HP Forum Archive 14

[ Return to Index | Top of Index ]

HP 33S CPU speed Message #1 Posted by Ed Look on 17 Sept 2004, 10:28 p.m. I know that the 49G+ is powered at its core by a 75 MHz ARM CPU. If I'm not mistaken, the 48GII has an ARM running at what, 25 MHz? But what is the 33S' processor and how fast does it run in terms of clock speed? (Is it still slower than an 80286?... j/k ;) )

Re: HP 33S CPU speed Message #2 Posted by John Limpert on 17 Sept 2004, 11:41 p.m., in response to message #1 by Ed Look It uses a Sunplus SPLB31A, which is an enhanced 6502 core with a LCD controller, integrated RAM/ROM and other stuff. Not sure what the clock rate is. Fast enough for a calculator.

Re: HP 33S CPU speed Message #3 Posted by Ed Look on 18 Sept 2004, 12:46 a.m., in response to message #2 by John Limpert Hey John, thanks! The only reason why I ask is that I took my HP 33S and programmed in a (silly?) infinite series that converges toward a value, and I made the cutoff to be if a term is less than 1 E -6. Man! It took over six hours! (Well, that isn't precise; I let it run overnight and checked it when I got up.) I was just wondering after that, "How fast(/slow) is that thing, anyway??" But, if I made the cutoff to be 1 E -3, it was over in some minutes. At 1 E -4, it was many more minutes; over a half hour at least if 1 E -5. So, as I said, I was just wondering.

Re: HP 33S CPU speed Message #4 Posted by Norris on 18 Sept 2004, 3:50 p.m., in response to message #2 by John Limpert There are some specs for the SPLB31A at: http://w3.sunplus.com.tw/products/lcdcontroller/splb.asp Don't actually understand the specs myself, but they indicate a "Max Speed" of 5 MHz, if that helps

Re: HP 33S CPU speed Message #5 Posted by Ed Look on 18 Sept 2004, 4:11 p.m., in response to message #4 by Norris Norris, thanks for the link. So, it's top speed is 5 MHz. Okay. It's like the 80286 or 80486, then... sort of; I know it's comparing apples to oranges. Now at least I know why the calculation too a little while on my 33S. (Is this faster or just as fast as the 32S and 32SII?)

Re: HP 33S CPU speed Message #6 Posted by martin cohen on 18 Sept 2004, 8:58 p.m., in response to message #5 by Ed Look That's slow! The original 6502 in the Apple ii and Commodore PET was 1 MHZ - in 1977!

Re: HP 33S CPU speed Message #7 Posted by Ed Look on 19 Sept 2004, 1:00 a.m., in response to message #6 by martin cohen LOL! ... and I used to look down on the Commodores because my baby brother had one, and he only played games or programmed images with it. I guess I considered it a toy, but 1 MHz! That's a serious number for any machine in those days. I'm amazed! Does anyone know the speed of the processor (I don't think it's a Saturn, though) of the 32SII?

Re: HP 33S CPU speed Message #8 Posted by John Smitherman on 20 Sept 2004, 6:50 p.m., in response to message #1 by Ed Look Hi Ed. I am not sure of the absolute CPU speed but the relative speed that I measure for the 33s using the method given at: http://www.thimet.de/CalcCollection/CalcPerformance.html is about a 44. This places the 33s as follows: 42s: 13 32sii: 25 32s: 33 33s: 44 48GII: 46 28s: 93 * Note: higher numbers represent faster performance on this benchmark. Have others run speed benchmarks on the 33s? Regards, John

Re: HP 33S CPU speed Message #9 Posted by Ed Look on 20 Sept 2004, 9:37 p.m., in response to message #8 by John Smitherman Thanks, John. It isn't too bad as far as that list is concerned, anyway. But the 48GII doesn't look so good compared to these others! I assume the 49G+ is a little better?

Re: HP 33S CPU speed Message #10 Posted by John Smitherman on 21 Sept 2004, 5:23 p.m., in response to message #9 by Ed Look Ed, there are results given for many models of calculators at the web site in my previous post. Specifically, there are 2 results (46 and 221) given for the 48GII which vary significantly. I do not understand the footnotes which I assume help explain the difference. It does not show a result for the 49G+. Happy computing! Regards, John

Re: HP 33S CPU speed Message #11 Posted by Richard on 21 Sept 2004, 9:28 p.m., in response to message #10 by John Smitherman Ed, The HP-33 is 2x faster then the HP-32SII for the integral (4 X sqr(1-x^2)dx)with full precision (pi to 12 digits)14 sec / 28 sec (8 sec on 48G+) Bad LCD contrast display but faster CPU then HP-32SII... Regards

Re: HP 33S CPU speed Message #12 Posted by Ed Look on 22 Sept 2004, 1:29 a.m., in response to message #11 by Richard I used to say to myself that I'd be really happy if the 33s came packaged in a Pioneer or Pioneer-based form. While that is always going to be very true, I have found the 33s growing on me: I don't really see the chevron shape so much anymore, unless I somehow recall our moanings about it. The presence of a built-in constants list does make life a little easier for me. As you guys have all pointed out, it is pretty fast for most calculations I need, infinite series convergence calculations notwithstanding. And, despite the small relative increase in program memory when translated into 32SII terms, it IS STILL an increase, and any extra RAM in any electronic device is always welcome; the more the merrier. Well, a few more registers or labels would be very welcome, too. The silvery overall color isn't really so bad, nor the display font... ... but that radix... a decimal point that an AFM couldn't pick up... that will still take some getting used to, especially since it really doesn't look all that different from the comma! Oh, yeah, I don't mind green and purple on a silver background, but I think they could have picked shades of green and purple that might have contrasted with each other and the metal of the faceplate a little more. All in all, not a bad little calculator for (serious) everyday use, especially in a laboratory. Another oh yeah- pretty nice case it came with, with a cool, embossed "hp" logo.

