HP 32SII
#1

Hello! Does anybody have first hand experience of this calculator? Is it worth aquiring for day to day use, how much does it have in common with the 42s programming and how much should one be willing to cough up for a good specimen?
Kind regards

#2

Hello Marnus,

Quote:
Is it worth aquiring for day to day use

Depends on what you need. Please compare the features of the 32S -- it's the 32SII without fractions. Compare the two layouts, too. And there is a 35S readily available being -- well -- a 32SII in todays quality for a reasonable price. Please search this forum for a lot of information about the 35S.

Quote:
how much does it have in common with the 42s programming

A lot. Please check this museum.

Quote:
how much should one be willing to cough up for a good specimen?

Up to you d:-)

HTH,

Walter

#3

A 42S is far more capable, but the 32S II is a nice machine to have. If you plan for everyday use, the 33S or 35S may be better... only because they are available now and you can buy a replacement if needed; if you lose a 32S II or other old machines, you will just be sorry. Suggestion: if price is not too high, just buy it.

#4

Quote:
... how much does it have in common with the 42s programming

If you search elsewhere on this site, you will see that the programming paradigms are somewhat different; the 42s is more like the 41C in that regard.

As mentioned above, there are two newer variants of the 32sII, and the one that resembles it most in terms of capabilities is the 33s - relatively ugly but a great machine nonetheless. Since you are looking for a 32sII experience, I recommend it over the 35s for reasons explained elsewhere on this site such lack of rectangular to polar conversions.

They are expensive and can cost you upwards of $200 on EBay, so unless you are adamant about getting one right away, try one of the newer and less expensive models first.

Jeff Kearns

#5

Yes, I have one, and use it occassionally, mostly for fraction work, and as back-up for an employee who has one she uses daily.

My take is this: If you just want the functionality, the 35s is readily available brand new at a consistent price of about $60, has a lot more memory, and a two-line display vs. one-line, and thus would be a better choice. Personally, I could never get used to the ugly 33s.

However, if you, like me, prefer the look and feel of the Pioneers, and don't need the extra memory, then the 32sii is a better choice. As far as price and availability, they are always available on eBay, and can be had for as little as $80 for a used specimen in nice condition, to well over $200 for a an "as new" one, as Jeff points out. Shop smart.

Edited: 17 Aug 2009, 10:27 a.m.

#6

Quote:
Does anybody have first hand experience of this calculator?

Yes. I acquired a 32SII after reading many positive comments here. I was very disappointed.

Personally I wouldn't spend $5 for a 32SII for daily use (I'd just resell it).

I was raised on the 15C and now use the 41CX and 42S daily (emulated on my iPhone). My 15C is on my desk, I still break it out to marvel at its design from time to time. The 32SII was a giant leap backwards. But, that is just my opinion.

If you really like the 32SII feature set, then save yourself some coin and try a 35s, its the follow-on to the 32SII/33s (unless you need it to fit in your pocket).

#7

Quote:
if you lose a 32S II or other old machines, you will just be sorry

But you will have had the fun of using it until you lost it. That fun is priceless.

Bill

#8

The HP42S is far more capable and elegant. I bought an HP32SII to allow me to retire my ten year old HP-15C. I mistakenly assumed that the 32SII would be at least as capable as the much earlier 15C. Bad assumption. Calculations using or producing complex number results are extremely poorly implemented on the 32SII, for just one example. Completely ignoring any collector value, the HP42S is still the best all-round RPN calculator ever made in many people's judgement, 21 years after its introduction. IMHO, the HP-15C is next best, yet far behind, while the 32SII doesn't even rate a mention. The only things the 32SII has better is more speed and larger characters on a less informative LCD.

The later butt-ugly HP33S and the oversized HP35S also have very poor complex number support similar to the HP32SII. What is it about HP's calculator design philosophy that continues this persistent defect in all their calculators except the 15C, 42S, and the RPL models? It would only cost a little firmware development to duplicate 42S capability.

#9

Quote:
Personally I wouldn't spend $5 for a 32SII for daily use (I'd just resell it).

I have a solution. Egan could sell his 32sii to Marnus. :-)

#10

Quote:
What is it about HP's calculator design philosophy that continues this persistent defect in all their calculators except the 15C, 42S, and the RPL models?

It's called marketing a "new" calculator model on the cheap.

#11

Well, thanks to everyone.

I use a 42s and a 50g on a daily basis. (The 42s stays at home, the 50g being the real workhorse) I also do own a 33s but find the keys and layout not very practical. That said it is much faster than the 42s on similar tasks, but there is a lot the 33s can't do that the 42s can. Also, all examples in the 33s is for algebraic mode, which really sucks.

