Indeed...magical. They actually create CPU cycles.
But seriously, yeah, I was going to ask about that. I am very surprised it wasn't even mentioned since over 100% utilization is just a silly thing to say. Does Apple count each core as 100% or something?
Under OS X's activity monitor in the list of processes you have CPU utilization as the percentage of a single core, so yes, 140% means it's using 70% of two cores. This is actually common in many non Windows OSes. If you've ever used top in Linux you'll see the same thing.
This is how multiple cores are typically measured since it's possible to have one core at 100% and another idle.
However, in this test I am confused as the article states that they are using only a single core. This makes sense to see the differences between the new and old Flash. It's very easy to turn off all but one core on Mac OS X with a developer tool Processor pref pane.
Also, it would have been great to get an HTML5 YouTube comparison thrown in. My tests with 360p, 480p, 720p and 1080p show Flash is using 3x to 6x as much processing. Some of this if Flash overhead but some could be how more efficient the WebKit framework is with H.264 decoding. I only tested Safari, not Chrome.
None of the "sponsored" YouTube pages have HTML5 at this point. I have to think it's a security reason because it's certainly H.264 video. The problem with that argument is that it couldn't be easier to grab a local copy of any YouTube video, so I don't know what the reasoning is.
Thats not true at all. If a game has versions on Windows, Mac or Linux, it's just about always the Win version that will run better, and sometimes a lot LOT better. Mainly because Windows has vastly better and more up to date drivers by both NV and ATI.
Flash not running well is more than likely down to Apple. It's taken this long for GPU accleration to happen with Apple as they only just opened up there low level API's for Adobe.
This is what you get when you go Apple and support a closed platform.
I know I'm kinda late to the party, but remember that a test like this was kinda done (although I think it should be re-done). This also brings up the issue for Anand if the browser has any impact on GPU Acceleration like it did for HTML5 video.
Wouldn't it make more sense to compare the new GPU accelerated to flash to the previous Flash 10.1 beta without acceleration? Instead of comparing it to 10.0?
Good question, one I want to know too. There are modern, aluminum iMacs with ATI graphics hardware in them; to not support them would be disappointing.
It you want this to work with an ATI card use Windows. It's been working with my ATI's for months.
Otherwise it's more than likely up to Apple and not Adobe/ATI with regards to it working with ATI hardware on OSX. Apple have to basically allow Adobe access to the API's for it to work. At the moment as far as i know only 3 NV cards work with apple and Flash acceleration.
This is what you get when you support a closed system.
I tried the new player on my 8800GT and it doesn't seem to work. This seems to be Apple's doing considering their tech note on the topic.
It's now even more depressing that my E8400 Dual core PC can still play back Flash video better than my 2xQuad core Mac Pro with identical video cards.
I'm not sure who your 'they' refers to, but it's Apple that created the API for Adobe to connect to in 10.6.3 and those are the only GPUs they included. I have to expect that more support for current and older GPUs will come, but remember this *is* Apple.
I think it's a safe bet the 9600M works since it's architecturally similar to the 9400M. It appears that Adobe was pretty conservative about what GPUs they mentioned in that blog post - I'd wager that we'll find out the real list empirically as people test and report back.
Adobe has to fix ALL of their Linux Flash implementations to stop crashing or running away with the CPU, not just add acceleration to their wildly unstable alpha 64 bit drivers.
It can actually play 720p flash video's just fine. It's also fine with 1080 from my TV tuner, it's just the flash won't use it. It's so old now anyway though that it doesn't bother me, I'm just going to have to finally upgrade it, and pretty much everything else.
Since I do a lot of online video deployment, I've been running some tests tonight as well. Using The Muppets: Bohemian Rhapsody in 1080p as a test video: http://www.youtube.com/watch?v=tgbNymZ7vqY
Running on a 2.4ghz 13" Unibody MB (9400M) and tested in Safari 4.0.5 and tonight's nightly Webkit build... results are identical.
