(keitai-l) Re: Mobile 3D Users & Future Trends

From: Paul Beardow <PBeardow_at_superscape.com>
Date: 04/26/04
Message-ID: <9F79740E2C1322438FB75E88D42113D6012BAF5C@hook-exch.superscape.com>
Hi,
I actually think the secret is a well-balanced system. A faster processor
can only do it's job properly if it can get data and instructions from
memory fast enough, the same applies to graphics processors, so the key is
good system analysis to make sure when a real app runs, data passes around
effectively. 

One problem I think we'll face is that graphics benchmarks often focus on
back end rendering performance, rather than completed games, so they don't
give a full picture of how well a handset can play games. A benchmark can be
arranged to fit in the processor cache, or be set up so that texture RAM
size isn't exceeded, so you aren't getting a true measure of overall
performance. In a real game, memory is being accessed all over the place and
textures often need swapping in and out and you don't have fine control over
memory when writing in Java.

As for memory and processing speed, a good processor cache size helps a lot
- as does burst mode if you are using PSRAM. As for power consumption, I
still think it's LCD backlights and RF amplifiers that run down the battery
fastest. 

Fast processors always help with gaming - all of the devices coming out from
the people we work with are 100Mhz+ ARM9 and several a lot faster, and of
course heavily optimised code is a must! Battery life seems fine on these
devices too - not up to Japanese levels, but Western phones tend to use
different battery technology and the network RF code is far more complex, so
consumption is higher.

One final point - M3G MIDlets don't run all the time, paints are done on
demand, so if the scene isn't changing, you don't need to update the screen.
There's no additional threads running all the time either.

regards

Paul
Received on Mon Apr 26 13:26:21 2004