H.264 Standard

[jcicc]

I was speaking with someone at Polycom regarding H.264 and was told that it works up to speeds of 512K. In that past everything I've ever read has been that H264 caps at 384K. Can someone give me a reality check here?

[11B-33T]

We've used our VSX 7000 & 8000 to connect via H.323 to H.264 compliant/capable codecs up to 512k. Any rate above defaults to H.263. We're still dinking around via ISDN to see the caps limit.

[pzielie]

H.264 itself does not top out at any particular speed but H.264 is a very processor intensive compression scheme and any given processor will have a limit it can encode at. H.264 also uses a lot more processor on the decode side so there will also be limitations there. You also have to balance latency against encoding quality limiting how much data you can get through the system.

Any H.26X standard is a decode standard. The encoder only has to present the data in a form that can be decoded by a standard decoder. How the image is compressed is up to the encoder so video quality may actually be better with an improved compression algorithm (more processor intensive) at a lower bandwidth than a lousy algorithm at a higher bandwidth.

[trapehzoid]

I was speaking with someone at Polycom regarding H.264 and was told that it works up to speeds of 512K. In that past everything I've ever read has been that H264 caps at 384K. Can someone give me a reality check here?

H264 does not cap at these speeds.. but your polycom endpoint might. What you are hearing are about implementation in a specific product.. not H264 limitations in itself.

One of the primary benefits of H264 is improved video at lower bitrates.. so low-end is where the first benefits of H.264 will be. as the power levels increase in the systems, they will be able to implement h.264 at higher speeds and actually get some benefit from it.

[Sean Lessman]

H264 does not cap at these speeds.. but your polycom endpoint might. What you are hearing are about implementation in a specific product.. not H264 limitations in itself.

One of the primary benefits of H264 is improved video at lower bitrates.. so low-end is where the first benefits of H.264 will be. as the power levels increase in the systems, they will be able to implement h.264 at higher speeds and actually get some benefit from it.
TANDBERG MXP endpoints support up to 2Mbps (30fps starting at 192kbps) with H.264 or up to the maximum rate of the particular unit (i.e. 768 for the T770).

H.264 does not specify a cap on speeds.

Sean

[Kevin]

H264 does not cap at these speeds.. but your polycom endpoint might. What you are hearing are about implementation in a specific product.. not H264 limitations in itself.

One of the primary benefits of H264 is improved video at lower bitrates.. so low-end is where the first benefits of H.264 will be. as the power levels increase in the systems, they will be able to implement h.264 at higher speeds and actually get some benefit from it.


The Codian MCU does H.264 on all ports all the way up to 2Mbps on all ports with full continuous presence and no loss of functionality or port count.

Certainly a lot of the gain is achieved is at the lower bitrates, though you can definitely see improvements at higher bandwidths too.

Be careful though. You can get into situations where H.264 will reduce your call quality if you are using older equipment. Sending H.264 to older underpowered endpoints can result in you receiving worse video back from it as it has to use too much of its resources to decode the video you are sending it so not enough processing power is left for its encode - this was discussed a while ago in this post on Wainhouse:

http://www.wainhouse.com/ubb/Forum3/HTML/000382-2.html

See the posting by william408 near the end.

Kevin.

[trapehzoid]

Thanks for the ads.. we now take you back to our regularlly scheduled discussions

Plus.. I don't really put much weight into william's analysis ( isn't that another codian guy?) given you don't have insight into how the units are actually operating.

I guess codian should make an endpoint since they seem to have this whole encoding thing down pact

[Entropy3XD]

Thanks for the ads..
.....and the smarta$$ comment. Sean and Kevin are simply referring to the H.264 specs of the equipment their companies make and did not make any comparisons to other vendor equipment. Both Sean and Kevin are up front in showing who they work for and I would not expect these guys to concern themselves with products from other manufacturers, leaving their comments to the equipment they know best.....their own. It would be most welcome if other vendors stepped up to discuss the particulars of their own equipment in order to give a little more balance to the forums. Unfortunately some of the other vendors have not come forward to take advantage of this.

