Handheld Engine benchmark and cam quality

[Jeffry]

Click here:

http://pocketgames.jp/modules.php?op=modload&name=Reviews&file=index&req=showcontent&id=107

[Shogmaster]

BOO HISSS~~~~~~~~

More expensive, slower, and smaller yet less bright screen!?! No freakin' thanks!

UX50 is offically off my list. Down to NX80 and maybe TT3.

[hausman]

LOL not knowing too much about the benchmarking here (and unable to read japanese and not trusting babelfish for this sort of thing)...I still wonder who needs a supercomputer to look up a contact in the addressbook. According to the marketing hype, the UX has what it takes when it comes to CPU-intensive video playback...other than that I can't imagine doing much of anything on a PDA that requires a 400MHz processor...at least not on a Palm PDA (this coming from a Zaurus user who does indeed run a staging server with Apache and mySQL on the PDA - so I know what it's like to need MHz - I just don't expect to be doing that on my UX)

[cykalan]

no no no, benchmark 2.0 isn't a very good benchmark program for all OS5 CLIEs.

i will wait for actual program usage results and/or speedy benchmark results.

[OcellNuri]

Considering the CPU is scaling, how do we know it's going to full speed for the benchmark? It may view the benchmark program as a less power hungry app, and downclock to well under 100mhz. I agree with hausman and cykalan. The true benchmark is going to be hands on. This new CPU seems too tricky to apply an old benchmark program to.

[TheZodiac]

I do believe this new clie cpu is optimized for the Palm OS6, and this is an interim model to see how the form factor does in the public eye. I was going to buy but now I think I will *have to* wait for a POS6 device.

And like Sony said - they will not be suprised if other manufactures will want to license the Handheld Engine. Which, takes some of its technology path from the new PSX architecture.

[Unregistered]

I don't think UX 50 can out run Z71, let alone T|C. It might have fast graphic accelerator, but that won't change a thing until somebody write a software for it.

[pdagal]

Quartus Benchmark hasn't been updated since 2000 and was written back in the days of the Motorola Dragonball (hey, the Visor Platinum was the reference for speed then!). I have a feeling it's not properly gauging new CPUs. My NX80 benchmarks at 271%, comprared to the 33 MHz Dragonball Visor Plat at 204%. It should come out 6x faster based on raw clockspeed, let alone taking into account the difference between a 68000 family chip and the ARM RISC processor.

[Unregistered]

nonetheless even compared to other ARM models the HHE comes out really short. (OMAP, Xscale)

the benchmark 2.0 reports a whopping 500% for the T|C. It is true this first number is preliminary, but it is a bad start. And personally I don't believe Sony can suddenly deliver a design coup that can out do what TI and Intel has done.

and strangely enough, Sony hasn't shown a single slide presentation showing the usual comparison between HHE against other ARM chip in the same class.

It is made using .18 micron btw. It will be a good while till Sony can make a .13 or .9 micron chip that can out run Samsung's or Intels top end.

[X Destruction]

Guys, it's probably going at 8mhz-33mhz on that old program which needs almost no speed. That's nothing to base it on. If it's a more cpu hungry program, it'll use the the full 133mhz and all the other help the engine holds.

[EdFrmBrighthand]

Benchmark 2.0 is a horribly out of date application. It was written for 68k Dragonball processors and so everything it does is in emulated 68K code and it doesn't make any use of ARM code. I'm not surprised it says a new ARM-based processor isn't very fast.

My friend Eugene, who lives in Tokyo and has access to the UX units on display there, ran yCPUBench on the new models. This is a benchmarking app designed to test Palm OS 5 handhelds and runs a variety of tests. I'm not spending 20 minutes putting in all the results so you'll have to accept that this is representative. It is a floating point math intensive benchmark. It calculates, using double precision math, a Mandlebrot set (depth of 100) for a 32x32 size monochrome bitmap.

