Does the Elgato Turbo.264 really make much of a difference on Intel hardware compared to, say, VisualHub?
I was not able to find a good answer before buying a Turbo 264 on a Black Friday sale.
For example, if you watch YouTube - Elgato Turbo.264 vs QuickTime vs VisualHub of a performance test on MacMini CoreDuo 1,66Ghz 2GB Ram. AppleTV video conversion, the Turbo is a clear winner. However, they ran all 3 comparisons at the same time, which means that VisualHub and QuickTime were both fighting for the processor. This is, to put it clearly, stupid. Such a test is designed to show off the Turbo.264 (which is essentially a dedicated processor for video). But no one in their right mind would run all 3 of those programs at once.
A better test would be to take the same file and try it with each program.
So that’s what I did.
In fact, I tried the same file on 3 different computers using 3 different tests.
My Hardware:
1) A 1.5Ghz Powerbook G4 with 1gb of RAM
2) A 2.0Ghz MacBook Core Duo (first generation) with 2gb of RAM
3) A 2.8Ghz iMac Core 2 Duo EXTREME! with 2gb of RAM (I have no idea why it is called Extreme, it just is)
Note that clearly the Powerbook is at a disadvantage because it has less RAM. This is not a clear Apples-to-Apples comparison. It is, however, what I had on hand.
My Video:
The EyeTV tests were done with a 28 minute clip of one of my son’s shows, recorded via EyeTV 2.5.1 in MPEG-2 (DVD quality). The commercials had been edited out and the ‘compact’ feature in EyeTV had been used. The file was simply selected and I clicked the “Apple TV” button, first without the Turbo.264 attached, and then again with it attached.
The final test was done using an MPEG exported from EyeTV of the same video because VisualHub cannot re-encode EyeTV files directly, but EyeTV allows for easy saving of files as MPEG program streams. Exporting took 2 minutes 2 seconds on the Powerbook, 1 minute 10 seconds on the MacBook, and 45 seconds on the MacBook. That time should be added to the VisualHub totals below.
UPDATE 2007-12-11: VisualHub can process EyeTV MPEG files, however, make sure that you drag the .eyetv files to the VisualHub rather than using the + to add them to the VH queue. I re-ran the conversion on the iMac and the MacBook and got the same exact times as shown below. I’m not going to make the poor ol’ Powerbook spend 3 hours re-converting this video again, but will assert that VH will convert an .eyetv file or a .mpg file of the same video in the same amount of time. The difference is that you do not have to export it to mpg first, which is a nice thing. It only took 2 minutes on the Powerbook (45 seconds on the iMac) but it was another step and another copy of the file that you had to have on your hard drive. When dealing with a 700+MB file, you don’t want any more copies around than necessary.
VisualHub was set to the ‘high’ setting for each test, and the “Low CPU” preference was turned off, giving VH the most access to the CPU possible.
| Powerbook | |
| Exported from EyeTV | 270 minutes |
| Exported via Turbo | 043 minutes |
| Converted with VisualHub | 168 minutes |
| Converted via Handbrake | 195 minutes | MacBook |
| Exported from EyeTV | 040 minutes |
| Exported via Turbo | 029 minutes |
| Converted via VisualHub | 031 minutes |
| Converted via Handbrake | 39 minutes | iMac |
| Exported from EyeTV | 024 minutes |
| Exported via Turbo | 051 minutes |
| Converted via VisualHub | 021 minutes |
| Converted via Handbrake | 023 minutes |
(I hate writing HTML tables, even for tabular data. What a PITA.)
A few things really jump out.
1) If you are planning to do any video encoding on a G4, get the Turbo. It was almost 4x faster than VisualHub, and over 6x faster than a regular export from EyeTV. (I would really like to know how a dual G5 would do in this test.) The Powerbook+Turbo was nearly as fast as a MacBook without the Turbo (43 minutes vs 40).
2) The Turbo makes a significant difference on the MacBook: 29 minutes instead of 40, plus the fact that the CPU usage with the Turbo is minimal. Without the Turbo, plan to spend 40 minutes with the fan running on your laptop and doing very little else with the computer at that time. Note that this gets us very close to “real time” encoding: a 28 minute file in 29 minutes.
3) VisualHub is very fast. On the Powerbook, VH took 2 hours and 48 minutes compared to a whopping 4 hours and 30 minutes with the regular EyeTV. However on G4 hardware, you can’t beat the Turbo. On the MacBook, VH is nearly as fast as the Turbo, but again the difference is that the Turbo does free up your main CPU.
4) Perhaps most surprising are the iMac results. I’m sure most of you noticed that the Turbo actually slowed down the encoding on the iMac. I found that so hard to imagine that I re-ran the test just to make sure. Yes, it took over twice as long (51 minutes vs 24) using the Turbo. Explanation? Simple: the Turbo’s processor isn’t as fast as the iMac — which really shouldn’t come as any surprise. If that wasn’t surprising enough, how about the fact that VisualHub was faster than EyeTV and almost 2.5 times faster than the Turbo?
VisualHub performance gains are clear on PowerPC hardware, but really starts to scream as you move up the Intel food-chain. 21 minutes for a 28 minute file is (if my math is right) a 25% improvement over “real time” encoding. It should also not be overlooked that VisualHub also has a lot of options beyond just exporting for AppleTV, and several orders of magnitude of preference settings available over the Turbo. It’s like comparing a bag lunch with a full menu at a fancy restaurant with a good chef. The Turbo is not trying to be a super-configurable device, it is trying to be a simple plug and play improvement. On that basis it does succeed for all but the most high end hardware, given that the iMac tested is currently the fastest one available. I do not have access to a MacPro to check how it would perform there, but my guess is that it would scream.
It should also be noted that the Turbo does not improve times for MacTheRipper, Handbrake, or VisualHub. It will improve any program which exports video using QuickTime. It also does not improve the time that EyeTV takes to compact video files — a process usually done after you have edited the video in EyeTV. Note that you do not have to compact the video before exporting it from EyeTV, however A) I always do because it frees up hard drive space and B) I assume (and I may be wrong) that VisualHub does not support the markers that EyeTV uses internally to indicate edits, so if you use VH to re-encode EyeTV files which have been edited but not compacted, your edits will not be in the resulting file.
So What Will I Do?
So now that I have these 3 machines and the information at hand, how will I use them?
Will I return the Turbo.264?
No. The Turbo will remain hooked to the MacBook when processing EyeTV video. Why? Because it will then be able to multi-task by running an EyeTV export along with a Handbrake conversion at the same time, due to the fact that the Turbo will offset the EyeTV CPU usage.
Is the Powerbook really good for anything when it comes to this workflow?
Yes! The Powerbook will be used to record shows via the EyeTV, and possible mark edits. Those files will then be moved either to the MacBook or iMac for editing and encoding for Apple TV. (This is one of those times when having Gigabit Ethernet comes in really handy. Even if you don’t have a Gigabit hub, plug the cable directly from the Powerbook to the other computer and voilà!)
However, clearly the VisualHub on the iMac is the clear winner for all my other conversion needs, whether they be EyeTV files, or avi files, or mpg files, etc.
(Possible followup: Compare Quicktime Pro (non-Turbo) on Powerbook vs MacBook vs iMac on converting movies to Apple TV format.)