I know that 1080p is an extreme and not the norm for browser playback, but the CPU temp never went over 154ºF while in the Youtube/HTML5 player and the fans never ramped up speed from the base 1800RPM. However, using the Gala preview, the temp climbed up to 200ºF as the fans slowly ramped up to 6000RPM to rein it in, eventually settling in at 172ºF at 6200RPM.. at least the frame rate didn't drop and playback was fluid. But still, 154ºF/1800RPM vs 172ºF/6200RPM. So Adobe has offloaded A LOT to the GPU, which is great, but the underlying performance issues are still there. GPU acceleration is great, but Flash on OSX still sucks at drawing anything to the screen.
Oh, and just for fun, the 480p version of this video is encoded in VP6 with Flash 10.1 Gala using 60% and Safari using 8%
I think that if both playback methods are watchable, then I guess the big question would be which approach has better battery life? It seems to me that the savings from the drop in CPU usage would be completely lost by starting up the GPU. Perhaps it's worth comparing? If it makes it worse, then I'd probably pass on that upgrade until they support the IGP (if I owned a mac, that is).
I've tested it on Macbook Pro 13 inch 2.53 ghz with 9400m on karate kid 1080p. CPU utilization seems to be slightly lower about 10-20% with Gala, but what I cared about more is battery life. I was monitoring the wattage my battery is draining, both Flash Gala and the current flash player drain about equal about of wattage, perhaps slightly higher with Gala at about 26-33 watts, the difference is not very noticeable. With HTML5 though however, it's lowered to 24-28 watts. Well I'm disappointed to say the least. I thought 10.1 is gonna further boost battery life on laptops, but I guess it's only true for the androids right now.
It's only using the vp4 video processor on the graphics card, right? Loading that alone must be a better alternative than loading a corei5 i would think.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
38 Comments
Back to Article
michal1980 - Wednesday, April 28, 2010 - link
They can utilize more then 100% of the cpu.MadMan007 - Wednesday, April 28, 2010 - link
Indeed...magical. They actually create CPU cycles.But seriously, yeah, I was going to ask about that. I am very surprised it wasn't even mentioned since over 100% utilization is just a silly thing to say. Does Apple count each core as 100% or something?
Anand Lal Shimpi - Wednesday, April 28, 2010 - link
Under OS X's activity monitor in the list of processes you have CPU utilization as the percentage of a single core, so yes, 140% means it's using 70% of two cores. This is actually common in many non Windows OSes. If you've ever used top in Linux you'll see the same thing.Take care,
Anand
MadMan007 - Thursday, April 29, 2010 - link
That's what I figured. It would more properly be called 'Core % usage' then.solipsism - Wednesday, April 28, 2010 - link
This is how multiple cores are typically measured since it's possible to have one core at 100% and another idle.However, in this test I am confused as the article states that they are using only a single core. This makes sense to see the differences between the new and old Flash. It's very easy to turn off all but one core on Mac OS X with a developer tool Processor pref pane.
Also, it would have been great to get an HTML5 YouTube comparison thrown in. My tests with 360p, 480p, 720p and 1080p show Flash is using 3x to 6x as much processing. Some of this if Flash overhead but some could be how more efficient the WebKit framework is with H.264 decoding. I only tested Safari, not Chrome.
Anand Lal Shimpi - Thursday, April 29, 2010 - link
I was just clarifying that it's % of a single core being used (e.g. 140% of one core or 70% of two cores). I'll clean up the text a bit.HTML5 does appear to use far less CPU than Flash, however I managed to pick one of the only trailers without HTML5 support in YouTube :-P
Take care,
Anand
solipsism - Thursday, April 29, 2010 - link
None of the "sponsored" YouTube pages have HTML5 at this point. I have to think it's a security reason because it's certainly H.264 video. The problem with that argument is that it couldn't be easier to grab a local copy of any YouTube video, so I don't know what the reasoning is.lyeoh - Thursday, April 29, 2010 - link
Why does Flash perform so much worse on nonwindows platforms?I find that puzzling. Many 3D games run as fast (if not faster) on non-windows platforms (when the developers bother to port them over).
