Do Browsers Matter? What About Full-Screen?
Does Your Browser Matter?
Surprisingly, yes. It's not something many people talk about, but remember that Flash is technically provided as two plug-ins. There is an ActiveX version written specifically for IE8 and another one for every other browser.
| Playback: YouTube 1080p on Asus UL20AAero Enabled, Hardware Acceleration Enabled | CPU Usage: Windowed |
|---|---|
| IE8 32-bit: 8.0.7600.1685 | 35% |
| Firefox: 3.6.12 | 49% |
| Opera: 10.63 (build 3516) | 25% |
| Chrome: 8.0.552.215 | 28% |
| Safari: 5.0.3 | 26% |
As far Adobe is concerned, the differences between the two Flash versions should be negligible. From a performance standpoint, the company has no specific example to suggest one browser is better for Flash than another.
Our own testing, however, reveals some interesting results. Generally speaking, we do see high CPU usage on Firefox and IE8. All browsers and all configurations are able to play video without stutter (24/30 FPS), provided the browser is in windowed mode and has hardware acceleration enabled.
We actually reran the numbers on a few systems after looking at the results from our first round of testing. The second round of numbers showed similar margins, which is why we believe this is a not an aberration. From our conversations with Adobe last week, it seems likely that there is something within each individual browser that affects how they are handling Flash video.
We also noticed a bit of a glitch unique to IE8. If you move the cursor rapidly during Flash video playback, skipping occurs. The degree of frame dropping depends on the CPU. Even on a Sandy Bridge Core i7-2820QM-based laptop and a Core i7-920-powered desktop we see the same results, which is downright strange considering the raw horsepower under each machine's hood. Because this is limited to IE8, it seems to be an artifact of the software-based difference between ActiveX and the plug-in version of Flash Player. Adobe seems to be unaware of this glitch, but we are informed that they are looking into the issue.
What About Full Screen?
| Playback: YouTube 1080p on Asus UL20AAero Enabled, Hardware Acceleration Enabled | Fullscreen |
|---|---|
| IE8 32-bit: 8.0.7600.1685 | CPU Usage: 52%27.0 FPS |
| Firefox: 3.6.12 | CPU Usage: 60%27.8 FPS |
| Opera: 10.63 (build 3516) | CPU Usage: 37%15.5 FPS |
| Chrome: 8.0.552.215 | CPU Usage: 39%14.0 FPS |
| Safari: 5.0.3 | CPU Usage: 39%13.8 FPS |
Full-screen matters. On the Windows side, remember all post-processing data operations, such as scaling, occur on the CPU. So, we weren't surprised to see higher CPU usage. What did surprise us was an inverted relationship between CPU usage and FPS. Even with lower CPU overhead, the FPS count is almost half of what it should be in order to see fluid playback on Sarfari, Opera, and Chrome.
Strangely, there seem to be two factors in play here: the browser we use and the available CPU horsepower. On a beefier system like the ThinkPad T510 (Core i5-540M), we don't see that browser choice matters, beyond some impact on CPU usage, since playback is completely fluid. Across the board on non-desktop-class systems, we do see this phenonmemon occurring--perhaps another artifact of the way individual browsers handle Flash video.
