Speed of the hp calculators family

[wojtek]

Where could I find data about comparison of speed of HP41c, 41cv, 41cx, 42s, 28s and other machines (48g, 50g, Prime) ?

[Accutron]

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

[wojtek]

Thank you. So the champion is 50g!
I had read somewhere that DM-41L is much (a few times) faster than regular 41c. According to this benchmark there is almost no difference. Was I seeing things?

[Tugdual]

I see that the 50g is assembly while the Prime is formula and list (whatever this means). I think you need to compare aple with aple to make a call cuz I can't believe the Prime is slower than the 50g.
This being said, speed is one thing but knowing your way is more important.

[Gerson W. Barbosa]

(03-01-2016 10:25 PM)Tugdual Wrote:  I see that the 50g is assembly while the Prime is formula and list (whatever this means). I think you need to compare aple with aple to make a call cuz I can't believe the Prime is slower than the 50g.
This being said, speed is one thing but knowing your way is more important.

Again, apples not being compared to apples. But at least the Prime wins hands down here :-)

http://www.thimet.de/CalcCollection/Calc...mance.html

[Claudio L.]

(03-01-2016 10:25 PM)Tugdual Wrote:  I see that the 50g is assembly while the Prime is formula and list (whatever this means). I think you need to compare aple with aple to make a call cuz I can't believe the Prime is slower than the 50g.
This being said, speed is one thing but knowing your way is more important.

The HP Prime CPU runs at 400 MHz, versus 192 MHz on the assembly benchmark for the 50g. The CPU on the prime could in theory run the exact same assembly code in roughly half the time.
But the key word is "could", because in reality it "can't", by being a very closed system. So the benchmark is actually quite accurate: you CAN use the 50g in C or assembly and get that kind of performance to solve N-Queens and many other problems. On the Prime you are limited to the interpreted language, which performs as shown on the list.
If they only made the Prime a little more open it would certainly take the throne, but "as-is" is not the fastest of the bunch.
Lots of respect to Xerxes for compiling that list, it's probably the most complete benchmark for calculators you can find on the internet.

[walter b]

(03-02-2016 02:19 AM)Claudio L. Wrote:  Lots of respect to Xerxes for compiling that list, it's probably the most complete benchmark for calculators you can find on the internet.

1+
Impressive list!

d:-)

[Raymond Del Tondo]

(03-01-2016 11:42 PM)Gerson W. Barbosa Wrote:  

(03-01-2016 10:25 PM)Tugdual Wrote:  I see that the 50g is assembly while the Prime is formula and list (whatever this means). I think you need to compare aple with aple to make a call cuz I can't believe the Prime is slower than the 50g.
This being said, speed is one thing but knowing your way is more important.

Again, apples not being compared to apples. But at least the Prime wins hands down here :-)

http://www.thimet.de/CalcCollection/Calc...mance.html

While we're at apples and peas: The HP 48GX
runs the benchmark in 0.336 seconds (HP-48GX Assembly / Saturn @ 3.7-4.0 MHz)
at a CPU clock of less than 4MHz , whereas the 50g and Prime need a 192+ MHz modern ARM CPU core to reach a similar speed range;-)

[wojtek]

(03-02-2016 08:37 AM)Raymond Del Tondo Wrote:  While we're at apples and peas: The HP 48GX
runs the benchmark in 0.336 seconds (HP-48GX Assembly / Saturn @ 3.7-4.0 MHz)
at a CPU clock of less than 4MHz , whereas the 50g and Prime need a 192+ MHz modern ARM CPU core to reach a similar speed range;-)

50g assembly has two modes: native ARM and Saturn emulation mode. Perhaps you are talking about emulation mode. In native mode it would be much faster than 48g (IMO)

[emece67]

(03-02-2016 08:37 AM)Raymond Del Tondo Wrote:  While we're at apples and peas: The HP 48GX
runs the benchmark in 0.336 seconds (HP-48GX Assembly / Saturn @ 3.7-4.0 MHz)
at a CPU clock of less than 4MHz , whereas the 50g and Prime need a 192+ MHz modern ARM CPU core to reach a similar speed range;-)

0.000150 HP-50G Assembly / ARM9 @ 192 MHz / Fast Mode x2.7

That's 2000+ times faster, whereas the CPU clock speed is only 50x faster. Obviously, the newer 32 bit ARM processor surpasses the venerable Saturn architecture.