33S colors Message #13 Posted by Gene on 23 Sept 2004, 1:32 p.m., in response to message #12 by Ed Look "Oh, yeah, I don't mind green and purple on a silver background, but I think they could have picked shades of green and purple that might have contrasted with each other and the metal of the faceplate a little more." Gene: And this is exactly one of the examples I'll be presenting at HHC2004 this weekend. I have established a numerical calculation for contrast between colors and examined quite a few HP choices of current (and older) calculators. Some are pretty good, some are, ahem, not so good. After the conference, I'll see about putting the powerpoint file up for anyone to download, if interested.

Re: 33S colors Message #14 Posted by Ed Look on 23 Sept 2004, 10:28 p.m., in response to message #13 by Gene Go get 'em, Gene.

Re: 33S colors Message #15 Posted by Hugh Evans on 25 Sept 2004, 9:45 a.m., in response to message #13 by Gene Please do so. I could really use such a document!

Benchmarks Message #16 Posted by Tizedes Csaba [Hungary] on 23 Sept 2004, 7:00 a.m., in response to message #11 by Richard Hello, I was send a letter about HP32SII and HP15C benchmarking with correct running time measures in 5 case (built-in looping/DSE,ISG/, counting, storing, basic arithmetics and subroutine calling), (about in 27th February 2004) but I cant to find that... Where is my response??? :) Csaba

The 28S has a P.I. of 54, not 93... Message #17 Posted by Karl Schneider on 22 Sept 2004, 1:43 a.m., in response to message #8 by John Smitherman ... but thanks for the link. I might try several, out of curiosity. I suspect that the Voyagers (at least the 15C) could be speeded up by GTO'ing a line number rather than a label. -- KS

Re: The 28S has a P.I. of 54, not 93... Message #18 Posted by Vieira, Luiz C. (Brazil) on 22 Sept 2004, 3:02 a.m., in response to message #17 by Karl Schneider Hi, Karl; About GTO'ing a line number, I'd like to know one thing. Should this sort of GTO be faster if # of bytes for program line is fixed-ratio only? I mean, if each program holds a one-byte instruction, I see that the GTO (line#) is a linear biased operation. Otherwise, if a different byte count is found in line numbers, there's a need for computing before jumping. As the only original Voyager with a different byte count for program instructions is the HP15C (I'm not counting the HP12Cp), I guess that in some cases it is probably slower than an HP11C if [GTO][I] is executed when [I] has a negative number (backwards relative jump). Did anyone try/benhcmark this? Just reasoning about. Luiz (Brazil) Edited: 22 Sept 2004, 3:02 a.m.

GTO line # on HP-11C & -15C Message #19 Posted by Karl Schneider on 22 Sept 2004, 9:57 p.m., in response to message #18 by Vieira, Luiz C. (Brazil) Luis -- I did benchmark the test program on my HP-15C and my recently-acquired 11C, with and without the "GTO line #" technique. Valentin Albillo advocates it as a speed-up trick. Both the 11C and the 15C took 22-25 seconds to run the program, with "GTO lbl" and "GTO I" [with -line # in I]. BTW, this contradicts the table, which suggests taht the 15C is considerably faster than the 11C. I suspect that "GTO I" [with -line # in I] could execute faster than "GTO lbl" because the calc is being told exactly where to branch, rather than searching for a label. How much run time is saved would depend upon number of loop executions and how many instructions must be scanned before the label is found. BTW, "GTO I" [with -line # in I] goes directly to the specified line # on the 11C and on the 15C. This is documented on pp. 108-109 of the English-language HP-15C manual. (I'm still awaiting my 11C manual via the USPS' fine Media Mail service.) There is no "backward relative jump". -- KS Edited: 22 Sept 2004, 10:17 p.m.

Oops! Message #20 Posted by Vieira, Luiz C. (Brazil) on 23 Sept 2004, 2:18 a.m., in response to message #19 by Karl Schneider Hi, Karl; You're right. I confused with the HP67/97 "GTO (i)" mechanics. In these calculators, if I-register contents is a negative number, the program counter jumps backwards the number of steps defined in the integer part of the number. It even goes through line 000- when needed. (sometimes I should read manuals prior to write answers... In fact, always!) Cheers and thank you for your measurements. Luiz (Brazil)

Re: The 28S has a P.I. of 54, not 93... Message #21 Posted by John Smitherman on 22 Sept 2004, 9:29 a.m., in response to message #17 by Karl Schneider You are right Karl. Looking back I think that I intended to type 48s and 83 instead of 28s and 93. Must have been having a bad day! I ran the benchmark on the 33s with values of 10, 20, 50 and 100 in the outer loop to get a sense of how much variability there would be. I then took and reported the average of the 4 test results. Regards, John

[ Return to Index | Top of Index ]

Go back to the main exhibit hall