GPU Accelerated H.264 Decode

Both AMD and NVIDIA have fully accelerated H.264 decode on their GPUs and chipsets with integrated graphics, something Intel just recently added with G45. Apple has historically done a terrible job of taking advantage of GPU accelerated video decode features in its OS, which is partly why it took us so long to get somewhat acceptable deinterlacing in Apple's DVD app, despite solid deinterlacing support by the GPU vendors.

Part of the problem is that unlike on the PC platforms, companies like ATI and NVIDIA don't write the entire driver for Mac OS X. The GPU vendors provide the hardware interface portion of the driver but Apple handles much of the rest. It's up to Apple to take advantage of the various features supported by the GPU, and most of them aren't high on Apple's priority list.

With Apple heavily pushing H.264 as the codec of choice and offering high bitrate HD H.264 movie trailers at www.apple.com, the move to NVIDIA's GeForce 9400M appeared to be the perfect time to take advantage of GPU accelerated H.264 decode. For the first time ever, Quicktime will use the GPU for the majority of the decode pipeline when playing back H.264 content.

To test the impact of the GPU offload I measured CPU utilization while playing back the 160MB 1080p trailer of The Spirit from Apple's movie trailers website:

  New MacBook Pro 15" Old MacBook Pro 15"
CPU Utilization 10% 30 - 50%

 

The old MacBook Pro saw 30 - 50% of its dual core Core 2 Duo 2.5GHz CPU used up to decode the trailer, while the new MacBook Pro only needed 10% as the rest was done on the GPU. The GPU is also a more efficient place to perform H.264 decoding so you'll actually see an increase in battery life when playing back hardware accelerated content.

As OS X still lacks any official Blu-ray support, the H.264 decode acceleration isn't very useful beyond playing these sorts of files, but it's a definite start. It's also unclear how easy it will be for 3rd party developers to tie into the GPU acceleration hooks or if Quicktime will be your only hope for that.

Performance Battery Life: Take Two Steps Forward, and Two Steps Back
Comments Locked

66 Comments

View All Comments

  • Johnmcl7 - Thursday, January 15, 2009 - link

    I didn't say it was for *nix, that's why I said *nix applications which still use the middle mouse button in other operating systems. There are many times when there isn't space for using a mouse, hence it's a laptop.

    As for keyboard shortcuts, they're not faster when using a mouse as it means a break from the sequence rather than just clicking with the mouse that's in use anyway
  • themadmilkman - Wednesday, October 22, 2008 - link

    Why don't you head to a store and try it? It's much more intuitive than you give it credit for.
  • Sunrise089 - Wednesday, October 22, 2008 - link

    There was a time when cars were changed and tweaked every single year, often for purely aesthetic/emotional reasons. That is no longer true. The average enthusiast car shopper is no less spec-conscious than PC geeks. And likewise the majority, and especially in the high-end/luxury market (Lexus, Apple) that is composed less by knowledgeable enthusiasts and more by people craving a certain image or experience, tend to shop based upon style, price, or other easy-to-understand factors.
  • headbox - Saturday, October 25, 2008 - link

    wrong. If people were "spec-conscious" about what they drive and getting performance was priority #1, then we'd see thousands of motorcycles on the freeways instead of dozens. You can spend $8,000 and get a motorcycle that is faster than any car made, gets 50 mpg, and can still carry a few bags of groceries.

    People buy nice cars because they can afford it and like the aesthetics.
  • RaynorWolfcastle - Wednesday, October 22, 2008 - link

    Just a note, but I've read elsewhere that under Windows, the graphics on the MBP always use the 9600 chip; I'm sure this accounts for part of the difference in battery life (assuming you ran the OSX tests using the integrated 9400 video.
  • ltcommanderdata - Wednesday, October 22, 2008 - link

    The Windows vs OS X battery life tests were done on a MacBook Air so discrete GPU has no effect.
  • jonmcc33 - Wednesday, October 22, 2008 - link

    Maybe you should test the power settings with Vista on Power saver setting? My Latitude D610 lasts over 3 hours with Vista. I wouldn't use Balanced unless it was plugged into the AC adapter.
  • Calin - Thursday, October 23, 2008 - link

    What about testing under XP?
  • jonmcc33 - Thursday, October 23, 2008 - link

    Nobody cares about Windows XP and it would be REALLY bad to compare to the latest Mac OS X product.
  • BushLin - Tuesday, October 28, 2008 - link

    I don't see why, XP isn't a limitation on anything useful unless you were just talking about the eye candy of OS X... See how many businesses still supply their laptops with XP rather than the junk they're supplied with because they're not tethered to Microsoft like the manufacturers.

Log in

Don't have an account? Sign up now