[Sean Lessman]

.....and the smarta$$ comment. Sean and Kevin are simply referring to the H.264 specs of the equipment their companies make and did not make any comparisons to other vendor equipment. Both Sean and Kevin are up front in showing who they work for and I would not expect these guys to concern themselves with products from other manufacturers, leaving their comments to the equipment they know best.....their own. It would be most welcome if other vendors stepped up to discuss the particulars of their own equipment in order to give a little more balance to the forums. Unfortunately some of the other vendors have not come forward to take advantage of this.
Sorry guys, wasn't trying to plug the gear. Was trying to give an example of how equipment on the market is already doing higher than the gentleman (allegedly) implied from Polycom. I do not talk about other people's stuff unless they put the gauntlet down first -- and then I am happy to discuss the pros and cons of anyone's gear :)

Sean

[smoody]

Personally I applaud Sean Lessman for his active participation in this (and other forums) as a representative of his company. Not only have I have gained many valuable insights from his comments, but enjoy being able to have a direct dialogue with someone who can help implement changes without having to go through a "barrier of bureaucracy" .

I too wish that someone from Polycom (which I've personally requested from them many times) and other manufactures would not only publicly join these forums, but offer comments and respond to inquires in the professional manner Mr. Lessman has.

Keep up the good work Sean!

[trapehzoid]

.....and the smarta$$ comment. Sean and Kevin are simply referring to the H.264 specs of the equipment their companies make and did not make any comparisons to other vendor equipment.

The question wasn't about the specs of either of their gear tho. The question was about H264.

[Entropy3XD]

The original post

I was speaking with someone at Polycom regarding H.264 and was told that it works up to speeds of 512K. In that past everything I've ever read has been that H264 caps at 384K. Can someone give me a reality check here?
Yes, the question was about H.264 but the original poster only had only Polycom to reference. In this thread it was stated there is no cap on H.264 and Sean and Kevin mentioned their own equipment as another reference to show this as fact.

[trapehzoid]

Either way.. its the trend my reply was into mostly, not the bits and bytes.

I'm wondering how much of these replies are generated from macros...

[Glen Sykes]

Feel the love people :knockedou

If the manufacturers aren't welcome to contribute how they are applying their technology for the implementation of H.264, then how are we supposed to have a decent discussion?

Something that strikes me already from this thread is the vast difference in bandwidths that polycom and codian are capable of operating at with H.264.

Add Tandberg into the mix with support of H.264 up to speeds of 768 and we see a picture of processing capabilities unfolding.

I'm also interested in the comments about decoding of H.264 requiring the most processing power, and would like an elaboration on this.

Finally, and somewhat ironically, I notice that in Trapezhoids first reply to this thread that he says:

"as the power levels increase in the systems, they will be able to implement h.264 at higher speeds and actually get some benefit from it."

Which I find rather an odd thing to say given that major vendors already support H.264 at 384Kbps, and Trapezhoid doesn't seem to think that vendors should be focusing their efforts on higher speeds, such as 1Mbps.

.....and he accuses me of contradicting myself :D :rolleyes:

[mazzarak]

group hug everyone group hug!

anyway Trapehzoid, would you feel happier if I said that Tandberg and Codian could do it? ;)
having said that, I also think you have the right to question the tone of posts on the forum - just as long as the abuse is kept nice and polite ;) I don't think the guys in question are trying to sell anything, they have sales people to do that ;)

[trapehzoid]

You guys missed the point. its not about vendors posting here..

I guess I'm alone here.. but if someone wants to ask a question about what H.264 can do.. I don't need a vendor's reference for what they are doing now (besides.. they only implement a very narrow portion of H.264).. as its very straight forward that the info the guy was getting was bunk to start with.


Which I find rather an odd thing to say given that major vendors already support H.264 at 384Kbps, and Trapezhoid doesn't seem to think that vendors should be focusing their efforts on higher speeds, such as 1Mbps.