Apart from the above mentioned models, I also own a 48S and a 82240B printer.

I'll just make the guy an offer and see if I can get it fairly reasonable, then take it.

#12

Interesting; I have both a 42s and a 32SII, and the one that gets all the use is the 32SII. It is ideal for the kind of quick calculations that I need to do all the time; it is one of the very few objects that has a permanent place below my PC monitor. It has a clear, legible display, and direct access to most of the functions you are likely to use.

I have a 42s, and from an engineering point of view, I admire it, but I hardly ever use it. The display is too difficult to read in the sometimes dim light I use to work at the PC, and too much of its functionality is hidden in long menus -- it is no fun at all to have to go fish in the function catalog even the %CH (percent change) function, which I use relatively often. Yes, I could program it as one of the custom soft-keys, but then I lose access to top-level key functions...

I could never warm to the 35s. The three chief drawbacks, in my opinion, are: the STO key is shifted; when you type a long number, no "," separators are inserted while you type (only after you enter it), making it difficult to enter them correctly; and it has some bugs that decrease my confidence.

So in the end I always go back to the 32SII. It is the only calculator that, when I use it, I do not wish I were using something else.

The 42s used to be the perfect design point (or close to it) in an era when access to PCs was more difficult. Nowadays, when more complex things tend to be done on PCs, I think the 32SII is the best pocket calculator there is. Simple, fast, direct, well-engineered.

Luca

#13

The following link will get you the 32sii manual so that you have all examples in RPN format. That should increase your enjoyment of the butt-ugly 33s (which I kinda like...)

http://h10025.www1.hp.com/ewfrf/wc/manualCategory?cc=us&dlc=en&product=58450&lc=en&jumpid=reg_R1002_USEN

Jeff Kearns

#14

Quote:
The later butt-ugly HP33S and the oversized HP35S also have very poor complex number support similar to the HP32SII

Actually, the 35s has much better complex number support. Having the "i" key (with the shifted angle) was in my mind a great leap forward. The problem is that the functions which will accept a complex argument are limited, and of course the lack of either this menu:

or these these functions on the keyboard:


...

#15

I consider that real support of complex numbers only happens when complex numbers are accepted and results generated whenever they naturally occur. Examples:

sqrt(-3),
-5.1^-2.7,
i^i,
arcsin(1.5),
ln(-3.1),
cos(i4.2),
manipulation of matrices with complex elements,
etc., etc.

AFAIK, the HP35s does NOT naturally process most of these types of operations. The HP42S and the RPL models handle them with ease. The HP-15C does too, with a little bit more effort.

There is no valid excuse for calculators such as the HP35S (and anything coming out later) not having this capability which had been beautifully implemented in calculators introduced more than two decades earlier!

#16

Since you're accustomed to the HP42S, you'll likely find the HP32SII rather disappointing. It is several years later than the 42S, so it is faster, but only by about 50 to 100 percent depending on the operations performed. The display's characters are larger and clearer, but anything otherwise would be surprising since far less information is shown. Like the earlier HP-15C, it has no sound-making (beep) function, and it has no infrared printer output. It's quite a come down from its more accomplished older brother.

I wish they had produced an HP42SII and dropped the HP32S fifteen years ago, instead of the reverse.

#17

I agree that calculators that support complex number should do so all the way. However, I don't think that all scientific calculators need to support complex numbers -- certainly financial/scientific calculators like the 20b don't need them. The 32SII does a really poor job at complex number support and set a precedent for the 33S and 35S. Even though I don't often need complex numbers I still find some of what they do useful even for non-complex number manipulation. In particular, you can use the sequence <0> <ENTER> <COMPLEX +> to duplicate the last 2 entries on the stack.

Although I have and use virtually all the HP calculator models I *love* the 32SII. It's the one I use daily and keep one handy everywhere (they used to be cheap on ebay). Programming on the 32SII is a pleasure, you have nearly instant access to all functions, memory isn't too big to keep what you're doing in your head, there are only a handful of flags to know and the display is large and easy to see in bad lighting.

#18

I have to agree with Katie - I love the 32sII. It is really well-balanced calc with very little hidden, good feature set, great keyboard and ergonomics and despite the relatively small memory size, I actually like that because it encourages efficient programming that explores all the techniques and commands.

One area that people rarely mention when comparing the 32sII with the 35s is base conversion and the two calcs are very different here. With the 32sII, you set it to the base you want and carry on working with no special procedures. With the 35s, despite setting the base, you still need to specify each number to be in that base which requires totally redundant but enforced extra keystrokes. This feature drives me nuts and I cannot understand why it was implemented.