UX50:
Time (mS/Loop) = 796
kFlops/s = 190

NX80V:
Time (mS/Loop) = 512
kFlops/s = 296

This shows the UX50 is slower than the NX80V. No surprise, as we know the NX80V uses a 200 MHz XScale processor, while the top speed of the Sony Handheld Engine is 123 MHz.

He also ran Speedy 2.6 on it. This performs a processor speed test, a memory speed test, and a graphical speed test. The model for comparison is a Palm Vx, which has a score of 100. The UX50 got a score of 2174. The NX80V got a score of 3125.

Again, this shows the UX50 is slower than a NX80V.

Of course, this begs the question of why Sony used a processor it knows is slower. It has repeatedly emphasized that the Handheld Engine uses very little power. Plus it has the built-in Digital Signal Processor (DSP) and graphics accelerator chip, allowing it to show video at 30 fps at 320 by 480 pixels.

[Quantumduck]

All of this seems to me like the guy who buys a Ferrari because it is one of the fastest cars in the world, and then never drives it faster than 65 mph.  Why does one need the fastest processor in the world to run a datebook, a phone list, an MP3 player (very little processor load really), a wordprocessor (virtually nil), a spreadsheet (ok, now we need some speed, but only when recalculating) etc.

The real reason to have speed when using a Palm is on taking pictures, the voice recording or the video playback.  Comparisons are not going to be valid on these areas because of the build in DSP and the Video chip.  The raw number comparisons are apples to oranges. 

Why get upset when the orange isn't red?:confused:

[ksjenkins]

This is all great information, but.......... The NX80 uses a XScale, the UX uses an ARM core. So it would seem that there is a lot of interpolation. I'm sure if you ran performance tests developed for an ARM and ran the tests on a XScale the numbers would be skewed the other direction (am I assuming too much?). Plus the UX has the DSP and Graphic Co-processors that don't factor into these RAW CPU tests.

The whole is stronger than the sum of its parts, anyway it's going to be tons faster than my 33MHz Dragonball VZ equiped N760. And besides IMHO good speed and great battery life are the main goals in a PDA.

I think Intel has really gotten everyone worked up over the whole the more MHz you have the FASTER the platform is. This is a Desktop/Server religious debate that doesn't belong in the PDA niche. Of course unless you are using PocketPC then go right ahead and quote Moores Law or what-ever, have at it.....:eek:

Also the screen brightness pics are kind of misleading & besides who runs their PDA's at 100% brightness, not me I run mine at about 25% almost all the time.

And that's all I'm going to say about that!

:D

[Shogmaster]

What's the damn point if the user has no say in how it scales?!!? Taking control away from the user is NEVER cool.

The thing I love about the Lightspeed Xscale proc over/under clocker app is that I can tell the program (and the device) exactly WHAT applications to run HOW FAST. For instance, I want my Acid Image running fast as possible, but all my default Palm OS programs running slow as possible.

If there's no way for the user to tell UX how to scale HHE, then it's all freakin moot, IMO.

[Shogmaster]

Originally posted by ksjenkins
This is all great information, but.......... The NX80 uses a XScale, the UX uses an ARM core. So it would seem that there is a lot of interpolation. I'm sure if you ran performance tests developed for an ARM and ran the tests on a XScale the numbers would be skewed the other direction (am I assuming too much?).

Intel's XScales are revisions of their older Strongarm processors. ;)

[n2ifp]

I am not going to poo-poo the UX50 yet until it's out and I see what it does. In the meantime, I'll suffer with the NX80 :D!

[Burns]

Originally posted by Shogmaster


Intel's XScales are revisions of their older Strongarm processors. ;) I hate it when people resort to Strongarm tactics ;)

[chrisclie]

Originally posted by Shogmaster
What's the damn point if the user has no say in how it scales?!!? Taking control away from the user is NEVER cool.

The thing I love about the Lightspeed Xscale proc over/under clocker app is that I can tell the program (and the device) exactly WHAT applications to run HOW FAST. For instance, I want my Acid Image running fast as possible, but all my default Palm OS programs running slow as possible.