.....and he accuses me of contradicting myself :D :rolleyes:

Uh.. read the words 'get some benefit from it'. A point you constantly seem to miss.. do something for some benefit, not just to do it. You can do H264 at these higher speeds in many products today.. but is it worth it? Do you get notically better quality? As I said.. as the power of the systems continues to increase.. then we'll start seeing the benefits. At the high speeds today H264 isn't delivering a significantly better picture. H264 is nice for the flexibility but of limited use today at high speeds.

[Entropy3XD]

group hug everyone group hug!
LMAO. Thanks amigo.........now I got a warm fuzzy. I needed that on a Monday. http://www.vtctalk.com/images/smilies/banana.gif

[Yukikaze]

I'm also interested in the comments about decoding of H.264 requiring the most processing power, and would like an elaboration on this.


Decoding any standard based on prediction will never require more processing power than encoding, assuming the encoder actually considers the reference pictures when encoding(If it does not it will achieve lousy compression). The reason for this is that the encoder actually has to do the decoding transforms and filters to form its reference images for further prediction. In effect an encoder does decoding also.

However, decoding is a much larger part of a reasonable H.264 codec than it used to be in H.261 and H.263.

I'd also like to mention that the difference in perceived quality between H.263 and H.264 deminishes as bandwidth increases. This is the main argument for falling back to H.263 at some point on the bandwidth scale. You simply get less and less in return for the extra processing power spent encoding the more complex H.264 as you increase bandwidth, and this processing power could be used to run H.263 at higher resolution or framerate instead(a choice you might not have with the more demanding H.264 due to processing requirements), in effect making H.263 a superior choice from a perceived quality standpoint even though it is worse at compressing each individual image.

[Kevin]

I'm wondering how much of these replies are generated from macros...

And here is another reply from Kev's Macro... :D

Sorry guys - I certainly don't want to cause any offence on these forums by posting. All views that I share here are my own and not Codian's. However, given, that I work for Codian it would be kinda surprising if my posts in general weren't somewhat based on or biased towards Codian... Just as I would find it weird if Sean's posts were not in general based on TANDBERG kit. However, given these are the products we know best about surely it makes sense if when giving examples we base them on our products. Its not as if we hide where our allegiences lie - just take a look at the left <--- :) I'm sure the users on the forums can all make their own decisions about our posts.

Anyway - back the real business of this board... I posted on this thread to highlight 2 points. The first was to give another example of equipment that can do H.264 above the speed that was posed in the original question.

The second was to highlight that care must be taken - just because you are using H.264 does not mean you are going to get better video quality, even at lower bandwidths - users need to test this out for themselves with their equipment to see.

However, decoding is a much larger part of a reasonable H.264 codec than it used to be in H.261 and H.263.

Given that H.264 decode takes much more processing power older underpowered endpoints that have had H.264 "enabled" on them can be left with too little processing power to do a decent encode (as they have no choice but to do the decode fully).

This is something that we noticed in the early days of H.264 (of course the endpoints in question code may have been better optimised since then).

I don't really put much weight into william's analysis ( isn't that another codian guy?) given you don't have insight into how the units are actually operating.

Lets agree to disagree about this. I would have thought that coming from a video conferencing equipment manufacturer would actually give unique insights to how video conferencing equipment operates. (I was one of the guys that designed the hardware of the Codian MCU - hopefully that has helped with my understanding of processing requirements too.)

Anyway - My advise to users out there who are looking into using H.264 on older equipment - do back to back tests with H.263 to check you are actually getting better quality - and check that both ends are seeing better quality, as you *may* be better off sticking to H.263 at any bandwidths.

[Glen Sykes]

Uh.. read the words 'get some benefit from it'. A point you constantly seem to miss.. do something for some benefit, not just to do it. You can do H264 at these higher speeds in many products today.. but is it worth it? Do you get notically better quality? As I said.. as the power of the systems continues to increase.. then we'll start seeing the benefits. At the high speeds today H264 isn't delivering a significantly better picture. H264 is nice for the flexibility but of limited use today at high speeds.

Please enlighten me, why else would you run H.264 at higher bandwidths if it wasn't to experience better quality video?