On top of that, the 35s has a poor keyboard, bad ergonomics, an ugly display and even though it has a lot lot more memory, you still only have 26 labels and have to use line addressing - hardly encouraging for writing large programs.

The 35s has a few extra useful commands (binary logic for example) but given a choice between this and the 32sII, I would pick the 32 without any hesitation. It really is a great, fun little calc.

Mark

#19

Quote:
...it is faster, but only by about 50 to 100 percent depending on the operations performed.

My numbers are wrong. I've just run some extended looping comparisons between the 32SII and the 42S. These show a 32SII speed advantage of only about 35 percent.

#20

I'll have to agree with the two above posters; the 32SII is a pleasure to use, though after having used the 33s and 35s, the memory space allotted for programming might just be a tad small on the 32SII. Otherwise, it was the first calculator I reached for after my 34C went down, until recently, because I am... afraid... of wearing it down, too, so now I prefer the 33s (over the 35s).

#21

I own a 32SII. I would emphasize one point that others have already made: the memory is really very small indeed. Don't get the idea that you can store a library of equations there, plus a few useful programs: you can't. Apart from this, I find it to be an excellent machine for day-to-day use. However, I prefer the HP-33S because of the greater memory and the two-line display.

Nigel

#22

Quote:
However, I don't think that all scientific calculators need to support complex numbers -- certainly financial/scientific calculators like the 20b don't need them. The 32SII does a really poor job at complex number support and set a precedent for the 33S and 35S.

I also agree, but if you are going to do it, then do it right, or at least as good as before, especially if you have a history of doing so already.
#23

Many have stated that the 35s is slower than the 33s. Have you compared the speed of the 35s to the 42S?

#24

Well, Marnus, you certainly have a full variety of opinions to choose from!

#25

I don't have an HP35S. The firmware quirks that have been pointed out over the past two years have discouraged me, plus it is unfortunately (and unnecessarily) a little too large for a pocket calculator.

I use the very simple, well-known, and often criticized Savage Benchmark, very slightly modified, to get an initial estimate of speeds. It is a simple 2500 iteration loop equivalent of:

A = TAN(ATAN(e^LN(SQRT(A)^2)))+1

How close the final result is to 2500 gives a feel for the precision and effects of rounding errors, and how long it takes can be used for relative speed comparisons.

For the HP42S (Runs in 605 seconds):

01 LBL "SB"

02 2500
03 RAD
04 0
05 LBL 01
06 1
07 +
08 SQRT
09 X^2
10 LN
11 E^X
12 ATAN
13 TAN
14 DSE ST Y
15 GTO 01
16 BEEP
17 DEG
18 END

For the HP32SII (Runs in 460 seconds):

S01 LBL S
S02 2500
S03 STO A
S04 RAD
S05 0
A01 LBL A
A02 1
A03 +
A04 SQRT
A05 x^2
A06 LN
A07 e^x
A08 ATAN
A09 TAN
A10 DSE A
A11 GTO A
A12 DEG
A13 RTN

If someone with an HP35S programs the equivalent and reports the run time, we'd have a better feel for relative speeds.

FYI, the HP50g runs the equivalent program in USER RPL in about 64 seconds, while the HP-15C runs it in about 6100 seconds. The HP-34C requires about 7900 seconds, and the HP-41CX full-nut requires about 3300 seconds.


Edited: 17 Aug 2009, 10:03 p.m. after one or more responses were posted

#26

Hi Mike. I timed 635 seconds on my 35s.


Regards,


John

#27

I did an initial run on the 35s and it took just short of 10 minutes! I will run it again later for a more accurate time but I would say between 550 and 575 seconds. Rather slow...

Jeff Kearns

#28

Could someone run this benchmark for me please on a regular 32s? I have been puzzled by certain results I've had in the past with a 32s compared to 42s and 32sII so an independent result on this particular benchmark would be greatly appreciated.

Thanks :)

Mark

#29

Quote:
FYI, the HP50g runs the equivalent program in USER RPL in about 64 seconds, while the HP-15C runs it in about 6100 seconds.

Do you have the listing of that UserRPL program please?

Thanks :)

Mark

#30

Quote:
(Runs in 605 seconds on the HP42S)

... and only 7 seconds on the iPhone's 42s!


Edited: 19 Aug 2009, 6:58 p.m. after one or more responses were posted

#31

Quote:
Many have stated that the 35s is slower than the 33s.
I entered the 32SII program above on an early-production 33S, and got a time of 7:55, or 475 seconds.

Based on the times listed above, the 33S appears to be slightly slower than the older 32SII, but significantly faster than the younger 35S.

Edited: 17 Aug 2009, 7:13 p.m.

