TJ25 is twice as fast as a UX50 !

[hotpaw4]

Here are some preliminary yCPUBench 0.7 numbers:

68k sieve - TJ25 61 bogo_68k_MHz - UX50 29.5 bogo_68k_MHz
arm sieve - TJ25 1620 units - UX50 1012 units (gcc compiled)
68k Mandelbrot - TJ25 205 kiloflops - UX50 102 kiloflops
arm Mandelbrot - TJ25 2.44 megaflops - UX50 1.43 megaflops (CW v9.2 compiled)

PACE looks better optimized for the MXL processor; the Sony HHE only runs legacy 68k code only half as fast. For native ARM code, the performance ratio looks more similar to the ratio of CPU clock rates.

These were 2 random units on display at Fry's in Sunnyvale, so I don't know whether some hacks/etc. could have been slowing down either unit.

yCPUBench is here: http://www.hotpaw.com/rhn/palm

[riversen]

Great stuff... only for someone like me, can you translate these numbers? I am an analyst myself, but it would be nice to see how these sets of numbers are referenced? Perhaps, a simple graph could be generated showing from baseline to better. It's just a thought.

[s_n_m]

The UX scales its power dontcha know? In movieplayer and flashplayer it bumps up to max speed, while in most apps it stays pretty low.

[hotpaw4]

Well, the native ARM code benchmark numbers scale pretty closely with the peak CPU clock speed. So on those, I doubt the HHE/UX50 was running much slower than 123 MHz, it's rated max according to the Sony press releases. But perhaps the UX runs PACE/legacy/68k apps at a lower power setting?

[s360clie]

remember the ux has a seperated dsp

[hotpaw4]

dsp? Irrelevant from the point of view of benchmarking or running most applications. Is there any software which actually uses the dsp except for the bundled movie/voice/music players?

[AWS]

Isn't that is the point of the "superior" design of the UX50... Having a dsp coupled with a scaling processor for maximized efficiencies?

I think it is hard to do straight benchmarking when comparing some of the newer models that use dsp chips for specialized purposes. Raw processor speed is not necessarily an indicator of the best PDA. For most things that folks do on a PDA, minimal power is necessary, so why waste battery power when the power is not needed?

[hotpaw4]

Originally posted by AWS
Raw processor speed is not necessarily an indicator of the best PDA. For most things that folks do on a PDA, minimal power is necessary, so why waste battery power when the power is not needed?

I agree, but the SJ25 is not only faster in basic PIM performance, but also smaller, cheaper and has a longer battery life than the slower UX50. That's a quadruple win. The UX only wins on multiple features (movies, keyboard, comms).

[harryleung]

does Sony use their own build chip for TJ model?

[hotpaw4]

Originally posted by harryleung
does Sony use their own build chip for TJ model?

The TJ25/35 use the Motorola MXL processor, similar to the old Dragonball line, except with a much faster 200 MHz ARM9 CPU built-in.

The Sony manufactured HHE chip in the UX50 includes a slower variable-speed ARM9 CPU, but includes lots of local memory and a DSP for multimedia acceleration (for movies, etc.)