Toshiba e740 Pocket PC 2002 benchmarked

[Jason Dunn]

<a href="http://www.pdabuyersguide.com/toshiba_e740.htm">http://www.pdabuyersguide.com/toshiba_e740.htm</a><br /><br />PDABuyersGuide takes the e740 for a spin and does some benchmarking on it - the results, as expected, are mixed. It seems faster in many areas, but can't play video to save its life. Go have a look!<br /><br />"It seems that the e740 does not run noteably faster than StrongARM 206 MHz processor Pocket PCs such as the iPAQ 3800 series. It most certainly doesn't run twice as fast, as one might first expect given that the XScale runs at 400 MHz and the StrongARM at 206 MHz. Both pocketnow.com and pocketmatrix have posted benchmarks comparing performance of the iPAQ 38xx vs. e740 (and other Pocket PCs using the StrongARM processor). We've run benchmarks using VOBenchmark from Virtual Office Systems and we've recorded the dropped frame rate in some popular movie files using Pocket DivX and have to say that the XScale is not wildly faster than the StrongARM (see Benchmark tables below)."

[Duncan]

Preliminary tests with the Loox are showing a dropped frame rate of 18 for that Matrix Reloaded trailer - playing from a CF card on battery power and with the cache disabled. With the cache enabled it is dropping zero frames. I don't know whether they enabled the cache or not.

[fastcar]

Wow, that LOOX sounds great!
How do you enable/disable the cache? Why would Toshiba (or any sane company) ship it with the cache disabled?

Preliminary tests with the Loox are showing a dropped frame rate of 18 for that Matrix Reloaded trailer - playing from a CF card on battery power and with the cache disabled. With the cache enabled it is dropping zero frames. I don't know whether they enabled the cache or not.

[kaiden.1]

With all the tests that have been done, the x-scale is still doing remarkably well. I think that the problem here is expectations! Lisa did say also in her report that even though there was frame dropping, she didn't even notice it, and she has professional experience dealing with that very thing. Over all, I think that x-scale is a great addition and that we are all to critical, as time moves on, and developers tweak in the necessary adjustments within their software for x-scale, I would bet in Vegas that we would all see that all our worry and FEAR was really just silly feelings over hyped up expectations!

I have a Toshiba E-740 and it still works and out does everything that I used to do PLUS a lot more. This is one great PPC, and I would still recommend it to anyone!

[Duncan]

fastcar,

Enabling or disabling the cache is something that you do with the DivX program - it isn't something that Toshiba did. I don't know whether the reviewers enabled the cache or not - thus I tested both ways.

kaiden.1 is right - the dropped frames count does seem to suggest that the Loox is performing better BUT the number of dropped frames mentioned is not going to be very noticeable to most people.

[pdagal]

For the DivX tests, the cache option in DivX was turned off for both the e740 and iPAQ 3835.

regards,
Lisa

[farnold]

As much as I like the idea of comparing XScale and StrongArm 206... right now I get the feeling that everybody blames XScale just for the performance the e740 shows.
I understand that not-optimized software is a problem, but I want to see how it looks like if you compare IPAQ 39xx, TOSHIBA e740, LOOX and others.
I wouldn't be surprised to see some major differences there... TOSHIBA was faster, but maybe others deliver a better adoptation

[pdagal]

We'll be getting an iPAQ 3970 tomorrow, and (as long as it actually arrives!) the first thing I will do after I receive it is do the same benchmark tests on it. I'll post these in the e740 benchmark table in the review referenced in this thread ( http://www.pdabuyersguide.com/toshiba_e740.htm) . So you should see this info on Weds. July 10th. In the following days, an iPAQ 3970 will follow.

My educated guess is that the numbers in VO Office will be similar, but the iPAQ will do better in multimedia playback than the e740 (the DivX test). I think the SD card speed of the 3970 will be better than the 3800 because I have an unconfirmed suspicion that Compaq used a 4 bit bus rather than the 3800 series 1 bit bus. This would put it on par with the Toshiba e740 SD slot. In his review at pocketpcpassion.com, Dale Coffing found that a pre-release 3970 had much better battery life than many other Pocket PCs. Can't wait to test that too!

I also can't wait to take an in person look at the LOOX!