#32

In my opinion the 32sii is minimalism at it's finest. Well really the 32S is but the 32sii is a close second. Especially If you can get one with a brown bezel (is that the right terminology), that would be best. It does everything that anybody living in a Newtonian world would ever need. It may be short on memory for programming and equations but as an everyday utilitarian machine it can't be beat. I own a 33s and 35s and I just don't feel the love for these two.

IMHO

#33

My UserRPL Savage benchmark program is below. I store the following program as SB:

<< RAD TICKS 0. 1. 2500. START 1. + SQRT SQ LN EXP ATAN TAN NEXT SWAP TICKS SWAP - B->R 8192. / DEG 800. 2. BEEP >>

The program ends by displaying in stack level 2 the final value after 2500 iterations, and displaying in stack level 1 the time (seconds) that it took to complete the looping. This is followed by an 800 Hz tone.

Typical results for an HP50g:

2: 2499.99946106

1: 64.5230712891

In the listing above, SQRT is the square root symbol, and B->R is the binary-to-real conversion function. It is important to have the decimal points on the constants within this program, otherwise they will be treated as binary integers (bints). This this can almost double execution time!


Edited: 18 Aug 2009, 4:52 p.m.

#34

HP32S Sn: 3038Sxxxxx

Execution time = 354s

Christoph

#35

Quote:


... and only 7 seconds on the iPhone's 42S!


PocketPC ARM PXA310/640MHz

Emu42PPC v1.11beta2 08/18/09

HP32SII emulation = 15s

HP42S emulation = 28s

Christoph

#36

Quote:
I consider that real support of complex numbers only happens when complex numbers are accepted and results generated whenever they naturally occur. Examples:

sqrt(-3), -5.1^-2.7, i^i, arcsin(1.5), ln(-3.1), cos(i4.2), manipulation of matrices with complex elements, etc., etc.

AFAIK, the HP35s does NOT naturally process most of these types of operations. The HP42S and the RPL models handle them with ease. The HP-15C does too, with a little bit more effort.


I agree, the 35s should do as you say. But I would say that accepting complex inputs for all valid functions is only part of the story. The other part is entry and display of complex values. In my opinion, the 35s' method of entry is far superior to both the 15C and 42S, and its display is almost as good as the 42s and again superior to the 15C. Of RPN machines, the 35s is the only one in which four complex values can be naturally and easily entered onto the stack. While it certainly has shorcomings, I think the 35s deserves some credit for the things it got right.

Quote:
There is no valid excuse for calculators such as the HP35S (and anything coming out later) not having this capability which had been beautifully implemented in calculators introduced more than two decades earlier!

It is not my intention to make excuses for HP. I will note that the 35s was an extension of the 32S/32Sii/33s line. Aside from the improvements in entry and display, it also offered a full 4-value stack depth, and eliminated the need to tell the calculator you want to treat x and y (and maybe z and t) as components of a complex value by pressing "COMPLEX" before executing a function. So, again in my opinion, I think the 35s succeeded in improving on the complex number capabilities of the 32S/32Sii/33s line.

#37

Quote:
One area that people rarely mention when comparing the 32sII with the 35s is base conversion and the two calcs are very different here. With the 32sII, you set it to the base you want and carry on working with no special procedures. With the 35s, despite setting the base, you still need to specify each number to be in that base which requires totally redundant but enforced extra keystrokes. This feature drives me nuts and I cannot understand why it was implemented.

Although I never use the base converions, I agree that the method implemented on the 35s is baffling.

Quote:
On top of that, the 35s has a poor keyboard, bad ergonomics, an ugly display and even though it has a lot lot more memory, you still only have 26 labels and have to use line addressing - hardly encouraging for writing large programs.

I like the feel and arrangement of the 35s keyboard. I'd like it to be a bit smaller, but that is my only ergonomic complaint, unless shifted STO is an ergonomic complaint. As far as programming, I think it would be fair to say that 26 labels and 384 bytes (not sure of exact figure) of memory is also hardly encouraging for writing large programs.

Quote:
The 35s has a few extra useful commands (binary logic for example) but given a choice between this and the 32sII, I would pick the 32 without any hesitation. It really is a great, fun little calc.

Don't get me wrong, I like the 32s/sii but given the choice (a choice that I have), I pick up the 35s, primarily for the complex capabilites discussed above (or below, depending on how you view your threads.)
#38

Just for grins:

mymac> cat savage.c
#include <math.h>

int main(int argc, char *argv)
{
float a =0.0;
int i;

for(i=0; i<=2500; i++)
a=tan(atan(exp(log(pow(sqrt(a),2))))) + 1;
}
mymac> gcc -o savage savage.c
mymac> time ./savage

real 0m0.002s
user 0m0.001s
sys 0m0.001s

2.53 GHz Intel Core 2 Duo. Mac OS X 10.5.8

Regards,
Howard

#39

The loop should terminate <2500 not <=2500.

