-
1. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
AngeloLorenzo Jul 18, 2013 6:09 PM (in response to MD Optofonik)I've never seen a deep universal reference but that's because you can't compare compression schemes apples-to-apples. I wrote this a while back http://www.fallenempiredigital.com/blog/2013/02/08/a-guide-to-common-video-formats-contain ers-compression-and-codecs/ which is more of a primer; the world of codecs, on a professional level is small.
characteristics as it relates to system performance, i.e., codecs that put the least amount of load on system resources.
Well that really comes down to a trade off. You either put load on your storage systems and throughput with less compressed material, or you put stress on your CPU with more highly compressed material.
Since it's such a broad topic, perhaps you should go into detail about an actual workflow issue you're trying to solve.
-
2. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
MD Optofonik Jul 18, 2013 7:07 PM (in response to AngeloLorenzo)You either put load on your storage systems and throughput with less compressed material, or you put stress on your CPU with more highly compressed material.
Finally, I understand something that I've read elsewhere but in the most convoluted terms. Thanks for the concise insight. I'll check out your blog post.
-
3. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
Kranex1 Jul 18, 2013 7:16 PM (in response to MD Optofonik)Wouldn't that depend mostly on your system configurations? For example, a very fast CPU will experience less strain than a slower one. Anyway, AVCHD is typically a CPU intensive codec.
-
4. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
MD Optofonik Jul 18, 2013 10:20 PM (in response to Kranex1)Here's the thing, all things being equal, there is a codec that can be considered to have the smallest throughput hit and there is a codec that has the least CPU hit on any system. It really is that simple. Now, whether or not anyone has done the research and has a comprehensive understanding of such things could be another story.
There is a car out there that can be considered the slowest and a car out there that can be considered the fastest. Now, anyone can complicate the issue as much as they want by introducing variables that make themselves appear knowledgeable but there are others who will answer the question without trying to appear clever because they're just knowledgeable about the subject.
-
5. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
AngeloLorenzo Jul 19, 2013 4:46 AM (in response to MD Optofonik)True but besides being an academic curiousity, you still have to satisfy other goals to varying degrees of trade off. Without the context of a workflow, it's meaningless. Even then, unless you're under an extreme budget or hardware limitation, it's money well spent to scale hardware for the fastest overall workflow. Tools are usually cheaper than man hours in the long term.
If we want to be simple about things then we could say that h.265 is the most processor intensive as it involves the most advanced compression scheme currently slated for mainstream usage. As far as least impact, probably uncompressed YCrCb 8-bit (anything higher quality and you're burning CPU just to deal with throughput).
To play off your analogy, I could have an Audi R10 or a BMW Isetta, but if my goal is to drive my family around, neither fits the bill.
-
6. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
Kranex1 Jul 19, 2013 10:15 AM (in response to MD Optofonik)Here's the thing, all things being equal, there is a codec that can be considered to have the smallest throughput hit and there is a codec that has the least CPU hit on any system. It really is that simple. Now, whether or not anyone has done the research and has a comprehensive understanding of such things could be another story.
Maybe. But you'd also need to include more data since degrees of performance can depend heavily on conditions and other criteria.
In other words, the question would be easier to answer if it were more specific. Say a comparison between two codecs, like A and B would help that. Generally speaking, the codec with the least amount of data and/or compression.
There is a car out there that can be considered the slowest and a car out there that can be considered the fastest. Now, anyone can complicate the issue as much as they want by introducing variables that make themselves appear knowledgeable but there are others who will answer the question without trying to appear clever because they're just knowledgeable about the subject.
In that case you limit their ability to "complicate the issue" by using specifics. For example by fastest do you mean 0-60 or top speed? Therefore, I'd say which car is fastest a Corvette Z06 or Porsche 911 from from 0-60? Or something similar to that.
-
7. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
MD Optofonik Jul 19, 2013 12:30 PM (in response to AngeloLorenzo)Using an 8-bit YCbCr colors space would help with throughput? What codec could that be married to in order to reduce CPU load?
BTW: ThrustSSC / Benz Patent-Motorwagen. It really is that simple.
-
8. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
nw42 Jul 19, 2013 2:02 PM (in response to MD Optofonik)such a comparison quickly becomes rocket science that no "normal user" can understand any more not even the experts itself some times ;-)
I good approach would be to ask: "What would be a suitable codec for a given task?"
So you quickly could see that H.264 isn't the best option as a mezzanine codec or that a DPX file sequnce doesn't fit the requirements for web streaming... ;-)
-
9. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
MD Optofonik Jul 20, 2013 2:16 PM (in response to nw42)Editors may not be rocket scientists but are they not highly skilled technicians who should be familiar with the myriad technical specifications of their craft?
-
10. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
Kranex1 Jul 20, 2013 2:53 PM (in response to MD Optofonik)Philosophically speaking? Maybe? But from a practical standpoint, they would like be to busy.
As for codec disclosure, which ever one you pick will say their the best. For sure that's the case with prores. According to Apple, if you're using anything else, you're not getting the best there is.
-
11. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
MD Optofonik Jul 20, 2013 9:10 PM (in response to Kranex1)Kranex1 wrote:
Philosophically speaking? Maybe?
Perhaps this is the crux of Adobe's market share problem with Premiere. It's not a philosophical question.
-
12. Re: Is there a reference anywhere of codecs, their relative file sizes, and performance characteristics?
Mark Mapes Jul 20, 2013 11:05 PM (in response to MD Optofonik)For starters: http://en.wikipedia.org/wiki/Comparison_of_video_codecs
As AngeloLorenzo pointed out--and as you may be well aware--codecs serve various purposes: in-camera compression, digital intermediate, final distribution... Those created for one purpose may not be much good for others. Unless you compartmentalize this kind of discussion, it won't get very far. If the parenthetical reference in your first post to "tape based and card based" was your way of narrowing the scope, then I gather you're basically asking about the characteristics of recording formats and how they perform for editing (assuming you don't transcode to a DI). But that's just a guess, so please clarify. In case I'm on the right track, I'll offer one quick observation: interframe compression formats like AVCHD and HDV tend to yield a higher compression ratio (i.e., smaller file sizes for a given resolution and duration), but decoding them is more CPU-intensive; intraframe compression formats like AVC Intra and DVCPRO tend to have a lower compression ratio but take less horsepower to decode.
BTW, I'm rather curious what you think is "Adobe's market share problem," and how Kranex's observation speaks to it. Particularly given that Kranex's comment came in reference to whether it's reasonable to expect professional editors to "be familiar with the myriad technical specifications of their craft" and had nothing directly to do with Premiere Pro.



