Nvidia's Ion: A Video Evaluation
Continued from ‘Nvidia’s Ion: A Hands-On Try-On‘…
Because Nvidia began as a standalone graphics technology supplier before expanding into today’s graphics-augmented core logic chipsets and other products, reviews of Nvidia-inclusive products tend to be heavy on gaming and other 3D-centric applications. I’ll leave such tasks to others who’ll inevitably be publishing their writeups around the same time mine appears, coincident with the lifting of Nvidia’s embargo. For one thing, I question the practicality of any sort of meaningful gaming experience on a 12" or smaller LCD, although I admit that the ability to run Windows Vista with all of its eye candy options enabled is a plus, and I also realize that customer, competitor and/or governmental pressure may compel Intel to allow Atom’s use in more mainstream system form factors going forward.
I also believe that at least one other key application, video playback, begs for review attention and will likely be overlooked (or at minimum under-explored) by others’ coverage. Video decoding, trans-coding and other processing tasks handled in part (if not completely) by the graphics subsystem can notably offload the CPU. This function re-balancing act is critical with diminutive microprocessors such as Atom and especially so with battery-powered systems, which greatly benefit from more energy-efficient hardware-accelerated processing versus the software-centric traditional approach. Native support in the graphics subsystem for latest-generation video codecs such as H.264 (aka MPEG-4 AVC, MPEG-4 JVT or MPEG-4 Part 10) and VC-1 (a derivative of Windows Media Video 9) is beneficial both for playback of physical media such as Blu-ray Disc and for online-accessed content coming from sources such as Amazon Unbox and Netflix Watch Instantly (VC-1), DivX (MPEG-4 Advanced Simple Profile), and iTunes YouTube and high-definition Hulu (H.264).
As such, I focused my attention on two MPEG video clips Nvidia supplied with the review hardware, further augmenting the test suite with a WMV9-encoded clip already in my possession:
|
Title |
The Dark Knight (trailer) |
Plush Life |
Industrial |
|
Length |
2:29 |
1:50 |
0:40 |
|
File size |
192 MBytes |
183 MBytes |
3.43 MBytes |
|
Video codec |
H.264 |
MPEG-2 |
WMV9 |
|
Frame rate |
30 fps |
30 fps |
24 fps |
|
Resolution |
1080p |
1280×720 |
640×352 |
|
Peak video bitrate |
<30 Mbps |
>20 Mbps |
550 Kbps (CBR) |
|
Average video bitrate |
~12 Mbps |
~10 Mbps |
550 Kbps (CBR) |
|
Audio codec |
AAC |
MPEG |
WMA9 |
|
#Audio channels |
5.1 |
2.0 |
5.1 |
|
Peak audio bitrate |
<350 Kbps |
224 Kbps (CBR) |
192 Kbps (CBR) |
|
Average audio bitrate |
~250 Kbps |
224 Kbps (CBR) |
192 Kbps (CBR) |
Here’s the version of CyberLink’s PowerDVD v8 that Nvidia provided with the Ion review hardware:
Along with key video processing parameters for the program:
When I initiated playback of a clip, the following warning screen would pop up:
I chose to leave the settings as-is, as a worst-case (from a processing burden perspective…best-case from a quality standpoint) test approach.
Below you’ll find the Windows Vista Task Manager CPU load measurement during playback of The Dark Knight.
Disregard the blips at the beginning of the clip (a reflection of PowerDVD start-up) and end (caused by my activation of Windows Vista’s Snipping screen capture tool); instead, focus your attention on the average load throughout the playback interval. You’ll likely be impressed, as I was, by the degree that the Nvidia 9400M off-loaded the single-core Atom CPU, thereby delivering a robust experience even on a system with fairly minuscule microprocessor resources. I occasionally noticed ‘tearing’ and other image artifacts, which I suspect are the result of initial software driver imperfections versus some fundamental hardware flaw. Glitches aside, the notable potential of the 9400M was quite evident.
Next, let’s look at Plush Life playback on the Nvidia Ion system, using PowerDVD:
Again, the degree to which the 9400M shouldered MPEG-2 video processing should be obvious to your eyes. Finally, take a look at what happened when I played back Industrial in PowerDVD:
This one was a surprise. Even though the WMV-encoded clip was sub-VGA in resolution (versus its high-definition predecessors), the CPU load this time was notably higher than with the earlier MPEG-2 and H.264 tests. I repeated the test using Windows Media Player 11, in case PowerDVD was deficient in its Windows Media codec support, but got similar results:
Nvidia’s hardware-accelerated an increasing percentage of the Windows Media Video processing pipeline through several generations of graphics technology spanning many years, so I suspect that this shortcoming simply reflects a 9400M initial-development focus on MPEG instead of Windows Media. Considering that Apple was Nvidia’s lead customer on the 9400M, and that Windows Media is a rarely-encountered codec with OS X, this MPEG prioritization would make sense. I’ll run these results by my Nvidia contacts and let you know via a follow-up post if I hear back anything interesting.
For comparative purposes, I also decided to play back the three clips on my Windows XP-based MSI Wind U100 netbook. Note that I haven’t yet boosted the system’s memory to 2 GBytes; it still contains the factory-default 1 GByte DDR2-533 SDRAM allocation. However, its Atom N270 CPU is for all intents and purposes identical to the Atom 230 in the Nvidia Ion reference platform; the Atom N270 has lower power consumption specifications but identical features and clock speeds, therefore delivering comparable performance. Here are the specific PowerDVD video processing settings I used:
And the pop-up (which I again ignored) that appeared when I initiated playback:
Unfortunately, the OEM build of PowerDVD v7 that comes with the MSI Wind U100 is incapable of being codec-augmented with the HD-264 Pack, and I haven’t yet upgraded this particular system to PowerDVD v8 Deluxe or Ultra. As such, I wasn’t able to try out the trailer for The Dark Knight, although rest assured that the lack of H.264 codec support in Intel’s chipset would have severely loaded down the CPU, at minimum making the system unusable for other simultaneous systems and likely also leading to frame drops and other undesirable outcomes. However, look what happened when I played back the MPEG-2-encoded Plush Life in both PowerDVD:
and Windows Media Player:
Intel’s chipset does have native MPEG-2 acceleration support, as has been the case for many years now, driven by the longstanding popularity of DVD. As such, the results are comparable to those earlier found with the 9400M. And when I fired up WMV9-encoded Industrial in both PowerDVD:
and Windows Media Player:
I again got 9400M-reminiscent results.
So is Intel’s Atom-paired chipset deficient from a video processing standpoint, compared to Nvidia’s upstart alternative? The answer to that question depends on your codec current usage and future adoption rate assumptions. Certainly, I strongly suspect, Nvidia’s in the lead right now with respect to H.264, a codec that’ll become increasingly popular over time courtesy of (for example) iTunes and various Adobe Flash 9-based services. But with MPEG-2 and WMV9/VC-1, which both chipsets support, the race is essentially a dead heat. Such is also the case with codecs that neither chipset hardware-accelerates, such as On2’s VP6 (used with older-generation Flash content) and VP7 (harnessed with ABC.com’s online video service).
Followup: Some other Nvidia Ion analyses for your perusal:
Bradman commented:















