HP 32SII speed differences

[Dave Britten]

I have both an older 32SII (3116A, 1991, ROM 0) and a newer one (ID108, 2001, ROM 1, the green/purple/silver design). For kicks, I ran the n-queens benchmark on both of them to see if any subtle manufacturing differences would have them finish a couple of seconds apart. I certainly wasn't expecting this stark of a difference.

The older ROM 0 model finished in 00:05:03, much faster than reported for the 32SII on the benchmark page. The newer ROM 1 model finished in 00:05:45, which is just about on the nose with the list of results. When both of them are running the program, the "PRGM" indicator on the older one flashes visibly faster than on the newer one.

Can anybody else reproduce similar results with the two different versions? I'm wondering if it's due to the ROM version, or some other hardware difference. Does battery voltage affect speed to this degree?

[rprosperi]

Did both machines start with the exact same memory conditions? E.g. did one have other programs or variables defined that the other didn't? I know nothing of the internals, but amount of available memory could well affect the performance.

Also, had the program been run more than once on one or the other machines? Are internal jump distances stored after the first run, etc.?

I'd do the exact same setup on both (full h/w reset, enter program, run program 1st time) to be sure of consistent conditions. These are good practices of course, but in this case, I'd guess they make no difference.

[Dave Britten]

(01-30-2020 02:22 AM)rprosperi Wrote:  Did both machines start with the exact same memory conditions? E.g. did one have other programs or variables defined that the other didn't? I know nothing of the internals, but amount of available memory could well affect the performance.

Also, had the program been run more than once on one or the other machines? Are internal jump distances stored after the first run, etc.?

I'd do the exact same setup on both (full h/w reset, enter program, run program 1st time) to be sure of consistent conditions. These are good practices of course, but in this case, I'd guess they make no difference.

Yes, I did a full On-A-F reset on both machines, typed in the program, compared them side by side, verified checksums matched, etc. and started them running at the same time. I used the stopwatch on my 41CX to mark their completion times, so I might be off +/-1 sec on the readings due to reaction time. The first time I ran them, I wasn't timing formally, as I only expected a couple seconds difference between them at most, but the second run where I timed a ~40 sec difference appeared consistent with the first attempt.

I'm very curious to find out if this is typical for the older ROM 0 models. To check your ROM version, run the self test, and at the 32SII-OK display, look at the flag annunciators for the ROM version.

[rprosperi]

I just ran it on a "new" style 32SII => 5:42 [Ver Flag = 1]

Update:

I just ran it on an "old" style 32SII => 5:31 [Ver Flag = 1]

So in my case, the old machine is a bit faster, but not as fast as yours.

We'll need more samples to reach any conclusions.

BTW - Joe Horn found one of the 32SII bugs that caused the Rev 0 -> 1 upgrade. As I recall it had something to do with the new fractions feature introduced with the 32SII, but perhaps he found some other bugs which might explain this change. It seems unlikely to me that a bug fix for the fraction feature would have caused this performance decay, but you never know.

[RMollov]

Not sure which style mine is, but for the record - 05:30, ver. 1

[J-F Garnier]

On these calculators, the CPU clock frequency is set by capacitors (RC or LC circuit) not crystal and the speed can easily vary from +/-5%. It may be that the frequency was reduced in the 32SII production lifetime to improve/secure the production yield and reliability.
Or just you got two extreme samples.

BTW, I long ago emitted the hypothesis that the ROM version was indicated by the flag during self-test, but it's the first time I got a direct confirmation, thanks!

J-F

[Paul Dale]

I wonder what the differences between the ROM versions are.

Pauli

[Pjwum]

Just to add my machines using the summation based benchmark with 1000 counts:

HP-32sII (blue/orange 1997 ROM Ver. 1): 225 s
HP-32sII (green/purple 2001 ROM Ver. 1): 225 s
HP-32s (1990): 183 s

No difference here, only that the more enhanced functionality of the sII comes with a performance penalty. The 32s program had even one more program step because it lacks the nth root key.

I have to say I like the newer version because of the green/purple colour scheme and the silver bezel. Dave, do you also have the infamous sticky key clicks on some keys? I also feel, the LCD in the newer version has a very slightly darker grey and less contrast.

Patrick

[Dave Britten]

(01-30-2020 08:40 AM)J-F Garnier Wrote:  BTW, I long ago emitted the hypothesis that the ROM version was indicated by the flag during self-test, but it's the first time I got a direct confirmation, thanks!

J-F

I would still take it as somewhat hypothetical, but my 0 version has the fraction denominator bug, and the newer 1 version doesn't, so it seems like very strong evidence.

[Dave Britten]

(01-30-2020 11:41 AM)Pjwum Wrote:  I have to say I like the newer version because of the green/purple colour scheme and the silver bezel. Dave, do you also have the infamous sticky key clicks on some keys? I also feel, the LCD in the newer version has a very slightly darker grey and less contrast.

Patrick

On my new version model, the +, -, *, and / keys have a sort of sticky clicking sound when you release them, but the key-press feel and response are fine. Is that a common issue on the later ones? I had just assumed a few drops of something might have gotten spilled in that area at one point (though it looks immaculate). Haven't noticed any obvious difference in screen contrast - I'll have to take a closer look in better light.

[Joe Horn]

(01-30-2020 03:51 AM)rprosperi Wrote:  BTW - Joe Horn found one of the 32SII bugs that caused the Rev 0 -> 1 upgrade. As I recall it had something to do with the new fractions feature introduced with the 32SII, but perhaps he found some other bugs which might explain this change. It seems unlikely to me that a bug fix for the fraction feature would have caused this performance decay, but you never know.

The only bug I found was the one that made flag 9 unreliable for fractions very close to certain boundaries. Their temporary "fix" was to include a paper insert with the manual which said not to use flag 9. Unfortunately they kept including that insert for a while even after fixing that bug. I highly doubt that this bugfix caused any kind of system slowdown, but it *might* cause different timings in programs which use fraction display mode.

[Pjwum]

(01-30-2020 12:43 PM)Dave Britten Wrote:  On my new version model, the +, -, *, and / keys have a sort of sticky clicking sound when you release them, but the key-press feel and response are fine. Is that a common issue on the later ones?

Yes, it is a manufacturing problem since production went to Indonesia. Try to bend the upper case and lower case slightly apart right next to the clicking keys and the click will disappear.
It will of course come back.

[Csaba Tizedes]

(01-30-2020 01:01 PM)Joe Horn Wrote:  The only bug I found

And another from me: HP 32SII SOLVE bug

Maybe find someone interesting.

Csaba

[xerxes]

I'll update the list with 5:03 for ROM0 and the average result of 5:38 for ROM1.
Thank you for testing.

[Dave Britten]

(02-02-2020 05:10 PM)xerxes Wrote:  I'll update the list with 5:03 for ROM0 and the average result of 5:38 for ROM1.
Thank you for testing.

Note that the correlation is still a bit hypothetical until we can get a few more ROM 0 tests, but it's as good a theory as any at this point.

[EdS2]

(01-30-2020 12:38 PM)Dave Britten Wrote:  ... the fraction denominator bug...

Just for reference, see here via here via here.

Quote:

Code:

From: sjth...@cup.portal.com (Stephen J Thomas)
Newsgroups: comp.sys.handhelds
Subject: HP 32SII Fractions
Message-ID: <41829@cup.portal.com>
Date: 30 Apr 91 01:49:50 GMT