CPU For Transcoding?

[KarylFStein]

I currently have a Q6600 CPU with 8G of RAM in my SageTV server. This set-up has worked fine for me for a while, but it seems like I’m now having transcoding issues with playback of CableCard captures in Placeshifter. I don’t know if CableCard is harder on the transcoding than the OTA and analog captures I did before, but that seems to be the culprit.

I’m looking for suggestions on a CPU that can for sure keep up with transcoding, (I may also try streaming to Android devices). The server also runs comskip and some other low-impact things, (except even HD300 playback can stutter when the server is doing backups, but I’m guessing that’s more of a disk/network thing.)

Cost is a big concern as I’m sure I’ll have to also upgrade the MB and RAM, (unless going with a Q9550 would do it--those don't seem too bad on eBay). If I do have to move off of Socket 775, I have no problems using something a generation or two back that does the trick and is a value in the used market.

[Fuzzy]

Placeshifter just doesn't use a very modern transcoder. It would be nice if sage would release their source changes they made to ffmpeg to turn it into SageTVTranscoder.exe. There have been quite a few updates to the core ffmpeg code since the last SageTVTranscoder build a couple years ago.

That said, the Q6600, as I'm sure you realize, is quite a ways behind even mid-range cpus of today. A mid level i5 or Phenom II is going to out perform the Q6600, and likely use a fraction of the power.

For reference, look at the Q6600 on this list: http://www.cpubenchmark.net/high_end_cpus.html

[KarylFStein]

It looks like I can build a Phenom II X6 1045T-based system for ~$200, (CPU/MB/RAM). I hate sinking any more money into the server, though, especially when not 100% sure that it's causing my Placeshifter issues .

Does anyone know if CableCard captures, (Ceton InfiniTV 4 if that makes any difference), are harder to transcode? And if I were having transcoding problems, would I expect to see the CPU usage on the server really high, (it's not)?

[Fuzzy]

Originally Posted by KarylFStein

It looks like I can build a Phenom II X6 1045T-based system for ~$200, (CPU/MB/RAM). I hate sinking any more money into the server, though, especially when not 100% sure that it's causing my Placeshifter issues .

Does anyone know if CableCard captures, (Ceton InfiniTV 4 if that makes any difference), are harder to transcode? And if I were having transcoding problems, would I expect to see the CPU usage on the server really high, (it's not)?

Cablecard should not be any harder than anything else to transcode It's really just MPEG-2, and it should actually be easier than ATSC, based on it potentially having a better signal strength, minimizing corrupted packets for teh decoder to have to deal with.. That said, the X6 may not help you any. SageTVTranscoder is single-threaded, so having 6 slower cores is going to be worse on the placeshifter transcoding front than even 2 really fast ones would be. The Phenom II's are still a good choice for this, as when running a high demand single threaded load, they will shutdown 3 of the 6 cores, and run the other three at a higher clock speed (3@3.2GHz vs. 6@2.7GHz). Looking at the single-thread, single clockspeed benchmarks here, it looks like, clock-for-clock, there hasn't been much change in the Phenom II cores - so you may get better bang-for-your-buck with a little older gen quad-core (Deneb based chips are still reasonably available, and may be quite a bit cheaper than the Thuban line - unless you find a good sale on the Thuban).

Overally, the Thuban is going to use less power than the Deneb - so it may end up washing out the savings. And either are going to be faster than that old Core2Quad.

[Fuzzy]

Of course, for not a lot more, you can possibly get something like an FX-4170 for $120, and break the 4GHz level... Of course, all these choices are what I believe keep AMD much more appealing to me. You can buy an AM3/AM3+ motherboard now that can run anything from a 3 1/2 year old Sempron 140 2.7GHz single-core, to a brand new FX-8350 eight-core 4.0GHz chip...

[KarylFStein]

For some reason I bought a Xeon E3-1235 platform. It's not just for SageTV, (really Minecraft for my kids and their friends which is a hog), but SageTV is really all that's on there now. No change in Placeshifter. So it's gotta be something with the transcoding or recording path as the CPU is out as a problem as far as I'm concerned *sigh*.

[Gustovier]

FYI.. As ffmpeg is open source Sage did provide the source to the changes they made to it. I never got a chance to really dig into it. I was wanting to do the same thing and update to a newer version of ffmpeg. But anyways I do have the source.

[FlyingShawn]

Originally Posted by Gustovier

FYI.. As ffmpeg is open source Sage did provide the source to the changes they made to it. I never got a chance to really dig into it. I was wanting to do the same thing and update to a newer version of ffmpeg. But anyways I do have the source.

Do you have a link to that? I don't have the experience to do anything with it myself, but perhaps someone else in the community does.

[Gustovier]

This is the source to both ffmpeg (sage transcoder) and mplayer.

http://download.sage.tv/ffmpeg.zip

http://download.sage.tv/mplayer.zip

[Fuzzy]

Originally Posted by Gustovier

This is the source to both ffmpeg (sage transcoder) and mplayer.

http://download.sage.tv/ffmpeg.zip

http://download.sage.tv/mplayer.zip

gonna take this a bit off topic, but really it doesn't look like he did much in the way of changes, only adding another avformat for the placeshifter streaming. It looks like if you don't do placeshifter, and just transcoding, then you may very well be able to simply drop in the new ffmpeg binary, renamed to SageTVTranscoder.exe, and get the better conversion speed - but it will break placeshifter.

[Gustovier]

Originally Posted by Fuzzy

gonna take this a bit off topic, but really it doesn't look like he did much in the way of changes, only adding another avformat for the placeshifter streaming. It looks like if you don't do placeshifter, and just transcoding, then you may very well be able to simply drop in the new ffmpeg binary, renamed to SageTVTranscoder.exe, and get the better conversion speed - but it will break placeshifter.

The modded ffmpeg also sends status of the conversion to sage as well. Not quite sure how it does that. I think maynne through another output stream.. They added in a few more command lime args from what I remember as well (haven't looked at it in about a year)

[Fuzzy]

Originally Posted by Gustovier

The modded ffmpeg also sends status of the conversion to sage as well. Not quite sure how it does that. I think maynne through another output stream.. They added in a few more command lime args from what I remember as well (haven't looked at it in about a year)

I couldn't find any indication of that. I'm guessing they are parsing the console output in some way to convert to progress.

[FlyingShawn]

I know so little about programming that I'm not even going to try looking at the code, but it sounds like the two of you have a somewhat-reasonable grasp of what you're seeing.

Is this as tediously-simple as finding the original ffmpeg version Sage was working from and comparing them side-by-side to find the changes? Could those changes be then copied over to the current ffmpeg code to make an update and keep Placeshifter? Going forward, would keeping it updated be as simple as following a sheet of "insert this line here" instructions to add the changes for yourself?

[tmiranda]

Originally Posted by Gustovier

The modded ffmpeg also sends status of the conversion to sage as well. Not quite sure how it does that. I think maynne through another output stream.. They added in a few more command lime args from what I remember as well (haven't looked at it in about a year)

IIRC there was another issue with the compiler options. I forget the details but I do remember somebody trying to replace the Sage mplayer with a custom built one and running into issues caused by using the wrong compiler options.