B3an - Sunday, May 2, 2010 - link
Thats not true at all. If a game has versions on Windows, Mac or Linux, it's just about always the Win version that will run better, and sometimes a lot LOT better. Mainly because Windows has vastly better and more up to date drivers by both NV and ATI.Flash not running well is more than likely down to Apple. It's taken this long for GPU accleration to happen with Apple as they only just opened up there low level API's for Adobe.
This is what you get when you go Apple and support a closed platform.
mmendoza27 - Thursday, April 29, 2010 - link
I know I'm kinda late to the party, but remember that a test like this was kinda done (although I think it should be re-done). This also brings up the issue for Anand if the browser has any impact on GPU Acceleration like it did for HTML5 video.http://www.streaminglearningcenter.com/articles/fl...
teryan2006 - Wednesday, April 28, 2010 - link
Wouldn't it make more sense to compare the new GPU accelerated to flash to the previous Flash 10.1 beta without acceleration? Instead of comparing it to 10.0?Anand Lal Shimpi - Wednesday, April 28, 2010 - link
The numbers are roughly the same between the two versions for these tests (just confirmed - ~150% using 10.1.53.21 in the 1080p Karate Kid trailer).Thanks for the heads up though, you had me worried for a moment :)
Take care,
Anand
Naton - Wednesday, April 28, 2010 - link
Is there any word on whether this might be expanded to support ATI cards at some point?LoneWolf15 - Thursday, April 29, 2010 - link
Good question, one I want to know too. There are modern, aluminum iMacs with ATI graphics hardware in them; to not support them would be disappointing.B3an - Sunday, May 2, 2010 - link
It you want this to work with an ATI card use Windows. It's been working with my ATI's for months.Otherwise it's more than likely up to Apple and not Adobe/ATI with regards to it working with ATI hardware on OSX. Apple have to basically allow Adobe access to the API's for it to work. At the moment as far as i know only 3 NV cards work with apple and Flash acceleration.
This is what you get when you support a closed system.
idontusenumbers - Wednesday, April 28, 2010 - link
I tried the new player on my 8800GT and it doesn't seem to work. This seems to be Apple's doing considering their tech note on the topic.It's now even more depressing that my E8400 Dual core PC can still play back Flash video better than my 2xQuad core Mac Pro with identical video cards.
UGH!
Ryan Smith - Thursday, April 29, 2010 - link
Which makes things all the weirder since the GTX 285 works.9000 and GTX 200 Series = VP2
9400M = VP3
330M (and presumably 320M) = VP4
So there's no reason why 8800GT shouldn't work, since it's VP2. The same goes for 9600M. VP2 is fully capable of full H.264 decoding.
jordanclock - Thursday, April 29, 2010 - link
Perhaps the mechanism that checks for compatibility isn't checking for VP2, but for the model number to be >9200 or something.B3an - Sunday, May 2, 2010 - link
Guys the only cards that work with Flash acceleration on OSX are:GeForce 9400M,
GeForce 320M,
GeForce GT 330M.
This is most probably down to Apple. They only just allowed Abobe access to the low level API's.
Please stop supporting a closed system.
drvelocity - Thursday, April 29, 2010 - link
They programmed in 9400 support but left out the 9600M? Odd..solipsism - Thursday, April 29, 2010 - link
I'm not sure who your 'they' refers to, but it's Apple that created the API for Adobe to connect to in 10.6.3 and those are the only GPUs they included. I have to expect that more support for current and older GPUs will come, but remember this *is* Apple.Brian Klug - Thursday, April 29, 2010 - link
I think it's a safe bet the 9600M works since it's architecturally similar to the 9400M. It appears that Adobe was pretty conservative about what GPUs they mentioned in that blog post - I'd wager that we'll find out the real list empirically as people test and report back.-Brian
Dâniel Fraga - Thursday, April 29, 2010 - link
Well, now Adobe needs to implement GPU acceleration on their Flash 64 bit (still alpha) client for Linux too...HangFire - Thursday, April 29, 2010 - link
Adobe has to fix ALL of their Linux Flash implementations to stop crashing or running away with the CPU, not just add acceleration to their wildly unstable alpha 64 bit drivers.HangFire - Thursday, April 29, 2010 - link
I meant "wildly unstable 64 bit alpha Flash", not drivers.Where's the edit feature?