You also need to output the result or the compiler is free to optimise the loop out of existence.

Even then, the compiler would be allowed to computer the result of the loop and just print that....

- Pauli

#40

I agree, but prefer to leave the 42S, and even the 32S II at home, and have a 33S at my lab desk, where it can be lost, dropped, etc. So I can use RPN at work, and use the older models at home.

That said, I remember in my university days, some students (me included) carried their HP 25 with them all the time. Such calculators showed some marks of heavy use, they were certainly "not mint" by today standards; but we knew how to apply them to many problems, and they helped us learn a lot more than the then current curriculum specified. There were other students, who also had an HP 25, but never have it on their hands or hanging from their belts. These calculators looked almost new (something a collector would appreciate nowadays), but their owners only bring them to class on final exam events. Needless to say, they barely know how to turn them on, so they benefited very little from the experience of using the calculator almost every day. Of course, I'm more in the side of "users", not that much in the "keepers", but I do reflect on the chance of never being able to replace a lost or broken "oldie".

Just my 0.02$

#41

Quote:
In my opinion the 32sii is minimalism at it's finest. Well really the 32S is

Agree on this. That was the reason why I recommended to Marnus to compare the keyboards of the 32S and the 32SII in my first response above.
#42

That's within a couple of seconds of what I got also - the second time while paying better attention... I missed a full minute on my watch the first time-round. JeffK

Edited: 18 Aug 2009, 4:46 a.m.

#43

Well-well, you guys certainly have been busy. I noted last night that with program execution the 33s is substantially faster when executing the same / similar programs, BUT NOT AT ALL much faster when solving equations. I used Colebrook-White/Darcy-Weissbach equation for flow in pipes and also the TVM implimentations (rewriting the 33s textbook equation so that the I%YR is entered as is, without the need to /12 beforehand).

Anyway, I've made an offer for the 32SII, thought I could always obtain a 35S later should I not like the 32SII, but the other way round would be more difficult.

Thanks for all the input

Edited: 18 Aug 2009, 10:42 a.m. after one or more responses were posted

#44

This Message was deleted. This empty message preserves the threading when a post with followup(s) is deleted. If all followups have been removed, the original poster may delete this post again to make this placeholder disappear.

#45

I don't believe the compiler will do any optimization unless you use -O or one of its variants (-O2, -O3), or request specific optimizations of it (e.g., -funroll-loops).

#46

Hi Norris. I timed 455 seconds on my early production 33s. I'm not sure why there is a 20 second between our times however I have noticed that this particular 33s slows down as the batteries get weaker. Could this be the case? Does your 33s have fresh batteries?

Regards,

John

#47

Hello Jeff (and others, too!)

Do we have a document on this site where the "programming paradigms" are listed and compared? Or maybe I didn't look in the right place!

Thanks,

Joel Setton

#48

32S S/N: 3038Sxxxx: 401s

20S S/N: ID747xxxxx; algebraic, keystroke programming, no DSE/ISG: 350s (I can post the code if requested).

#49

I wonder what the new fast 12C would attain?

#50

It doesn't have the trigonometric functions that are part of the standard Savage benchmark.

#51

http://www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/archv015.cgi?read=87806

The above link will provide some info on this. The best person to comment on this is Karl Schneider.

Jeff

#52

Hi Joel,

Here at the museum is a benchmark article, by Xerxes. It features a long list of calculators (incl. non-HP calcs), their benchmark times, and source code of the various N-Queens programs used for benchmarking. By the source you can contrast or identify programming patterns across various models.

http://www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/articles.cgi?read=700

I used FIND (and NEXT FIND) and "42s" and "32sii" to get around the article quickly.

Edited: 18 Aug 2009, 11:06 a.m.

#53

http://www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/articles.cgi?read=556

Also look at this article by Karl. It is very instructive.

Jeff

#54

Hi, Pauli

Taking a look at the assembler code produced by gcc, it's clear the loop is structured as coded.


mymac> gcc -S -o savage.s savage.c 

and a portion of the output:

        jmp     L2
L3:
cvtss2sd -16(%ebp), %xmm0
sqrtsd %xmm0, %xmm0
mulsd %xmm0, %xmm0
movsd %xmm0, (%esp)
.
.
(calls to libm for the rest of the computation)
L2:
cmpl $2500, -12(%ebp)
jle L3
.
.

