Author |
Topic |
|
ctacke
877 Posts |
Posted - 20 Sep 2002 : 17:10:54
|
Q: What kind of video playback rates will Eurotech systems provide?
There are many influencing factors to video playback with any system including color depth, media size, media format, production framerate and sound. Additionally since the StrongARM is its own video processor, the addition of hardware like a graphics accelerator will also have a marked effect.
We have tested some of our systems, both with and without accelerators, using a variety of formats, rates, etc. The media used for testing and the test results are available here.
Keep checking back as we will continue to test systems and add results to this page.
----------------- Chris Tacke, eMVP Applied Data Support |
|
ctacke
877 Posts |
Posted - 25 Apr 2003 : 11:45:31
|
Additional Testing
We've done some additonal testing using the BitsyX with a PXA255 processor at both 200 and 400MHz. The full result set is available from the link above, but here are the highlights:
Video | Format | SA1110, 206MHz | PXA255, 200MHz | PXA255, 400MHz | kevad | MPEG 130% QVGA | 5.9fps | 4.1fps | 7.5fps | kevad | AVI 130% QVGA | 17.8fps | 21fps | 21fps | humanbody | AVI QVGA | 24.5fps | 29.6fps | 29.8fps |
The PXA255 system running at 400MHz played MPEGs 27% faster and AVIs 22% faster than the SA1110 system. |
|
|
ctacke
877 Posts |
Posted - 03 Jun 2003 : 18:06:39
|
Additional testing We've done more testing with the BitsyX using a SIF (QVGA) sized display and performance is great.
Using a 2:31 MPEG1 file playing from a CF card we achieved a 98% efficiency (playing a 24fps video at 23.8fps). This video has full sound and is higly compressed but the BitsyX played it with almost no perceptible video flakiness or trouble.
Files we used for test include the following: |
|
|
ctacke
877 Posts |
Posted - 04 Jun 2003 : 10:13:05
|
Additional Information Video playback performance can be somewhat confusing as there are several factors that affect the end result that you see on the display.
Key factors to consider are:
- Media Format - AVIs play much better than MPEGs because they are typically uncompressed. This means the processor can dedicate cycles to playing rather than decompressing and then playing. To add to this, the media itself has an impact. Lots of motion and color change decreases performance, as does sound.
- Media Dimensions - The larger the media (height by width), the more data that must be pushed across the bus and the slower it will perform
- Display Dimensions - A larger (i.e. higher resolution) display will perform slower than a small display, even with the same media. This is because the framebuffer itself is larger and must be refreshed.
- Display Color Depth - Greater color depth means more data per pixel shown, which in turn means more data across the bus and slower performance.
- Media Location - Media performs best when played from RAM (as opposed to on the FlashFX Disk or on a CF card) because it doesn't have to travel across a bus or through an additional driver.
- Production Framerate of the Media - The video being played will be recorded at a certain framerate (common values are 20, 24, 30 and 36 fps. The human eye can discern about 26 fps). If a video is recorded at a framerate above the capabilities of what the device can deliver, frames are dropped. This results in not only "stutter" in playback, but it also burns CPU cycles. You will get the best playback results from media recorded at or near the device's capabilities than the same media recorded at a rate higher than the devices playback capabilities.
For example, in the above two posts you'll see that playing "kevad" is quite a bit slower than playing the "Matrix2" trailer on the BitsyX. The primary reasons for this are: - The Kevad video was tested on a VGA (640x480) display while the Matrix2 trailer was tested on a SIF (320x240) display
- The kevad video is substantially larger (352x288 = 101,376 pixels) in output dimensions than the Matrix2 trailer (320x176 = 56,320 pixels)
For more information see the following Support Forum Topics: StrongARM display bandwidth limitations explained Performance hit for 16 bpp color
|
|
|
|
Topic |
|
|
|