Are Pocket PCs Starving?

[Jason Dunn]

<a href="http://www.cewindows.net/faqs/pocketpcstaving.htm">http://www.cewindows.net/faqs/pocketpcstaving.htm</a><br /><br />Chris De Herrera has written a column on Pocket PC data speeds. It's an interesting topic since as we move into the realm of 400 mhz CPUs, can the other components on the Pocket PC keep up? I've always found it rather surprising that for a RAM-based computer, Pocket PCs are pretty slow.<br /><br />"The bus width in combination with the bus speed has a dramatic effect on the performance of computers. The bus width defines how many bits or bytes of data that can be read from the ram of the computer into the processor. The bus speed defines how fast the data can be read from the ram to into the processor. So think of the bus width as the number of lanes on a highway and the bus speed as the speed limit. Right now the StrongARM and XScale CPUs use a 100 mhz bus speed. Also the designs of the Pocket PC that are currently available use a 16 bit bus. This means that the maximum bus speed is 200 megabytes per second."<br /><br />I'd be interested in hearing some of you highly-technical users chime in with your own opinions. What hardware elements need to be improved in order for us to see radical speed improvements? I personally think the video sub-systems are dragging things down - without any true hardware acceleration, the CPU has to do a lot of heavy lifting. Just wait until that ATI mobile chip arrives in a Pocket PC...

[Paragon]

Just wait until that ATI mobile chip arrives in a Pocket PC...


AS in the ATI chip mentioned here http://paragon0.tripod.ca/ATI1.html
:) :)


Dave

[paulv]

Jason,

I reckon much of the issue is in the coding of Pocket Windows. There's so much of an effort to make sure that Windows is tackling a million threads that the poor old processor doesn't get much time to actually run user programs.

Given that even my Jornada is 206MHz that's a lot of MIPS and yet Pocket Excel (which is a lot more compact than the regular version - and it runs from Ram) doesn't run as fast the full version used to on my old P-120 desktop (previous computer).

I think that the design of these PC's are seriously dodgy. I'd be interested in finding out if Linux runs at a similar pace or whether it thinks a 206MHz processor is fine - maybe one of the IPAQ owners can say.

I reckon that the programmers are victims of having multi-MIPS available and have forgotten how to program tight code.

[Jason Dunn]

http://paragon0.tripod.ca/ATI1.html


Uh.....huh... :-) Where did you get that information/screen shot from?

[Paragon]

http://paragon0.tripod.ca/ATI1.html


Uh.....huh... :-) Where did you get that information/screen shot from?


Well while you were lounging around in Mexico some of us were digging around to see what we could find. :D

Dave

[Timothy Rapson]

It may be hard to notice with the new X-Scales as they will change support chips, motherboards and more along with the CPU. I have been reading that the Toshiba e310 is running some stuff 3 times as fast as other PPC 2002 models. Better graphics support chips?

I think the reason for the slowness of PPCs doing regular stuff is that MS built in so many walls to keep programs from corrupting data when they mess up. They got beat up so bad on the first version of CE (and it was just awful. I never bought a first generation CE model because I could virtually never find one at a demo stand that wasn't locked up to try it.) so they really protected the data in PPC 2000-2002.

The 400 MZ Ipaq may really run Pocket IE like a desktop.

[seanturner]

Is the iPaq bus speed really 100Mhz? If so, that will be enough to accomodate anything you're gonna throw at it till you reach processor speeds of around 800Mhz. Unless you're doing VERY graphics intensive applications, I can't see you needing anything more. But I do have a hard time believeing the bus speed is 100Mhz. I had assumed it was like 33.

[sweetpete]

http://paragon0.tripod.ca/ATI1.html


Uh.....huh... :-) Where did you get that information/screen shot from?


Well while you were lounging around in Mexico some of us were digging around to see what we could find. :D

Dave


Hmmm ... that looks awefully similar to the specs I saw on Infosync for the new HP WDA 575 (less the graphics chip info). http://www.infosync.no/show.php?id=1752

This looks to be an awesome device. I do wonder though if the Jornada will survive the merger... :cry:

[Paragon]

[

This looks to be an awesome device. I do wonder though if the Jornada will survive the merger... :cry:


I think it is a little yearly to see decisions like that made. I would think they would like to get the two companies merged before they start publicly announcing product cuts. I have heard a rumor recently that HP will announce the new Jornadas on May 15th IF that is the case then they would have to have thousands of them on hand now, so I doubt they would throw them in the dumpster just yet.