Regards,
Howard

#55

Oops, I forgot, it's a financial calc :(

#56

My test times (and the numeric result) are as follows:

HP35s:    629 seconds, 2499.99939863
HP33s: 481 seconds, 2499.99939863
HP-32Sii: 419 seconds, 2499.99946106
HP-32S: 355 seconds, 2499.99946106
HP-42S: 597 seconds, 2499.99946106

Note - for fun and to keep everything in the hp family, all timings were done using an hp-45 with crystal installed.


Edited: 18 Aug 2009, 11:51 a.m.

#57

Quote:
Does your 33s have fresh batteries?
Nope, they apparently haven't been changed since January 2006. That likely accounts for the time difference.

When I change batteries, I write the date on a piece of tape stuck to the back of the battery cover.

I took the PE exam with the 33S in early 2006, and apparently took the precaution of replacing the batteries before the exam. I have not used the 33S extensively since then. It got me through the PE exam, but I've never grown particularly attached to it. For the most part, I only pull it out of the drawer if the 11C is buried somewhere under the papers on my desktop

Edited: 18 Aug 2009, 12:44 p.m.

#58

My results with Free42/42s are similar to Reth at 7 seconds, with a result of:

2500 . 00000 00000 00001 704613

I compiled up a binary math version of 42s, and it runs in less then a second (hard to tell exactly how much less), and it gets a result of:

2500 . 00000 00015 816

#59

And in an albeit somewhat silly speed test, wouldn't one routinely turn the optimiser on? I know I would.

- Pauli

#60

If you remove the trig functions, I get a time of 260 seconds (result=2499.99999938) on the 32Sii and a time of 31 seconds (result=2500.000035) on a 12C+ (with the latest firmware). This isn't a totally fair comparison though since the 12c+ doesn't have DSE and x^2 and therefore requires more commands. Here's the code I used:

01 2
02 5
03 0
04 0
05 STO 0
06 0
07 ENTER
08 1
09 +
10 SQRT
11 ENTER
12 x
13 LN
14 e^x
15 RCL 0
16 1
17 -
18 STO 0
19 x=0
20 GTO 23
21 Rv
22 GTO 08
23 Rv
24 GTO 00

-Katie

Edited: 18 Aug 2009, 9:48 p.m.

#61

Quote:
This isn't a totally fair comparison though...


Although it does seem to imply that that if HP were to release a 15C+ or (heaven forbid) a 42S+, it would beat the socks off the 32's.

Interesting though that my 20S does not do too badly in the Savage test (350sec) even with no stack and no DSE.
#62

Quote:
HP32S Sn: 3038Sxxxxx

Execution time = 354s

Christoph


Thank you Christoph. On mine, I get 391.5 seconds with result being 2499.99946106.

So, I wonder why so many of us are getting different results for the same models? Could it be something simple like battery condition? Or could it be down to the use of memory in the calcs with other variables present, etc.? I cleared all programs out before running the benchmark but I didn't clear the variables.

Mark

#63

Bart - your 32S time is 10 seconds slower than mine but there are two other results posted that are in the ~350 second region which is a lot faster relatively than our timings.

How is it that what are meant to be identical calcs can produce such a wide variance of performance?

Mark

#64

Did another attempt but this time I did a "Clear All" before typing in the program just in case memory usage could have a bearing on the result. I reasoned that Clear All would be one step away from resetting the calc.

The result.

As near as damnit, exactly the same.

So, some of us have fast 32S and some of us have slow 32S. Is it temperature related? Could it be simple manufacturing variances? Could this mean that we can't get accurate benchmark results for any calc unless we take a wide survey from many machines?

Mark

#65

Quote:
How is it that what are meant to be identical calcs can produce such a wide variance of performance?

Those of you who ever saw 2 *identical* items, please stand up d8) Identity (?) is the consequence of bad sight d;)
#66

Quote:

So, I wonder why so many of us are getting different results for the same models? Could it be something simple like battery condition? Or could it be down to the use of memory in the calcs with other variables present, etc.? I cleared all programs out before running the benchmark but I didn't clear the variables.


You question is correct. Especially the program memory usage will have an influence. IFAIK the program dispatcher can only jump forewards. At the end of program the interpreter jumps to program top and continue searching.

My 32S was totally empty before doing the benchmark.

A 2nd reason is the frequency multiplier. IFAIK all Saturn based calculators use a 32768Hz crystal oscillator. A frequency multipier generates the CPU clock out of the crytal frequency. Several physical ambient factors have an influence to the frequency multipier, mostly the ambient temperature.

That's the reason why every Lewis chip based calculator measures the CPU speed before sending data over the Redeye interface. On these calculators the complete IR pulse timing is generated by the CPU.

