I've posted originally this message on the Dell support forum, but I believe this forum is a better place for such kind of questions.
I downloaded a game that was said to run at 50-60 fps on a handheld similar to Axim X5 400MHz that I own, with PPC 2003 A04 preinstalled. Now I run it and see 20-22 fps instead of the expected 50-60.
I got and ran VOBenchmark and the interesting part is that while it shows most numbers as listed in most Web reviews of Dell Axim X5, the video performance is approximatively twice smaller. For example, at http://www.pdabuyersguide.com/dell_axim_x5.htm the "Graphics Bitmap BitBlt" performance listed (measured with the same version of VOBenchmark that I have tried) is 56.60, while I get ~24. Most other numbers are similar to those listed there (most are actually a little bit higher because I believe they used a PXA250 and my palmtop has a PXA255), but everything related to video performance is a LOT slower. (interesting that in the review above iPAQ 3970 also shows a very poor video performance - just a bit faster than my Dell).
Could this be somehow related to the "poor PPC 2003 performace" that is often mentioned in all Axim forums? Can someone with PPC 2003 (how about those lucky people with the A05 ROM?) and with 2002 run the benchmark so that we can compare the results here?
To make a start, here are my results (just most important and the ones that are very different from those listed on the site above):
CPU: FP 12.38/INT 26.40
Graphics: BitBlt 24.36, StretchBlt 0.30
Filled Rectangle: 3.12
Text: 1.61 (maybe because I have enabled text antialiasing).
What's interesting is that most dramatic differences are seen in tests that involves intense video memory access (MediaQ chip problem?). Filled ellipses are just a little slower (just because they are more a matter of CPU than video memory access) than they should be and filled rectangles are twice slower (because CPU is almost not involved here, just plain storing of pixels into video memory).
Ok, so there are really some weird things with my palmtop.
I got a newer VOBenchmark 3.0 (and hacked my first ARM program in about 10 minutes to avoid installing those ads :-\) and also changed the CPU setting to "always 400MHz" (it was Auto in my previous test). The results, however, stayed almost the same (actually the BitBlt test was a little quicker - 28.23 parrots). Here are updated values for my palmtop with VOBenchmark 3.0:
CPU: FP 12.70/INT 26.94
Graphics: BitBlt 28.23, StretchGrow 8.10, Shrink 5.30
Filled: rectangle 4.90; ellipse 1.88; rounded 1.60
GAPI: Max FPS 47.528; Max Sprites 116
Lines: 35.40
Text: 3.50 (tried with and without ClearText - no difference)
Memory: allocation 11.33; fill 2.0; move 1.34
So here are some weird things that are visible at the first glance:
a) CPU and memory on our palmtops performs equally fast (so all claims that A04 has slow MEMORY performance are absolutely WRONG!)
b) GAPI performs equally fast on both palmtops.
c) BitBlt is almost twice slower
d) Filled things are twice slower
e) Drawing lines is FOUR TIMES slower
f) Drawing text is FIVE TIMES slower
wow, what a difference! Now could someone with PPC 2003 A04 post his results here, so that I can see I have a broken OS (surprise! :-|) and not a broken palmtop!
Actually I figured out that to disable ClearType I had to soft reset.
Well, this haven't changed VOBenchmark numbers except the "Text" test which is now 11.00 - anyways almost twice smaller than it can be.
__________________
Chris Leckness, Microsoft MVP Windows Mobile To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts. | To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts. | To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts. | To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts. | To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts. | To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts. | To view links or images in signatures your post count must be 10 or greater. You currently have 0 posts.
This may be OT but I am a fairly new Axim owner. Looking at the readouts, I notice the all of the Axim Advanced 400mhz processors are the PXA255 with Bios A05.
I have a brand new Axim basic, and it reads out as PXA255, 300mhz, Bios A05.
is PXA255 just the processor family? It strikes me as strange that the processors would have the same designation but different speeds.
pxa250 is the old one whereas pxa255 is the new one. Intel stopped producing pxa250 back in march03. Fastest speed for both is 400mhz, just that the pxa250 has a 100mhz system bus and runs at 1.5v whereas the pxa255 has a 200mhz system bus and runs at 1.3v
I don't know really. I was thinking of this myself when I was going to buy my Axim... however, the memory upgrade itself was enough to incline me towards Advanced model :)
It also can be that Dell enforced in hardware the lower performance of Basic models. For example, they could have a lower core voltage, a slower memory or such.
So the things are going stranger and stranger! Actually some tests show on PPC2002 lower results than PPC2003 A05! So 2003 is definitely worth the upgrade, but not to the A04 ROM :)
To summarize, here in percents are shown the results relative to ROM A05 of ROMs A03 (first number) and A04 (second number):
Ok, so you got the idea. This all looks fascinating! It looks like those M$ guys can't write even a optimized rectangle filling routine :-( Like they could ever...
So the A05 ROM definitely looks like it's best of the all ROMs (I'm supposing all Axim Advanced models have similar performing hardware).Oh well, it looks like it's time to set up a cross-compiler and help those guys at www.handhelds.org to finish the Axim port of Linux...
This night I tried to run the damn Dell update yet another time... and guess what... IT WORKED!!!
So the first thing I did after boot is running the benchmark. And it has shown the correct numbers, more or less like in Takenover83's post. The game I started my quest with runs at 30fps. And in general, the feeling is indeed that response time has substantially decreased - for example, with A04 it was a noticeable amount of time after system boot when PPC started to "feel" touches. With A05 this interval become almost unnoticeable.
Ok, thanks everyone who read all this crap... will go sleep now...