinced that the HP Prime is the only graphical calculator that is actually worth buying.

[Dave Britten]

I picked one up just a couple weeks ago too. I've already got a 9860G Slim (thanks to whomever that was that brought it to the HHC 2017 door prize table), and a pair of 9860GIIs that I scored for $17 each on clearance at Wal-Mart. I really like Casio's geometry app and its constraint-based design, and wanted to have that with a hi-res color screen (as well as having some simple color capabilities in the spreadsheet, which is very basic, but still useful).

So far, the Prizm is basically a 9860GII with a better screen, more flash RAM, enhanced connectivity, and a 3D graphing app (which, with the 3.11 update, allows for graphing arbitrary functions of z=f(x,y), or in parametric form expressing x, y, and z in terms of t).

I've always liked the Casios for number crunching, because they're very fast and responsive, have a pretty intuitive UI, and nice "math-print" input. Casio BASIC leaves something to be desired, particularly in terms of terminal I/O, but it gets the job done. I haven't tried the C/C++ SDK yet.

Some of the things I discovered after getting my hands on it:

No connectivity software required. It mounts itself like a flash drive, and you can transfer files, and also import/export programs and data in text/CSV formats. Having to go into the link menu, enter PC link mode, wait for it to initialize, do your transfers, and unmount is a little more cumbersome than just clicking a "Send" button in some connectivity software, but not having to rely on 3rd-party connectivity software is nice. It's a bit like the DM-42 in that respect.

The colored parentheses are awesome. The Prime and TI-84 CE need to steal this feature. When entering an expression in a program or other non-math-print input, the parentheses are all colored based on nesting depth. The outermost parentheses are blue, with red inside that, then green, etc.

The screen is somewhat trans-reflective, and very visible even in bright sunlight.

[Helix]

(08-03-2018 01:47 PM)Chasfield Wrote:  The Basic interpreter is surprisingly slow, despite the 116 MHz system clock. I ran a couple of small Basic program fragments to benchmark it against my ageing fx-9860G Slim monochrome unit.

There is C.BASIC, which is said to be compatible with Casio Basic, but much faster. It is still in alpha version for the CG-50 though.

[Chasfield]

Thank you for flagging up C.BASIC. It looks like it would be a good option when it is properly available on the Casio Prizm platform.

The inefficiency of the stock Casio Basic interpreter running on the fx-cg50 hardware is quite remarkable. Think of dropping a 426 Hemi motor into go-kart and still posting a poorer quarter mile time than an old man on a bicycle! After all, this wee handheld device, with its spartan OS, is packing similar CPU power to a Windows 95 486 rig. Bundled Qbasic, for example, was pretty lively on such hardware.

The core code of the Casio interpreter is 20 plus years old, so I guess that it could be running inside a number of abstraction layers, with bottlenecks that can thwart any amount of CPU power.

[Chasfield]

I downloaded C.Basic for CG Alpha. The comparative benchmarks for the fx-CG50 platform are as follows:

1) Iterate 10,000 addition operations:

Stock Casio Basic, 21 seconds
C.Basic for CG Alpha, way too short to measure!

2) Plot 2500 pixels in a 50x50 square:

Stock Casio Basic, 230 seconds
C.Basic for CG Alpha, ~0.6 seconds!

C.Basic is around 400 times faster at rendering pixels.

Years ago, I compared the then new HP39GII with the Casio fx-9860G (not the Slim version) on this forum using a Mandelbrot set plotting program on a field of 60x100 pixels as a benchmark.

I pulled up this code and ran it on the fx-CG50 using both interpreters. Here is the complete result table:

fx-9860G stock Casio Basic, 643 seconds
HP39GII, 44 seconds - quick
fx-CG50 stock Casio Basic , 608 seconds
fx-CG50 C.Basic for CG Alpha, 30 seconds - quicker still

Thus C.Basic is heading for 20 times as fast as stock Casio Basic with regard to number crunching and on a completely different scale with regard to screen rendering. It also pips the HP39GII. This is a good add-in.

[Jlouis]

As a posted before, the fx-cg10 has similar configuration and I have installed a CAS in it, so the new Prizm will be a pass for me, unless it goes real cheap.

Cheers

JL

[Dave Britten]

(08-04-2018 03:40 PM)Jlouis Wrote:  As a posted before, the fx-cg10 has similar configuration and I have installed a CAS in it, so the new Prizm will be a pass for me, unless it goes real cheap.

Cheers

JL

Is it the Eigenmath port? It has some issues running on the fx-CG50 anyway.

[Jlouis]

(08-04-2018 04:10 PM)Dave Britten Wrote:  Is it the Eigenmath port? It has some issues running on the fx-CG50 anyway.

Yes.

One reason more to pass the fx-cg50.

[ijabbott]

(08-04-2018 04:28 PM)Jlouis Wrote:  

(08-04-2018 04:10 PM)Dave Britten Wrote:  Is it the Eigenmath port? It has some issues running on the fx-CG50 anyway.

Yes.

One reason more to pass the fx-cg50.

I think it's been fixed now - see my post in the CASIO Graph 90+E thread: http://www.hpmuseum.org/forum/thread-107...#pid100868

Admittedly, I haven't tried it yet....

[Jlouis]

(08-04-2018 04:31 PM)ijabbott Wrote:  I think it's been fixed now - see my post in the CASIO Graph 90+E thread: http://www.hpmuseum.org/forum/thread-107...#pid100868
Admittedly, I haven't tried it yet....

Thanks, looks like Parisse made it, but still, I think it's not worth buying this new Prizm at it's price tag, having the old Prizm.

Cheers

[Dave Britten]

(08-04-2018 04:31 PM)ijabbott Wrote:  

(08-04-2018 04:28 PM)Jlouis Wrote:  Yes.

One reason more to pass the fx-cg50.

I think it's been fixed now - see my post in the CASIO Graph 90+E thread: http://www.hpmuseum.org/forum/thread-107...#pid100868

Admittedly, I haven't tried it yet....

Oh nice, thanks for the link. I'll have to try them out. I don't need a CAS often for my work, but better to have it and not need it than vice-versa.

[Chasfield]

More benchmarking:

I reformatted the HP39GII code for the 100x60 pixel field Mandelbrot set so that it would run on the HP Prime.

Code:


EXPORT Mand()
BEGIN
RECT();       
FOR X FROM 1 TO 100 DO 
 FOR Y FROM 1 TO 60 DO
   A:=(X-50)/30; 
   B:=(Y-30)/30;
   C:=A;      
   D:=B; 
   N:=0;    
   S:=1;  
   REPEAT
     N:=N+1;           
     G:=A*A-B*B+C;      
     B:=2*A*B+D;       
     A:=G;              
     If A*A+B*B > 4 THEN 
     S:=0;     
     END;
  UNTIL (N = 20 OR S=0); 
  If S=1 THEN
  PIXON_P(X,Y) 
  END;
 END;
END;
TEXTOUT_P("DONE",0,80);  
FREEZE;  
END;

Here is the updated benchmark table:

HP Prime, 13 seconds
HP39GII, 44 seconds
fx-CG50 C.Basic for CG Alpha, 30 seconds
fx-CG50 stock Casio Basic , 608 seconds
fx-9860G stock Casio Basic, 643 seconds

The Prime's much faster system clock wins the day.

[Helix]

Interesting results.

If you feel adventurous, you can try to overclock the calculator with Ptune3:
http://pm.matrix.jp/ftune2e.html
Ptune and C.basic are made by the same author.