Asus a730: Reports of the Death of Its Performance Have Been Greatly Exaggerated

[Jonathon Watkins]

<div class='os_post_top_link'><a href='http://www.pocketmatrix.com/forums/viewtopic.php?t=17647' target='_blank'>http://www.pocketmatrix.com/forums/...pic.php?t=17647</a><br /><br /></div><i>"I received the a730 a few weeks back and I immediately tested out the graphics and benchmarks because I, like the rest of you, have heard rumors of it being slow. My initial findings were looking grim. It was performing slow, very slow, and it didn't quite make sense to me. So I got talking with Dan East of Pocket Quake, as well as Picard of Beta Player. With the help of these two developers, we found the main cause for the Asus a730's slow speeds. It had to do with a poorly optimized dll which converts the 320x240 code to display on the a730's VGA screen (pixel doubling). Picard had sent me a dll to try on the a730 which had some better optimized formulas for the conversion. After benchmarking in quake and spb before and after this new dll, I seen significant gains (roughly 5 fps in quake)"</i><br /><br />Matt Keys over at PocketMatrix has been playing with a Asus a730 "final production unit" for a few weeks now and has seen problems with it's display speed. Happily Asus responded immediately after being sent details of the fix and the current a730 performance is significantly improved. Matt is withholding his final judgement until he sees a final "final production unit" :wink: , but so far he is impressed with the device and says that it has a lot of potential. Problems with complex devices like PPCs are bound to occur and it is really good to see Asus being so responsive and on the ball. Now if only the other PPC manufactures were willing to react to user feedback (let alone quickly). Oh well, we can hope, can’t we?

[mscdex]

Matt Keys over at PocketMatrix has been playing with a Asus a370 "final production unit" ...

Asus a370? Is that the low end model?

[Jonathon Watkins]

Matt Keys over at PocketMatrix has been playing with a Asus a370 "final production unit" ...

Asus a370? Is that the low end model?

Darn Dyslexia. :oops: Corrected. :) It was of course, the unreleased a730

[whydidnt]

While I commend Asus for fixing this when brought to their attention, it is a bit concerning that:

A) Their own quality control couldn't figure out that there was an issue.
and
B) A third party developer had to point them in the right direction to fix it.
and
C) WMSE support for 640 x 480 is so complicated for hardware vendors to implement.

[jkendrick]

SE VGA support is so complicated because OEMs and software developers now have to make sure their s/w works in QVGA, VGA, portrait, and landscape. This greatly complicates development and testing, as several developers have complained bitterly to me.

This problem must be the reason that Asus as announced the A730 will be delayed until later this year.

[whydidnt]

SE VGA support is so complicated because OEMs and software developers now have to make sure their s/w works in QVGA, VGA, portrait, and landscape. This greatly complicates development and testing, as several developers have complained bitterly to me.


I thought that this was the reason that MS decided to make it "easy" to just double the pixels. What is more complicated about this than for a normal PC? Nearly every application you run on your standard windows machine provides this capability. And even so, what does that have to with ASUS' "final production" machine not having optimized VGA drivers? Don't the core OS apps that come from MS already have the VGA funcionality built in?

Is there something inherently broken or wrong about how MS decided to implement VGA on the PPC?

[jkendrick]

Programs can be easily pixel doubled at the OS level but that doesn't provide high resolution screens, which is what VGA owners expect. Throw in that apps that aren't SE optimized often have dialog boxes that display off the screen when rotated to landscape orientation and thus rendered useless.

We're seeing a transitional period between FE and SE and VGA further complicates the switch.

The OS itself is optimized for VGA and landscape but even PIE displays images pixel doubled and not true VGA.

[The PocketTV Team]

> The OS itself is optimized for VGA and landscape but even PIE displays images pixel doubled and not true VGA.

That's a bug of PIE.

Or rather, MSFT has not yet implemented correctly a VGA-enabled version of PIE that displays images in VGA res. The current version of PIE is basically only drawing text with VGA res, i.e. PIE is is not yet hi-res aware (or if it is, it is not taking full advantage of the hi-res when available).

This problem with PIE currently exists on all VGA devices, not just the Asus.

[The PocketTV Team]

SE VGA support is so complicated because OEMs and software developers now have to make sure their s/w works in QVGA, VGA, portrait, and landscape. This greatly complicates development and testing, as several developers have complained bitterly to me.
Oh yes, it does!

Not to mention even more complexity when you want to take advantage of specific hardware and/or features only available in particular versions of the OS :)

[The PocketTV Team]

Regarding game performances and pixel doubling, there is two approaches:

1) use GAPI, in which case the GAPI emulation layer will do the pixel doubling for you, presumably using optimized code (hopefully using Wireless MMX instructions).

2) Use GETRAWFRAMEBUFFER (and not use GAPI for display) and do the pixel-doubling yourself. This is probably better/faster if you want to update only small areas of the display (sprites), since the GAPI emulation can only pixel-double either the entire screen or a horizontal stripe of it.

3) If available, use specific hardware features of the device. One example is to use hardware scaling of surfaces on devices that have the ATI Imageaon 3200, e.g. the Toshiba e800. This is more complicated, but can result in an enormous performance gain.