But how the CPU can measure there own speed? All newer Saturn based calculators have a 8192Hz timer which is directly derived from the 32768Hz crystal (dividing by 4) . When you now compare the CPU the with 8192Hz timer speed you get factor how much faster the CPU based on the 8192Hz signal is.

I hope this helps to solve some of the mysteries about speed differences.

Christoph

#67

Quote:
Those of you who ever saw 2 *identical* items, please stand up d8) Identity (?) is the consequence of bad sight d;)

-It's all in the eye of the beer-holder.


However: my 32S was made in the same place and the same week of the same year as Christoph's.

In the first runs I had 136B free mem. I have cleared all and retyped the prog. Now I get 369 seconds, 15s (~4%) slower than Christoph.

Another option is to try new batteries (I can't remember when last they were replaced).
#68

Quote:


Those of you who ever saw 2 *identical* items, please stand up d8) Identity (?) is the consequence of bad sight d;)


I'll need to check this but I think my Voyager calc timings have been pretty consistent with those published elsewhere. Different architecture than the Saturn?

Mark

#69

Thanks for the details Christoph. My instinct tells me that the warmer the calc, the faster it should run and today when I did the timing tests, it is a very hot day here in the UK. But does high temp actually equate to slower speed?

Regarding memory usage, when I cleared the calc before doing another test, that test was about 1.5 seconds slower than before when I left the vars as-is and only cleared program memory. I'll run some more tests for consistency.

Mark

#70

HP32S Sn: 2820Axxxxx 
Execution time = 387s

HP32S Sn: 2901Axxxxx (50th Anniversary)
Execution time = 370s

HP32SII Sn: ID7440xxxx
Execution time = 458s

HP32SII Sn: ID0100xxxx
Execution time = 417s
Execution time = 426s (after replacing weak batteries with fresh)

Did a Clear All before testing. Same numerical result.

I love the 42S for its power, memory, and 2-line display.

But for everyday use and simple programming and equation solving tasks, the 32S is simple and uncluttered. The display is very easy to read. And clearly it's a bit faster for complex tasks. It's also not too hard to obtain or replace so far.

The 32SII is obviously close, but the keyboard seems too busy after using the 32S. Just my opinion.

#71

If you modify the benchmark for the 32S very slightly by making it a summation of the returned values, you can re-write it to run in the Solver. This enables timings on the 17B, 27S, 19B, etc..

Timing for a 27S at ambient temp 25C is 274 seconds. The fastest of the Pioneers!!

Mark

#72

For those who cannot accept that the HP42S is slower than the HP32SII. Here's a small trick. The HP42S has a Lewis chip inside, and the Lewis chip has a RATE register controlling the CPU speed.

This register at 40300H can be modified with the internal memory scanner from 7H to FH for maximum speed. But don't exit the memory scanner with the regular <ON><SK3>. This would reset the RATE register back to 7H. Best, working on all High-End-Pioneers, is to execute address 00000H. This executes a warmstart but leave the RATE register and the memory content intact.

Pressing the <ON> key will also reset the RATE register. So start the benchmark just after the warmstart and enjoy the fastest original HP42S you ever seen. ;-)

Christoph

#73

... and the new just released 42s Version 2.2 (Free42 1.4.52) returns 2,500 exact for 6 second (before the BEEP)! Brilliant!
actually subtracting 2,500 from it gives 1.704613e-15

Edited: 19 Aug 2009, 10:08 p.m.

#74

Quote:
The HP42S has a Lewis chip inside, and the Lewis chip has a RATE register controlling the CPU speed.

This register at 40300H can be modified with the internal memory scanner from 7H to FH for maximum speed.


Gene Wright posted details about this procedure in 1997:

http://www.rskey.org/gene/hpgene/hp42fast.htm

http://www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/archv009.cgi?read=26022

I suppose that 7 hexadecimal is the middle speed; F hexadecimal is the maximum speed. Values lower than 7 would slow it down.

Only the original versions of the HP-42S (reflective, non-recessed display window) allow the speed setting to be changed by program.

-- KS

#75

Hi, Katie --

Quote:
The 32SII does a really poor job at complex number support and set a precedent for the 33S and 35S.

Although cumbersome for interactive usage, the capabilities are still quite useful for programming:

http://www.hpmuseum.org/software/32acxfer.htm

The built-in complex-number capabilities of the HP-32SII were the same as those of the HP-32S, which apparently used the HP-41 Math Pac as a template.

Quote:
... the 32SII (is) the one I use daily and keep one handy everywhere ... Programming on the 32SII is a pleasure, you have nearly instant access to all functions, ... and the display is large and easy to see in bad lighting.