And yes, this does take us back to the Lifesize thread, but hey, I've SEEN the lifesize product in action, using H.264 at higher bandwidths, and guess what...the video quality is better. HD resolution with no deterioration in visual quality.

What I struggle with here is that you refer to benefits of running H.264 at higher bandwidths, but flatly refuse to acknowledge HD as an improvement, and that this improvement will deliver benefit!

Put simply, IF a manufacturer has the processing power to use H.264 to deliver significantly better video, be it at the same or higher bandwidths than existing technology, then surely the person sat in front of the system benefits from this.

Given the limits in processing power on current vendors endpoints, and the factors so excellently explained by Yukikaze, it is no surprise that manufacturers switch back to H.263 at higher bandwidths.

[trapehzoid]

I don't get how you think I'm contradicting myself.

If you drink the LS koolaid then what I'm saying follows exactly.

I said "You can do H264 at these higher speeds in many products today.. but is it worth it? Do you get notically better quality? As I said.. as the power of the systems continues to increase.. then we'll start seeing the benefits"

LS claims to have that power.. so yes.. if they do.. then yes.. running H264 at high speeds would be benefical. (which is exactly what I said above).

But me.. I don't drink the koolaid.. and I don't define the rest of the market by one product. So I stand behind what I said.. there is no point of doing H264 if you are doing it just to do it. Do it for a REASON. If you get better picture from H263 vs H264.. don't do H264. It shouldn't take a scientist to figure that one out. If you get a better picture on your system from doing H264.. use H264. Today as it stands because of the wide varying implementations and power of the systems.. there is no black and white answer on when to use it or not.

Fact is though.. just like HD is doing now.. the last 12months people are insisting on H264 just for the sake of H264.. ignoring if they are actually getting any benefit from using it.

I'm not flawing H264.. I'm saying don't use it unless your system gives you some benefit from doing it. In the case of LS.. they use it to get the image format they want and still be standards based, so it makes sense to use it at those speeds.

Also.. a system built to do H264 at one speed, may not be all that good at another speed.. even if it has the power to do it. The software must be designed to do, and to do it well takes attention. Call it 'tuning'

[Glen Sykes]

I think that the important point you make here is defining the market by one product. You're absolutely correct with that.

I'm at risk of labouring the point here so I'll try to keep this brief. The reason I feel you contradict yourself is that your whole position thus far on HD / Lifesize seems to be that it is gratuitous, unneccesary and that it wont be of much benefit to anyone. I take this from the general theme of your posts.

However, the contradiction is where as I put before, on the one hand you say "as the power levels increase in the systems, they will be able to implement h.264 at higher speeds and actually get some benefit from it", and on the other you imply that HD resolution at 1Mbps is gratuitous and unnecessary. Which is it to be?

If Lifesize, or any other company for that matter can deliver videoconferencing experience that is measurably better than what they are experiencing today, irrespective of whether it's H.263 or H.264, then surely this is a benefit to the end user! It's as simple as that.

Something I WOULD agree with is there is no point doing H.264 for H.264's sake. The current endpoint vendors implementations of H.264 makes sense by having a cutoff point, because at higher bandwidths there is no benefit, and taken too far, actually decreases the quality. It could be argued that Codian's implementation of H.264 at 2Mpbs is gratuitous, given that no endpoint available today is capable of running this. To be honest though, I think that's just Codians' way of showing off their processing potential.

[Sean Lessman]

Add Tandberg into the mix with support of H.264 up to speeds of 768 and we see a picture of processing capabilities unfolding...

Or its possible some manufacturers just haven't written the software yet because most people use H.264 at lower speeds thereby bumping other features up on the priority list. Its not as black and white as you may believe.

The TANDBERG endpoints do up to 2Mbps H.264 and the MPS does up to 768kbps today. The endpoints used to do only 768kbps and in F2 software this was increased to 2Mbps. As we rarely see anyone doing 2Mbps in a Multipoint call, there are a few higher priority features for the MPS right now.