[walter b]

(03-01-2016 09:14 PM)Accutron Wrote:  http://www.hpmuseum.org/cgi-sys/cgiwrap/...i?read=700

Hey Pauli, did you know we had built the fastest RPN programmable ever?! Hope I didn't overlook anything significant.

d:-)

[wojtek]

(03-02-2016 12:16 PM)SOME SPAMMER Wrote:  hi friends,
good to see you all discussion and experience about Rolex watch. this is a very big and trusted brand in all world. for more details please visit....
[DELETED SPAM URL]

Tell us pls what pils are you taking

<<<MODERATORS NOTE: wojtek; that was a spammer that Katie had axed from the site, after you started writing this but before you posted. That's why it only shows up in the quote. Thanks for finding it though. END NOTE>>>

[Claudio L.]

(03-01-2016 11:42 PM)Gerson W. Barbosa Wrote:  Again, apples not being compared to apples. But at least the Prime wins hands down here :-)

http://www.thimet.de/CalcCollection/Calc...mance.html

I've never seen this benchmark before, thanks for reposting it. I couldn't resist doing the test, and the 50g using newRPL firmware gave the following results:

At 12 digit precision: (7.26 seconds for 1000 loops), performance index = 4681
At the default 32-digit precision: (12.23 seconds for 1000 loops), performance index = 2786

So it seems the Prime in this benchmark is about 2.6x faster than newRPL on the 50g at the same precision (let's not forget the CPU clock is 2x faster on the Prime).
This brings the 50g right below the Prime on that list.

[Gerson W. Barbosa]

(03-02-2016 02:03 PM)Claudio L. Wrote:  

(03-01-2016 11:42 PM)Gerson W. Barbosa Wrote:  Again, apples not being compared to apples. But at least the Prime wins hands down here :-)

http://www.thimet.de/CalcCollection/Calc...mance.html

I've never seen this benchmark before, thanks for reposting it. I couldn't resist doing the test, and the 50g using newRPL firmware gave the following results:

At 12 digit precision: (7.26 seconds for 1000 loops), performance index = 4681
At the default 32-digit precision: (12.23 seconds for 1000 loops), performance index = 2786

So it seems the Prime in this benchmark is about 2.6x faster than newRPL on the 50g at the same precision (let's not forget the CPU clock is 2x faster on the Prime).
This brings the 50g right below the Prime on that list.

I have an older 50g I don't use anymore. One of these days I will try newRPL on it :-)
Quite an interesting project!

[Raymond Del Tondo]

(03-02-2016 10:04 AM)emece67 Wrote:  

(03-02-2016 08:37 AM)Raymond Del Tondo Wrote:  While we're at apples and peas: The HP 48GX
runs the benchmark in 0.336 seconds (HP-48GX Assembly / Saturn @ 3.7-4.0 MHz)
at a CPU clock of less than 4MHz , whereas the 50g and Prime need a 192+ MHz modern ARM CPU core to reach a similar speed range;-)

0.000150 HP-50G Assembly / ARM9 @ 192 MHz / Fast Mode x2.7

That's 2000+ times faster, whereas the CPU clock speed is only 50x faster. Obviously, the newer 32 bit ARM processor surpasses the venerable Saturn architecture.

Agreed. Didn't notice the 50g assembly version.

(However at least I made the fastest 48G version;-)

[Jlouis]

Don't need to bother about being fast...calculators are ladies....see the 15c....and everybody knows that a woman needs more time....

[Jim Horn]

Good point. But today's speeds are a real help. I worked hard to get the HP-67 to be able to factor any input value within its integer range (up to 10^10). The worst case, 9999999967, took about 3 3/4 hours as I recall. And its battery pack seldom lasted that long on a charge.

That the WP 34s can do that in a fraction of a second is *very* much an improvement!

[Dave Britten]

(03-05-2016 10:12 PM)Jim Horn Wrote:  Good point. But today's speeds are a real help. I worked hard to get the HP-67 to be able to factor any input value within its integer range (up to 10^10). The worst case, 9999999967, took about 3 3/4 hours as I recall. And its battery pack seldom lasted that long on a charge.

That the WP 34s can do that in a fraction of a second is *very* much an improvement!

Was it the venerable mod 30 seive, or something craftier? Would be a good test for my DM-41L. I wonder if it can get power from the USB port...

[multilingualtricker]

- 0.000192 HP-50G HPGCC 3.0 / RegVars / Cross Compiler / Fast Mode x2.5 (75->192 MHz)
-
- 0.000150 HP-50G Assembly / ARM9 @ 192 MHz / Fast Mode x2.7

Is the latter of these 2 where you program in arm code directly on the 50g?
Phew that must be difficult!