As much as I like the idea of comparing XScale and
StrongArm 206... right now I get the feeling that everybody blames XScale just for the performance the e740 shows.
I understand that not-optimized software is a problem, but I want to see how it looks like if you compare IPAQ 39xx, TOSHIBA e740, LOOX and others.
I wouldn't be surprised to see some major differences there... TOSHIBA was faster, but maybe others deliver a better adoptation

[Jimmy Dodd]

...we'll be testing to see if that means better games performance, though given the sub-par movie player performance, it's doubtful.

Using Pocket DivX, an excellent and popular multimedia player that handles MPEGs, MP3s and AVI files, the iPAQ drops almost no frames playing the popular high quality movies listed above. The e740 drops a shameful number of frames, though it's interesting that I couldn't perceive this when watching the movies (and I used to work for Avid, a computer that makes video post-production systems, so I usually notice these things!).



OK, someone explain this to me. First, the e740 is described as having "sub-par movie player performance" and as dropping a "shameful number of frames." That sounds really bad. But then she says "I couldn't perceive this when watching the movies." So are we just splitting hairs here? Is there a perceivable difference between watching a movie on an e740 and the iPaq? Is the reported speed problem only detectable if you write an app to go look for the problem?

I am seriously considering an e740 soon, but the in store models don't have any apps installed that really tax the system to do a serious comparison. I'd like to see some new benchmarks in the future that focus on end-user perception. Instead of breaking down the benchmarks into memory move speed and ellipse drawing give me something I (and other users) can use to compare.

And before anyone tries to explain the purpose of these completely objective benchmarks to me let me say that as a professional software developer I use benchmarks like these everyday while testing and tweaking my code. But when it comes time to show the user what we've accomplished I have to put it in terms of their day-to-day use (the app loads x times faster, the task has been speeded up x ms, etc.).

Thanks,
Bwana Jim

[Dr. Smooth]

I am seriously considering an e740 soon, but the in store models don't have any apps installed that really tax the system to do a serious comparison. I'd like to see some new benchmarks in the future that focus on end-user perception. Instead of breaking down the benchmarks into memory move speed and ellipse drawing give me something I (and other users) can use to compare.

And before anyone tries to explain the purpose of these completely objective benchmarks to me let me say that as a professional software developer I use benchmarks like these everyday while testing and tweaking my code. But when it comes time to show the user what we've accomplished I have to put it in terms of their day-to-day use (the app loads x times faster, the task has been speeded up x ms, etc.).


As a developer of PocketPC apps, I will tell you that there is no perceivable speed difference between the e740 and any of the ARM-based PocketPCs. Our app does a number of processor- and memory-intensive tasks like running handwriting recognition engines, parsing and writing large (200K+) XML files, compressing and MIME-encoding large chunks of data. We have informally run races between the e740 and the Jornada 568 and various iPaqs. No perceivable difference.

As for apps opening and closing and general responsiveness of the interface, I can tell you it feels just like an ARM-based device. From what I understand, for the Xscale to shine, you need optimized code. None of the resident WinCE/PocketPC code is optimized for Xscale, and eVC++ can't compile for Xscale. The best you can do is use some libraries provided by Intel for some common multimedia operations. Those libraries are optimized and may provide some speedup. But as far as optimizing your app's entire codebase, it can't be done yet.

What worries me the most is that when WinCE .NET comes out, everything will be running through an additional layer of an interpreter. If that causes a performance hit, it may offset gains from the Xscale. It may be a long time before we see PocketPCs that exhibit the kind of speedup everybody longs for.

[Jimmy Dodd]

Thanks for the info, Dr. Smooth. Thats about what I expected.

As for CE.NET I've been running some apps with the beta .NET CF and it seems pretty speedy. Most of my work has been with MFC lately and the speed is comparable. Still definitely faster than eVB. Maybe not as quick as straight Win32 though. Bottom line is I haven't noticed a decrease in my test apps. Of course I don't have any benchmarks like memory move rates and frames dropped, but I am confident that my customers won't be asking "what the heck happened to the performance?!?!" when I deliver.

---
Bwana Jim

[pdagal]

Our iPAQ 3970 has arrived, and I'ved added it to the benchmark tables in the review ( http://www.pdabuyersguide.com/toshiba_e740.htm )

I can't add DivX benchmarks, as DivX won't run correctly on the 3970. I will be adding Pocket TV comparisons later today though.