Sean

[Glen Sykes]

Cheers for the update Sean, that one must have slipped by me!

glen

[trapehzoid]

I'm at risk of labouring the point here so I'll try to keep this brief. The reason I feel you contradict yourself is that your whole position thus far on HD / Lifesize seems to be that it is gratuitous, unneccesary and that it wont be of much benefit to anyone. I take this from the general theme of your posts.

You must be confusing my posts with other people's from wainhouse or something. My beef with the LS messaging is (and as I've posted here)

- I don't believe they can deliver HD quality at the speeds they are claiming within the acceptable norms people have of INTERACTIVE VTC today. I do believe there are tricks implemented to try to implement those types of resolutions at those speeds. If they are so high on HD quality, why didn't they show something like a Hollywood movie at Infocomm? I know it wouldn't demostrate the camera, but how about showing us what you can do?
- I don't believe doing mass market VTC at 1+mbit today is viable. I don't care what all these people are saying 'well I can do 1mbit'.. sure one guy at a time. Now lets get 10 calls going on your internet pipe.. now what? With the level of problems people have getting their networks to perform for VTC at 384+kbs in most situations.. how in the world do you think they are going to get 1-5mbits working reliably? 'bandwidth is cheap' is a cop out.. BW has been cheap for along time.. do you see the problems going away recently? I don't.
- I don't believe HD is the market catalyst nor do I think it helps the directions people have been driving which is cheaper, more accessible systems.

That's my beef with the LS groupies.. not if I think HD is beneficial or not. (tho in my last point I don't think HD is worth anything in the desktop space)

However, the contradiction is where as I put before, on the one hand you say "as the power levels increase in the systems, they will be able to implement h.264 at higher speeds and actually get some benefit from it", and on the other you imply that HD resolution at 1Mbps is gratuitous and unnecessary. Which is it to be?

See above.. I haven't argued that at all.

To be honest though, I think that's just Codians' way of showing off their processing potential.

Good for them. I find all the HD claims interesting and we'll just see how they'll play out.

I kinda think of it like interlaced video.. it has obvious benefits, but required so much more bandwidth, and without the ideal bandwidth typically delivered a poorer image clarity wise. Support for it with features in MCUs, etc was limited.. it became a 'boxed in' solution and how many people do you see using it or demanding it now?

I think HD is going to play out the same way. Vendors will use it to try to differentiate between each other.. without any real practical uses for most end-users (yes, some will benefit from it.. but not the masses). Eventually everyone will have some sort of implementation.. and regardless of who is better, etc.. most will not using it and it will fade away as another checkbox most people don't care about.

Another way to look at it would be.. If Polycom annouced HD capability in the VSX8000 with the same feature sets without LifeSize even existing.. would we all be having these same conversations and so many people acting like the world has changed? I don't think so. Its more of the 'chosen ones' thing and the fact the new kids are betting their whole marketing strategy on it.

Just my .02 from a lowly guy who has to use the stuff.

[Glen Sykes]

Thanks for the clarification,

I'd agree that at present, mass market 1Mbps is not feasible, business to business. That will change though, as I've argued previously. Bandwidth is becoming commoditised, and whilst the carriers continue to milk businesses for providing them with bandwidth, the price will continue to drop. This will be further driven by businesses that introduce applications that demand bandwidth, such as regular VC, VoIP, streaming video e.t.c. The major point of debate here will be when it happens. Carriers will make their money by providing other services I suspect.

I really don't think that when the change does happen, there will be much of a difference between a 384K stream and a 1Mbps stream as a proportion of the overall bandwidth.

In the meantime, the demand for HD will be driven by the lift in what is perceived as acceptable quality by the end users. They will see a dramatic change when HD hits the home, they will also see they change when LifeSize kit starts being demonstrated.

I beleive that you're entirely correct that these factors will not tip the market on it's head overnight, but it does mark the beginning of the end of CIF as the acceptable video quality in the meeting room.

On your technical appraisal of what LS are doing, we'll just have to agree to disagree on that point, the proof of the pudding will be in the eating. This one won't be ready to eat until september :happy:

[trapehzoid]

Or maybe later? :)

I thought their management software was released.. why did wainhouse say 'post-beta' was being shown at infocomm? what the heck is 'post beta'?