Yup, I use it most routinely for those reasons. It's also about 12 times as fast as an HP-15C, and better-suited to maintaining small programs in permanent residence (easier to follow and organize).

-- KS

#76

Interesting that fresh batteries actually slow down the calculator. I remember a similar observation on a battery operated bedside clock I had once, it gained time when the battery started getting weak.
So if you want a fast calculator, use weak batteries with the risk that it may die just before it completes the calculation :).

I too prefer the less cluttered 32S. I like menus. When I got my 28C (later replaced by a 28S), it seemed the natural progression for calculators to add more functions to without umpteen shift buttons (eg the fictitious 67CX, although the interest in this machine seems to indicate my views may be the minority).

#77

TNX for remembering Gene Wright's article.

The CPU speed is calculated by ((RATE + 1) * 524288 Hz) / 4.

So the nominal speed is ((7 + 1) * 524288 Hz) / 4 = 1048576 Hz

You can verify this in the single step self test mode <ON><SK4> with <SK1>.

Christoph

#78

KS-

Very nice program and an excellent fit on the 32Sii! Do you work in electrical power engineering?

-Katie

#79

Quote:
Very nice program and an excellent fit on the 32Sii! Do you work in electrical power engineering?

Katie --

Yes; at least several of us Forum participants do.

You may have noticed that several versions of my program have been uploaded, each tailored for the HP-32S, HP-32SII, HP-33s (RPN and ALG). I've also submitted versions for the HP-35s, but these were just crude, direct portings of the HP-33s programs.

The posted HP-32SII program omits a small refinement I made -- lines P29 through P33 should read as follows:

P29  STO P
P30 x<>y
P31 STO Q
P32 VIEW P Show P injected into line(s)
P33 VIEW Q Show Q injected into line(s)


A power-system equation I keep in the meager 384-byte memory is the following:

M=K*A*SQRT(3)/100/10     3-phase (MVA) = (kV)*(Amp)*sqrt(3)/1000

Expressing "/1000" as "/100/10" saves five bytes, because non-negative integers lower than 255 on the HP-32SII require only 1.5 bytes of storage for programs or equations.

Or, one can save a byte and gain some speed by entering a (trivially-)calculated constant in the equation:

M=K*A*1.73205080757E-3     

Certainly, direct-solution programs can also be stored for this purpose (and that is what must be done on the HP-32S), but, well, there's that storage-space issue...

-- KS


Edited: 24 Aug 2009, 11:27 p.m.

#80

Hi all,

Or... if you're concerned about 42's speed, replace the internal 32.768KHz Xtal with a 65.536KHz one and you won't need to worry about the [ON] key ever more... (unless you still want to get it doubled again (4x) by altering H'40300 register to H'0F... regrettably, it won't go beyond H'09 - a 2.25x from original speed - figures above this will render an unreadable display.)

BTW, once opened you may also consider a memory upgrade to 32KB... ;-))

I've made these mods for several users as well as on my own 42's and they really work!

However, in my experience, both models of 42's can be speeded-up by changing reg H'40300.

Cheers.

Diego.

Edited: 1 Sept 2009, 11:34 a.m.



Possibly Related Threads…
Thread Author Replies Views Last Post
  HP 32sii Fault Carlos Estinos 9 2,779 09-26-2013, 01:06 PM
Last Post: Csaba Tizedes (Hungary)
  HP 32sII with faulty keypad aurelio 6 2,518 09-21-2012, 05:08 PM
Last Post: aurelio
  HP 32sII Integration Error of Standard Normal Curve Anthony (USA) 4 1,601 03-14-2012, 03:18 AM
Last Post: Nick_S
  Find the hidden path game for HP-32sii Don Shepherd 0 1,014 09-27-2011, 09:46 PM
Last Post: Don Shepherd
  HP-32Sii jweisskohl 38 8,379 06-21-2010, 10:30 AM
Last Post: Katie Wasserman
  HP 32SII keyboard - dead column of keys fixed Jeff Johnson 2 1,180 03-27-2010, 02:34 AM
Last Post: geoff quickfall
  How to open my HP-32SII Sören Nilsson 8 2,535 05-20-2008, 09:45 PM
Last Post: bill platt
  How to open my HP-32SII Sören Nilsson 2 1,176 04-07-2008, 07:11 AM
Last Post: Lyuka
  HP 32SII "Clicking" Santeh 1 929 03-21-2008, 06:52 PM
Last Post: Thomas Radtke
  HP 32Sii Troubleshooting - Does not turn ON recnadniar 1 951 08-28-2007, 07:54 AM
Last Post: Chris Foley

Forum Jump: