45 replies to this topic

[Xerxes]

Calculator Benchmark

[eew]

How fast is the FX-9860G compared to the CFX9850GC plus?

[Xerxes]

I do not have a CFX-9850G, but in another thread it takes ~3.6 sec for this prog:

1000->X
Lbl 0
Dsz X
Goto 0

The FX-9860GSD takes ~0.4 sec only, that is 9 times faster. But IMO this test is not convincing enough for a benchmark.

My overall impression of the FX-9860GSD: a real improvement to the predecessor calcs.

[eew]

But it's clear the fx-9860 is much faster than the CFX-9850G.

[Xerxes]

You are right.

It depends on the calculators hardware. There are 2 possibilities:

-changing the ceramic resonator or quartz, e.g. FX-603P.

-changing a resistor or capacitor, e.g. FX-602P.

If you have tested the benchmark on a calculator you cannot find in the list, more results are welcome.

[Crazio]

I made a bomb program to my Casio fx-9860G. It counts seconds down to zero and when at zero it prints "BOOM". I got a clock working by making a for->next loop and forcing it to replay it 1850 times:

For 1->Z To 1850
Next

It made through this in one second, so I got a nice working timer. In theory boosting my processor would make the timer faster, but I have no intention to do that.

[Xerxes]

@Sergei:

This test is for non mathematical problems, like tools, games and etc. to see the differences of the calculators and languages.

I have also thought about a mathematical benchmark, but it is difficult to find a accurate way for comparing, however I'm working on it.


@Guest:

Thanks for testing. I have updated the table with your calculator.


@Crazio:

The FX-9860G is a very fast calculator in comparison. In the table it's the fastest calculator not using integers, so the intention to speed it up is not so great. But don't forget, there are many examples that your calculator cannot be fast enough.

[Xerxes]

Thank you for testing. I updated the Casio versions of the benchmark a few hours before you have posted your results because of a mistake. I hope you have tested the correct version, sorry if not.

Interesting is the FX-602P compared to FX-603P. It's a pity to not have a FX-502P in the list now.

I like the FX-5800P, but the speed is a bit disappointing, if I look at the result of the FX-4800P.

I agree that the formula method of array handling to test the Sharp is much to slow.

[Xerxes]

Thank you both for correcting results.

The speed of FX-7300P is really outstanding in comparison to the other 8-bit calculators. I guess the AFX is slower too.

@Mitiay:
The FX-602P can be speeded up to factor 3 by changing a resistor. The FX-602P turbo is as fast as the FX-603P.

[Xerxes]

An interesting bug of this calc. I have checked your rewritten program with the FX-7000GA turbo. Instead of 43.8 it takes 44.0 sec for executing. So I think it is ok to take the FX-8800G to the table with 147 sec. Note that on some calcs Dsz and Isz are only allowed with A~Z. So you have found a accurate way to avoid that bug.

If you have noticed, a structured version is now available for testing languages without goto or if faster, but only the FX-9860G ist faster now. I have tested the FX-5800P (264 vs 227 sec) and TI-85 (150 vs 113 sec) also, but as you can see both are slower. May be the TI-89 or TI-V200 are faster.

[eew]

I see there is no HP-41C in your list, only a HP-41CV and a HP41CX, I (or my dad) own a HP-41C but i can't take over the HP-41CV program, I don't know what y<>x means, and I can't find some command's. Is there somone who can help me whit this?

[Menno]

I see there is no HP-41C in your list, only a HP-41CV and a HP41CX, I (or my dad) own a HP-41C but i can't take over the HP-41CV program, I don't know what y<>x means, and I can't find some command's. Is there somone who can help me whit this?

y<>x means y is not equal to x like y!=x

[eew]

That mean's I have to read a huge manual, to find dat function?

[Xerxes]

It's a bit strange how to type in this command:

XEQ ALPHA 6 SHIFT SIN multiply 3 ALPHA

gives X<>Y?

If you want to test the speed, run the program by pressing

XEQ ALPHA SIGMA+ ALPHA

for about 10 seconds and then interrupt the execution by pressing R/S.

Now start executing again for measurement.

[eew]

SIGMA+???

Why?

[eew]

Ive done somthing wrog, after a vieuw secondes it beepes and output 8. No I can start again.

[Xerxes]

XEQ ALPHA SIGMA+ ALPHA

means execute program at label A.

Why start again? You can edit the program.

[eew]

I'am not realy good white that calculator it's from far before my time.

SIGMA+ doesn't work.

[Xerxes]

Turn the calculator off and after a few seconds on. Now type

SHIFT EEX R/S


If your German is better than English:

Mit SIGMA+ ist nat?rlich nicht die Funktion, sondern das darunter stehende A gemeint. Also nach dr?cken von ALPHA erscheint ein A im Display wenn du SIGMA+ dr?ckst. Falls du nicht weiter kommst, kannst du mich auch in Deutsch fragen oder anmailen.

[Xerxes]

For those are interested in the beginning of the programmable calculators. I have updated the table with some new results:

The first Basic programmable calculator from Casio the FX-702P (1981).

The first programmable LCD calculator from Casio the FX-502P (1979). I guess the FX-502P is the fastest calculator of its time for this benchmark. Well done Casio.

[Mikael]

Hmm... My TI-89 (HW2, 12MHz) seems to have failed miserably. 180 seconds to run the TI-V200 BASIC program. I'll have to try my Casio CFX-9850GB Plus later.

EDIT: Okay, changed how it handles results (approximate -> auto) and the time went down to 108 seconds.

EDIT2: Damn! Can't get the FX-9860GSD program to work on my CFX-9850GB Plus. I get a syntax error at the IfEnd and I can't seem to find anything wrong with the program. What's going on?

[Xerxes]

For a better overview I indented the listing of the FX-9860GSD first, but to avoid errors I have changed it to the normal view how it is displayed on the calculator. May be this is the reason for the syntax error. Before executing it is necessary to create the Matrix A[8,1]. In a few days I have the opportunity to test the CFX-9850GB+, so I will check the problem. Thanks for testing and your advice.

[Mikael]

I looked at the program in my calc again and stepped through it line-by-line and it's identical to what you've written. I still get the syntax error. I just can't find anything that would cause an error like this. I'll continue looking and report if I find the error.

[eew]

I can't get te fx-9860 program on a fx-9860

[Xerxes]

@Mikael:
I looked at the program too, but I cannot find a bug. May be the program itself is not the problem but the Casio-Basic interpreter causes the problem. I will check it and post the solution.

@eew:
Can you explain the problem on the FX-9860G? If you check the program length, it has to be 196 bytes. One reason may be the unequal sign represented by "<>" in the code, but in the program editor its a combination of "/" and "=" not representable in ASCII.

[eew]

No, i don't know exaclty what's wrong. I have not looked at it for af very long time.

[tutti]

Mat A[X,1]-1->Mat A[X,1]
WhileEnd
IfEnd
WhileEnd
LpWhile Y<>1

If you remove the first WhileEnd, it'll run fine, but the last three results will be negative, which I guess isn't right...

[Xerxes]

For sure not, because the values in the matrix A represent the position of the queens on chess board. Have you test it on a FX-9860G or another calculator?

[tutti]

Not the FX-9860G SD, though. Or at least, mine doesn't.

[Xerxes]

@CFX Master:
Thanks for clarification. This is a really big bug, that dosn't allow structured programming without complications.

I guess the FX-9860G has different ROM-Versions, because the benchmark is executable on my FX-9860GSD.

[Xerxes]

Good to know. In this case I will test the CFX-9850GB+ with the unstructured version to avoid this problem.

[tutti]

Huh. Yes, I did do something wrong, and it ran fine when I corrected it =/

Anyway, it took quite a few seconds, but it eventually returned 876. Also, Mat A=[[8,4,1,3,6,2,7,5]]... is that right? :3

EDIT: Nevermind, it is

[Mikael]

Good to know. In this case I will test the CFX-9850GB+ with the unstructured version to avoid this problem.

Well, I took the liberty of rewriting the FX-9860G BASIC program to something that will run correctly on the CFX-9850GB+. It required just a few small changes (I've put them in bold). This program will run slightly slower than what the original program would, due to an extra If clause. Shouldn't affect the times much, though. There might be a smarter way to rewrite the program, but this is what I came up with after a quick check.

The time needed for a CFX-9850GB+ to execute this program is 141 seconds. That's actually not that slow. For comparison, the TI-89 made it in 108s, which is less of a difference than one might expect. Then again, the TI graphs between two and five times faster (at higher resolution), so this test is probably more of a BASIC interpreter test than a hardware test.

The modified program:

0->A~Z
8->R
Do
Isz X
R->Mat A[X,1]
Do
Isz S
X->Y
If Y>1
Then Do (*These two rows replace the original While Y>1.*)
Dsz Y
Mat A[X,1]-Mat A[Y,1]->T
If T=0 Or X-Y=Abs T
Then 0->Y
Mat A[X,1]-1->Mat A[X,1]
While Mat A[X,1]=0
Dsz X
Mat A[X,1]-1->Mat A[X,1]
WhileEnd
IfEnd
LpWhile Y>1
IfEnd (*These two rows replace the original WhileEnd.*)
LpWhile Y<>1
LpWhile X<>R
S