Yes it is the specs for new HP 575 WDA.... so it seems

Dave

[entropy1980]

Here's a thread at PDABUZZ that talks about it
http://forums.pdabuzz.net/showthread.php?threadid=37600
looks legit enough....can't wait!

[charlie]

I think that the improvement with the biggest improvement in usable speed that someone could do for the pocket pc is better multitasking in the os. it's pretty bad, threads often lock up the whole machine exactly when I want to do some pim stuff, which makes my ipaq unusable. I usually reset instead of waiting for stupid apps to wake up. I've gotten rid of dashboard and pocket informant because they just don't run fast enough.

for hardware, I think offloading a lot of the IO tasks that the os handles to accessory chips (or circuits on the same die) would help out the cpu a lot. but since io differs so much from device to device, this might be hard to implement without some standards (like picking a platform for one). Although it might not seem like it today, 206 MHz is a LOT to work with, even for a RISC ISA.

charlie

[bbarker]

Timothy Rapson wrote:I think the reason for the slowness of PPCs doing regular stuff is that MS built in so many walls to keep programs from corrupting data when they mess up. They got beat up so bad on the first version of CE (and it was just awful. I never bought a first generation CE model because I could virtually never find one at a demo stand that wasn't locked up to try it.) so they really protected the data in PPC 2000-2002. I had two of the very first Windows CE devices, a Casio Cassiopeia A11 (with a keyboard) and a Cassiopeia E-10 (their first palm-sized model). Both were quick and reliable. I seldom had lockups and it worked great. My son still uses the E-10. The screen was a little dim and it only had 4MB RAM, but I liked it and never considered it "awful."

Even then I considered Windows CE superior to a Palm. I researched both of them when CE came out and chose CE. I never have agreed with people who keep writing about how terrible the pre-Pocket PC versions of CE were.

[seanturner]

Has anyone any done any benchmarking on the buss transfer speed on the iPaq?

Also, that IPCLK overclocking utility, does that adjust the cpu multiplier or the buss speed when it overclocks? Overclocking does speed my iPaq along a little, but, its not THAT drastic of an improvement going from 206 to 236ish Mhz. Back when I had a Palm 5, I overclocked it from 20 Mhz to 43 and that made a huge difference, but that is over a 100% increase. :D

Although, the XScale would probably be a drastic improvement, even at 400 Mhz, increasing the bus speed to anything above 100Mhz isn't going to make much of a difference.

[I. Bergman]

I think I've posted this before, but you can't simply compare the clock frequencies of a 206 MHz RISC processor (ARM, which we have in the PPC) to a 200 MHz CISC (X86, on the Desktop). My estimate is that a 206 MHz processor corresponds to about 40 or so MHz on the desktop. Now imagine you are runing W95 on a 486/40, I think this gives a pretty accurate picture of the speed you can expect on current PPC's.

[ChrisD]

Hi,
According to the specs on the Intel website, the clock speed of the bus is fixed at 100 mhz. So overclocking helps in some cases but not all.

Has anyone any done any benchmarking on the buss transfer speed on the iPaq?

Also, that IPCLK overclocking utility, does that adjust the cpu multiplier or the buss speed when it overclocks? Overclocking does speed my iPaq along a little, but, its not THAT drastic of an improvement going from 206 to 236ish Mhz. Back when I had a Palm 5, I overclocked it from 20 Mhz to 43 and that made a huge difference, but that is over a 100% increase. :D

Although, the XScale would probably be a drastic improvement, even at 400 Mhz, increasing the bus speed to anything above 100Mhz isn't going to make much of a difference.

[ChrisD]

Actually this is not a good way to compare performance of CISC and RISC processors! RISC processors execute a single instruction per cycle. CISC procesors can take as many as 10 cycles to execute a single instruction and the older the processor the larger the number of cycles it takes to execute one.

