42th anniversary of HP-35?

[jebem]

Is 2014 the 42th anniversary of the HP-35?
HP has lost an opportunity here to release a anniversary version of the HP-42S... maybe a HP-42SII running on ARM system?
But wait, the year is not finished and HP never release information on future products... nothing is lost yet.

[HP67]

Please, no more ARM-based junkware.

I know it won't happen, but they really ought to go back to proprietary designs and fab their own chips and own the product from end-to-end instead of using off the shelf parts like a bunch of high school kids. I would think by now the R&D is already paid up.

An anniversary 42S? Yes! Running on ARM? No thanks.

[Mike Morrow]

(06-22-2014 05:03 PM)HP67 Wrote:  An anniversary 42S? Yes! Running on ARM? No thanks.

There's nothing inherently wrong with an ARM-based calculator. What is very wrong with most attempts is using the original firmware of some legacy model on an emulation of the original hardware on an ARM. The emulation layer ruins the result, not the ARM.

The flaws of the HP-15C Limited Edition stem directly from emulation of the original HP-15C hardware on real ARM hardware. Emulation results in gross wastes of run time and battery power to support those run times, compared to the same ARM-platform with firmware native. The HP 30b runs on an ARM with native firmware. It calculates a 2500-iteration Savage benchmark in 6 seconds producing a 12-digit result of very good accuracy. In comparison, the HP-15C LE runs on an ARM with a clock speed very similar to the HP 30b, but it uses old firmware and an emulation layer. The same benchmark takes 48 seconds to produce a 10-digit result of far lower accuracy. Thus, the HP-15C LE takes eight times longer to run and eight times the power consumption to produce a significantly inferior numerical result compared to the HP 30b. Had that same HP-15C LE ARM hardware been executing native firmware, results would doubtless have been very similar to those of the HP 30b.

Speed improvements of the HP 49G+ and HP 50G resulting from introduction of ARM processors are minimal, due to the Saturn emulator layer that runs on the ARM. It would be lightening-fast without an emulation layer.

[Mobius]

(06-22-2014 05:03 PM)HP67 Wrote:  ...they really ought to go back to proprietary designs...

I never thought i'd ever read or hear that in this day in age hahah. Off the shelf chips with rock solid native code any day for me.

[Gerald H]

(06-22-2014 07:19 PM)Mike Morrow Wrote:  

(06-22-2014 05:03 PM)HP67 Wrote:  An anniversary 42S? Yes! Running on ARM? No thanks.

There's nothing inherently wrong with an ARM-based calculator. What is very wrong with most attempts is using the original firmware of some legacy model on an emulation of the original hardware on an ARM. The emulation layer ruins the result, not the ARM.

The flaws of the HP-15C Limited Edition stem directly from emulation of the original HP-15C hardware on real ARM hardware. Emulation results in gross wastes of run time and battery power to support those run times, compared to the same ARM-platform with firmware native. The HP 30b runs on an ARM with native firmware. It calculates a 2500-iteration Savage benchmark in 6 seconds producing a 12-digit result of very good accuracy. In comparison, the HP-15C LE runs on an ARM with a clock speed very similar to the HP 30b, but it uses old firmware and an emulation layer. The same benchmark takes 48 seconds to produce a 10-digit result of far lower accuracy. Thus, the HP-15C LE takes eight times longer to run and eight times the power consumption to produce a significantly inferior numerical result compared to the HP 30b. Had that same HP-15C LE ARM hardware been executing native firmware, results would doubtless have been very similar to those of the HP 30b.

Speed improvements of the HP 49G+ and HP 50G resulting from introduction of ARM processors are minimal, due to the Saturn emulator layer that runs on the ARM. It would be lightening-fast without an emulation layer.

Hear! Hear! & just for this reason my favourite calculator is the HP 49G.

[HP67]

(06-22-2014 07:19 PM)Mike Morrow Wrote:  There's nothing inherently wrong with an ARM-based calculator.

That's a matter of opinion. And I believe there is a lot wrong with it.

(06-22-2014 07:19 PM)Mike Morrow Wrote:  What is very wrong with most attempts is using the original firmware of some legacy model on an emulation of the original hardware on an ARM. The emulation layer ruins the result, not the ARM.

That's an implementation problem, not an indictment of emulation. Maybe the emulation just isn't good enough. Given buying any new device means having off the shelf el-cheapo commodity SOC crapware jammed down our throats I would prefer to go with known good firmware that behaves exactly like it did in the real device they're copying rather than a whole new set of problems caused by reinventing the wheel. Why mess with success? All they needed was higher clocks and maybe a little better display. The old HP 48 would be as good as ever with a faster processor.

(06-22-2014 07:19 PM)Mike Morrow Wrote:  Speed improvements of the HP 49G+ and HP 50G resulting from introduction of ARM processors are minimal, due to the Saturn emulator layer that runs on the ARM. It would be lightening-fast without an emulation layer.

They would also be lightning fast running on a Saturn chip clocked around 100 MHz, with good battery life and something that could actually be called an HP calculator, rather than a generic creation from Kinpo running on a generic platform just like all the other handheld devices in the world.

[Mike Morrow]

(06-23-2014 06:48 AM)HP67 Wrote:  That's an implementation problem, not an indictment of emulation. Maybe the emulation just isn't good enough.

You miss the fundamental problem. No one ever wrote an emulation layer between old firmware and new hardware that did NOT result in a system running many times slower than the same system would have, had new firmware native to the new hardware been used. You missed the very concrete example of the Savage benchmark running eight times slower on the HP-15C emulated on an ARM, compared to running on the HP 30b whose firmware is programmed native to a very similar ARM with similar clock speed. The HP-15C old firmware emulation on the HP-15C LE will take about eight times as long to produce a rather inferior result and use eight times the battery energy to achieve that. Were the HP-15C emulation perfect and perfectly optimal, there would doubtless be some improvement, but also gross difference in benchmark speed remaining. We are NOT talking about emulating a calculator on a PC...that's completely unrelated and irrelevant. Emulation of a calculator on calculator hardware is always very inefficient.

Quote:The old HP 48 would be as good as ever with a faster processor.

True, as long as there is little expectation attached to the phrase "good as ever". I own and have used the original HP 48SX and HP 48GX since first introduction. The performance and versatility of each is grossly inferior to the HP 50G...or even the infamous HP 49G+. Were they sped up with a processor upgrade, they would remain terribly limited machines that were now simply faster.

I do not understand the worshipful attitude that some have for earlier HP calculators, which seems to begin as soon as they are no longer sold by HP. I've used HPs since the first HP-35 units showed up at Georgia Tech in Spring 1972. For the most part, worship is undeserved. The 26-year-old HP-42S came closest to being a perfect machine...yet even it still had a long long way to go.

[HP67]

Maybe I misread your post but it sounds like what you're saying is that if I only knew as much as you do I would agree with you on everything.

No, I don't think so Mike. Not by a long shot.