If there's no way for the user to tell UX how to scale HHE, then it's all freakin moot, IMO.

Well Shogmaster, you're a power user, while the vast majority (probably > 95%) of the target market that Sony is going after are not. Most people don't care and don't ever want to know about overclocking/underclocking ... so HHE's auto scaling technology is a perfect fit :) ...

[aithir]

It is very well expected Sony's CPU to be less powerfull and maybe more buggy than Intel's. It's their first implementation...This is going to be a very good PDA CPU, after 2 or 3 generations. By then they would have solve most problems and deliver a really good and sophisticated chip. They are just testing it now.

[nauta]

Originally posted by aithir
It is very well expected Sony's CPU to be less powerfull and maybe more buggy than Intel's. It's their first implementation...This is going to be a very good PDA CPU, after 2 or 3 generations. By then they would have solve most problems and deliver a really good and sophisticated chip. They are just testing it now.

More buggy than Intel? Is that possible? ;)

If Sony were still testing the chip, they wouldn't release it into the mainstream as a "beta product". Could you imagine the PR sh*tstorm that would bring? The first generation of a processor doesn't mean it's going to be buggy; it depends on how much testing the company has done before they decide to let it out of the lab. Personally, I trust Sony more than Intel.

[SumDumGuy]

What makes you think the chip was built from scratch?

[nauta]

Originally posted by SumDumGuy
What makes you think the chip was built from scratch?

As far as I know, they based it on PlayStation 2 technology.

[jklfafa]

i think in the end the ux will actually be slower. But the advantages are that it consumes less power.it appears they are try to make cut backs here and there. THe zaurus on the other hand looks like a great device! high resolution screen but it uses linux. I have the zaurus 5500 with openzaurus which is kinda ok but if feels like a mac pc comparison cuz of little support. But this device has wifi and bluetooth. Indeed i can see how it was ment for coperate usages. BUt nonetheless i want to get my hands on this device. I really dont think we should be judging this device until an english speaker can get there hands on it and thoroughly explore it...Im using a jornada 690 which runs on sh31 33 mhz
regardless of speed its a great device, but slow...but im thinking this device will not be topping and new speeds, but simply as fast as the current mid range devices while conserving power.
Personally if i can get this device in the 300-400 dollar range while it is NEW and it does not WEAR out ever then id love it
im sooo sick of worry and estimating the life of my nx70
i have had pdas since 96 and all of them can run and my cassiopeia a11 is still in great condition. devices used to be much better quality. dumb sony just tries to dish one out that lasts for a year. WHere are the endurance tests??
ill wait.
i was so sceptical to look at the samsung a600 phone cuz its a swivel!!!
there are ways to make a swivel sturdy.
when my nx70v (primary device dies from wear and tare) ill use jornada 690 for primary... i dont know whats taking palm to get the t3 out so slow
if they had it out now
they can grab the market
sigh...i just want a sturdy device with wifi
my toshiba e740 was such a disappointment
batterylife butt ugly case, and very unstable system
i sold it to get my nx70v
im kinda happy but wifi has its advantages in college

[c0mm0n]

This news proc just make me think of Intel PenitumM ; low frequency low voltage but high performances and great autonomy.

Reading the topics, on several boards many people shout on the ux series because of the proc frequency, but we all now know it'sn't only about mhz...

[ConceptVBS]

Originally posted by Quantumduck
All of this seems to me like the guy who buys a Ferrari because it is one of the fastest cars in the world, and then never drives it faster than 65 mph.  Why does one need the fastest processor in the world to run a datebook, a phone list, an MP3 player (very little processor load really), a wordprocessor (virtually nil), a spreadsheet (ok, now we need some speed, but only when recalculating) etc.

The real reason to have speed when using a Palm is on taking pictures, the voice recording or the video playback.  Comparisons are not going to be valid on these areas because of the build in DSP and the Video chip.  The raw number comparisons are apples to oranges. 

Why get upset when the orange isn't red?:confused:

Not everyone of us are like you. There are other varieties in the fruit basket. :)