Is the iPaq bus speed really 100Mhz? If so, that will be enough to accomodate anything you're gonna throw at it till you reach processor speeds of around 800Mhz. Unless you're doing VERY graphics intensive applications, I can't see you needing anything more. But I do have a hard time believeing the bus speed is 100Mhz. I had assumed it was like 33.

[ChrisD]

If you want to know alot more about the ATI Imageon 100 see http://www.cewindows.net/reviews/ati-imageon.htm

I've seen it in action! It will make a difference in MPEG4 video playback that uses the hardware. Other than that, the OS has to take advantage of the 2D acceleration for graphics.


Just wait until that ATI mobile chip arrives in a Pocket PC...


AS in the ATI chip mentioned here http://paragon0.tripod.ca/ATI1.html
:) :)


Dave

[ChrisD]

Actually the way the Pocket PCs and Windows CE protects programs is using the ring approach. Each application is limited to 32 MB for execution and it can not access other applications data directly. All applications are address segment switched to the base addresss zero to execute. See http://www.cewindows.net/faqs/architecture.htm

It may be hard to notice with the new X-Scales as they will change support chips, motherboards and more along with the CPU. I have been reading that the Toshiba e310 is running some stuff 3 times as fast as other PPC 2002 models. Better graphics support chips?

I think the reason for the slowness of PPCs doing regular stuff is that MS built in so many walls to keep programs from corrupting data when they mess up. They got beat up so bad on the first version of CE (and it was just awful. I never bought a first generation CE model because I could virtually never find one at a demo stand that wasn't locked up to try it.) so they really protected the data in PPC 2000-2002.

The 400 MZ Ipaq may really run Pocket IE like a desktop.

[Foo Fighter]

I've always found it rather surprising that for a RAM-based computer, Pocket PCs are pretty slow.

Sacrilege!!! ;)

I have been quite tempted lately to go back to the PPC camp ever since the new Toshiba e310 rolled into town. But every time I play around with my friends iPaq 3835, I am immediately turned off by what I see. Despite all that impressive hardware...PocketPC offers a slow ride. Even the most basic app behaves sluggish at times. It's a stark contrast to the Palm environment where everything just opens instantly. Chris is right, the culprit seems to be the speed at which PPC loads code from ROM to RAM.

Still trying to decide whether to jump ship or not. The e310 has me lusting for the hardware. But the software comes up very short.

[Jason Dunn]

PocketPC offers a slow ride. Even the most basic app behaves sluggish at times. It's a stark contrast to the Palm environment where everything just opens instantly.


Remember how fast DOS was? Uni-tasking operating systems are FAST because they don't juggle multiple processes. There's a REASON that Palm is so fast. :wink:

[Foo Fighter]

Uni-tasking operating systems are FAST because they don't juggle multiple processes. There's a REASON that Palm is so fast. :wink:

But remember, Jason. Palm has the highly advanced ZEN kernel powering their OS. ;)

[seanturner]

I woudln't even start to compare it to a desktop processor because its an entirely different core and OS, although it would be interesting to see how it would benchmark... But at around 1Ghz, the Athlon, which is a psudo RISC processor starts to get memory bandwidth issues. But, my basic point was that I don't see the need for a lot of memory bandwidth in the applications I would be running on a PPC.

Oh, does anyone have any info on who created the iPaq overclocking utility? It would be interesting to see whether it is a bus or core overclock...

[Take1]

I too would like to see a direct comparison to the actual processing power of a 206MHz StrongARM processor compared to, say, a K-6 II or Pentium II chip of equivalent clock speed. Onboard cache and front side buses permit desktop chips to run smoothly, but I don't know if StrongARM chips use this kind of technology or if they simply run 'straight'. I find myself thinking of PPC processors in terms of desktop chips (erroneously) and also find the performance surprisingly slow for the MHz.

[seanturner]

I could be just a rambling idiot, as illustrated by my ineptness when it comes to spelling, but, I believe I once read that the StrongARM processors have either no or minimal cache.... Hmn... reminds me of the Intel Celeron... lol

[seanturner]

Oh, I took a look at the website of the guy that makes the iPaq overclocking utility and it said nothing of use there. I emailed the company and havn't gotten a responce.