Dell Axim X50v with Intel2700G

[hrishi2das]

one very important app (to me ) that I use is a emulator called Morphgear, basically it emulates GBC, GBA games. I think this software would run very slow because of the non-acceleration. Hopefully they will rewrite the software, but that does not at all look promising . this is really frustrating, having such a good hardware and not even passing the performance of older generation pocket pcs.

Cant a wrapper or something be written ?

 

[deviantchild]

another review stating somewhat iffy graphics performance:-

From a performance point of view, the Dell Axim X50v was somewhat disappointing, as one of the PDA's best highlights--the VGA screen--turned out to be its biggest downfall. That said, we noticed similar hiccups in other VGA models such as the Asus MyPal A730 and the Toshiba e805, and while not ideal as your primary media player or gaming device, the X50v is still a very capable PDA.

http://reviews.cnet.com/Dell_Axim_X50v/4505-3127_7-31138229-2.html?tag=top

their test suite:-
http://reviews.cnet.com/4520-6603_7-5110068-1.html?tag=txt

 

[glascwmn]

Have a look at this for an explanation. Shouldn't the PowerVR people be highlighting this ?

ftp://download.intel.com/design/pca/companion/applnots/graph_bnchmk.pdf

 

[deviantchild]

i didn't need the explanation, but thx, that's a great doc to have locked-&-loaded
i was merely showing what's out there unattested on popular review sites
who will set the record straight?... PowerVR will inform you that it's Intel's job; it's Intel's product
but, if it didn't sell well, who would get dropped?
anyway, seeing as that is Intel's PDF they seem to have predicted this so... no worries, i guess

 

[Kristof]

I think I posted the link to the Intel document a while ago. Anyway most of the "issues" seem to be with the SPB benchmark score... but lets have a look at it :

http://www.aximsite.com/articles/link.php?id=200

Graphics Index X50v = 487.6
DIB = 22.94 fps
DDB = 260.8
GAPI = 45.57

Graphics Index Asus A730 = 1015.4
DIB = 8.442
DDB = 139.9
GAPI = 146.6

Factors X50v in relation to A730 :

Graphics Index = 0.48
DIB = 2.72
DDB = 1.86
GAPI = 0.31

So in summary the Index says the x50v is overall 48% of the A730, yet the detail scores show that in DIB its 2.72 times faster and in DDB its 1.86 times faster while in GAP its running at about 1/3th the speed... now do those numbers warant an overall index of 50% ? IMHO a bit of a weird scoring system. According to the SPB website they think that DIB and GAPI warrant 40% weights and DDB a score of 20%... so if we take the usual formula this would translate to :

X50v : 40 x 260.8 + 20 x 22.94 + 40 x 45.57 = 12713
A730 : 40 x 139.9 + 20 x 8.442 + 40 x 146.6 = 11628

I guess it all depends on how you look at the numbers... <shrug>
Reminds me of my maths teacher : you have small lies, big lies and then you have statistics...

Just my personal 2c...

K-

 

[Loewe]

@Kristof

Interesting math but say this the guys who made the reviews.
The potential buyers don't read this forum, they read the reviews and we all know the numbers from there.

 

[Luminescent]

Gameplay vids of the X50v w/2700G in action available towards the bottom of this page: http://pocketnow.com/index.php?a=portal_detail&t=reviews&id=495&p=2

 

[Roger Kohli]

Those movies are great. Stuntcar Extreme reminds me a bit of Ultim@te Race, especially with the popup (?) in the distance. Enigmo looks like it could be fun to play too.

 

[Tala]

After discussion on PPC related websites i came to the conclusion that the actual GAPI implementation of the X50v is flawed, as they most likely use the same implementation as for PPCs without hardware acceleration and frambuffers allocated in system memory.
It woks as follows:

GAPI is inherently an API for QVGA devices only. Thus QVGA backuffer is allocated in system memory, which is presented to the user as frame buffer. As far as good. However with a call to GXEndDraw() the CPU reads the backbuffer and scales it to VGA, while writing the scaled content to the 2700Gs framebuffer. A much! better implementation would be, when the 2700g can read the QVGA system memory buffer per DMA and use the scaling hardware to write into its VGA frame buffer. Not only does this use only 1/4 of the bandwidth but it can happen asynchronously to the CPU. The next call to GXBeginDraw() should synchronize with the video hardware again (ie. it is blocking until the GPU is finished).
This way you would have minimal overhead compared to QVGA devices without GPU.

While it is easy to develop a better GAPI implementation as described above, because it is only a DLL, we have no chance in doing so without a proper documentation how to access the 2700Gs device registers using the driver from user mode.

Does anyone know who can provide such information? Is it Imagination Technologies, Intel or Dell? I would gladly look into the issue and try to improve GAPI when i get access to the information required.

Tala

 

[Luminescent]

Intel has a good set of detailed documentation on 2700g. Just do a search on the site for 2700g and search through the whitepapers.

 

[Tala]

Intel has a good set of detailed documentation on 2700g. Just do a search on the site for 2700g and search through the whitepapers.

I read all the whitepapers available for public access. They do not tell how to program the actual device.

Tala

 

[Loewe]

@Tala

I think its imgtec.
That here is the right forum to get contact with the guys. Kristof will read it, I am sure and if you don't get an answer there is no problem to bring you both in contact.