Necrosaro420 - Thursday, April 29, 2010 - link
Wish they would get the x64 flash working already....BlakJupitr - Thursday, April 29, 2010 - link
I can't even use it with my PC, considering the FX 5200 I'm still using. >=\Flash at this level.. *bangs head on keyboard*
piroroadkill - Thursday, April 29, 2010 - link
A 5200? I'm sorryBlakJupitr - Thursday, April 29, 2010 - link
Me too.. lolIt can actually play 720p flash video's just fine. It's also fine with 1080 from my TV tuner, it's just the flash won't use it. It's so old now anyway though that it doesn't bother me, I'm just going to have to finally upgrade it, and pretty much everything else.
vasu42 - Thursday, April 29, 2010 - link
Since I do a lot of online video deployment, I've been running some tests tonight as well. Using The Muppets: Bohemian Rhapsody in 1080p as a test video: http://www.youtube.com/watch?v=tgbNymZ7vqYRunning on a 2.4ghz 13" Unibody MB (9400M) and tested in Safari 4.0.5 and tonight's nightly Webkit build... results are identical.
Flash 10.1 RC2 = 166% Flash + 20% Safari
Flash 10.1 Gala = 92% Flash + 20% Safari
HTML5 Playback = 20% Safari
I know that 1080p is an extreme and not the norm for browser playback, but the CPU temp never went over 154ºF while in the Youtube/HTML5 player and the fans never ramped up speed from the base 1800RPM. However, using the Gala preview, the temp climbed up to 200ºF as the fans slowly ramped up to 6000RPM to rein it in, eventually settling in at 172ºF at 6200RPM.. at least the frame rate didn't drop and playback was fluid. But still, 154ºF/1800RPM vs 172ºF/6200RPM. So Adobe has offloaded A LOT to the GPU, which is great, but the underlying performance issues are still there. GPU acceleration is great, but Flash on OSX still sucks at drawing anything to the screen.
Oh, and just for fun, the 480p version of this video is encoded in VP6 with Flash 10.1 Gala using 60% and Safari using 8%
MonkeyPaw - Thursday, April 29, 2010 - link
I think that if both playback methods are watchable, then I guess the big question would be which approach has better battery life? It seems to me that the savings from the drop in CPU usage would be completely lost by starting up the GPU. Perhaps it's worth comparing? If it makes it worse, then I'd probably pass on that upgrade until they support the IGP (if I owned a mac, that is).Spivonious - Thursday, April 29, 2010 - link
AFAIK, Hulu doesn't support hw acceleration. It's a big gripe over at the AVS Forum.DukeN - Thursday, April 29, 2010 - link
You're not allowed to get mad at Father Steve.Please remember, all thoughts need 2-5 days of approval by Father Steve's committee before you can express them. Just like the apps.
KJ242 - Thursday, April 29, 2010 - link
I read this as Apple finally enabling this and not adobe:http://labs.adobe.com/technologies/flashplayer10/g...
B3an - Sunday, May 2, 2010 - link
Yep it's all down to Apple. Closed system POS remember.happycamperjack - Thursday, April 29, 2010 - link
I've tested it on Macbook Pro 13 inch 2.53 ghz with 9400m on karate kid 1080p. CPU utilization seems to be slightly lower about 10-20% with Gala, but what I cared about more is battery life. I was monitoring the wattage my battery is draining, both Flash Gala and the current flash player drain about equal about of wattage, perhaps slightly higher with Gala at about 26-33 watts, the difference is not very noticeable. With HTML5 though however, it's lowered to 24-28 watts. Well I'm disappointed to say the least. I thought 10.1 is gonna further boost battery life on laptops, but I guess it's only true for the androids right now.osmosium - Monday, May 3, 2010 - link
Oh Flash, how many ways can you be the infection vector for viruses... let me count the ways... now with OS X driver virus vector support!Os
mindbomb - Monday, May 3, 2010 - link
It's only using the vp4 video processor on the graphics card, right? Loading that alone must be a better alternative than loading a corei5 i would think.