Live TV without buffer

[whyukon]

When I watch Live TV, I rarely pause or rewind. Having the option to turn off Live TV buffering would allow NPVR to work like the old GBPVR "Live TV".

Why would I want to turn off buffering? Two reasons:
1. when watching some sports, I find my Live TV video to be a little bit jittery... not jerky, just not totally smooth. I think this MAY be because the overhead of recording the buffer results in some CPU starvation that results in not-totally-smooth live TV. There could be other reasons, but I did not find this jittery behaviour with GBPVR. This brings me to my second reason for wanting to turn off the Live TV buffer.
2. I use my PC for other tasks (on a second monitor) while watching Live TV. Those other tasks also consume valuable CPU cycles, and read/write from the hard drive. I have encountered situations where my CPU is running consistently between 95% and 100% due to other (unrelated) background tasks running while also trying to watch Live TV. It would help both my Live TV viewing and my other (non-NPVR) processing if I had the option of turning off the buffering.

This seems to me, perhaps naively, as if it should be an easy option to bring back. Since NPVR is actually doing two things at once in "Live TV", i.e. it is both displaying the true live feed (through all the processes that entails) and recording to the buffer (through the processes that entails), is it not fairly simple to just give an option to disable the buffering and its processes?

My system is a Dell Vostro 1320, Core 2 Duo 2.4 GHz, running 32-bit Vista. I am using a Hauppauge WinTV-HVR 950q and viewing and/or recording analog NTSC (cable) only.

Many thanks for considering this.

[mvallevand]

Not wanting to take anything away from the request (I know others have asked) I think the buffering impact on the CPU is minimal and you probably need to look elsewhere for the problem. I have a low end C2D on my Vista system and in a quick test when recording NTSC, the CPU is average 1.48 according to perf mon. In another really good test of the NextPVR CPU buffer impact look at this perfmon capture, of my currently played LiveTV of the Chicago/Phoenex game to my NMT which is HD h264 capture from my HDPVR and also include http web streaming. The numbers speak for themselves.

Martin

[whyukon]

I don't dispute your results, but I do know that they are not the same as mine. I probably should have originally posted my own perfmon results to illustrate. My resource monitor posting clearly indicates almost 24% Average CPU for NRecord.exe, while watching Live TV (NTSC Analog). When I start Live TV, my system's total CPU usage jumps by 40% OR MORE. I know from monitoring this on several different occasions, that this is typical for my system. So, while CPU usage may not be an issue for you, it is for me, and likely is for others, which may be why, as you have said, others have requested this feature (disabling buffering) in the past. BTW, the comskip process that you see here is a postprocessing batch on a tv show that was recorded immediately prior to my switching to Live TV, so don't be mislead into thinking something else is going on... the CPU for NRecord IS for Live TV.

I can see from your Perfmon posting, that the CPU usages for all of your tasks are lower than the corresponding usages of my tasks, so it seems as if your hardware is likely quicker than mine. If you, or anyone else, can suggest any way for me to substantially reduce the CPU usage of NRecord, short of buying different hardware, I welcome the suggestions. Failing that, I respectfully ask to leave my request on the wish list.

Thanks again.
Glen
[ATTACHMENT NOT FOUND]

Not wanting to take anything away from the request (I know others have asked) I think the buffering impact on the CPU is minimal and you probably need to look elsewhere for the problem. I have a low end C2D on my Vista system and in a quick test when recording NTSC, the CPU is average 1.48 according to perf mon. In another really good test of the NextPVR CPU buffer impact look at this perfmon capture, of my currently played LiveTV of the Chicago/Phoenex game to my NMT which is HD h264 capture from my HDPVR and also include http web streaming. The numbers speak for themselves.

Martin

[mvallevand]

Is it possible your tuner is set for software encoding?

Martin

[sub]

Yes, that 28% CPU will virtually all be software encoding the video for analog on the 950Q, which would have to happen regardless of whether it was buffering to disk or not.

[mvallevand]

I can see from your Perfmon posting, that the CPU usages for all of your tasks are lower than the corresponding usages of my tasks, so it seems as if your hardware is likely quicker than mine.

I doubt it I have an E4300 @ 1.8Mhz The reason it looks low is that I was streaming to the NMT which does the hard video decoding. I wanted to show you that the disk buffering was likely less than 1% of real CPU. I expect it would quite easy to lower those numbers with hardware encoding.

Martin

[whyukon]

Thanks. It seems clear that this is partially a hardware issue... the tuner hardware. Sub is, of course, correct that the 950Q is software encoding only. However, I'm probably not the only former GBPVR / now NPVR user who is using software encoding. I do know that under GBPVR without buffering, my total Live TV CPU usage was less than it is now with NPVR. So, then it seems to me (again, possibly naively), that eliminating the buffering WILL gain me back some valuable CPU cycles that are being consumed by NRecord.


I will try to revive GBPVR on my computer to see whether I can come up with some comparative statistics to confirm or deny this hypothesis.
Glen

I doubt it I have an E4300 @ 1.8Mhz The reason it looks low is that I was streaming to the NMT which does the hard video decoding. I wanted to show you that the disk buffering was likely less than 1% of real CPU. I expect it would quite easy to lower those numbers with hardware encoding.

Martin

[whyukon]

Maybe I'm barking up the wrong tree with this, and this has little or nothing to do with Live TV Buffering. Maybe it is other differences between GBPVR and NextPVR. However, for what it's worth, I can confirm a substantial increase in CPU usage (20% or more higher) with NextPVR's buffered LiveTV vs GBPVR's non-buffered Live Preview, on my system which as pointed out by Sub relies on software encoding. I ran the NextPVR test immediately after the GBPVR test, and tried to eliminate as many external influences as possible on the results (i.e. no other programs or unrelated disruptive background processes).

Here's the GB-PVR Live Preview CPU results:
[ATTACHMENT NOT FOUND]

And here's the NextPVR Live TV results:
[ATTACHMENT NOT FOUND]

So, sub, can you clarify, does it seem as if buffering is having a significant impact, or is this entirely due to other differences between GBPVR and NextPVR, or possibly something else.....?

Thanks again!
Glen

[sub]

GBPVR only supported a very limited number of soft encoder devices in it's very last release, and had limited capabilities with the few device it'd work with. For live preview live tv it effectively hooked the capture device directly to the audio/video renderers. NextPVR on the other hand MPEG2 encodes the output of these soft encoder devices to pass it through same logic as used by digital devices and hardware encoder devices, always ready to record/pause/skip etc.

There is zero chance I'm going to change NextPVR to work without encoding soft encoder devices.

[whyukon]

Thanks for the clarification on that, sub. I had just switched from GBPVR to NextPVR a month or two ago, so I apologize for digging up old issues. I had hoped to stick with my Hauppauge 950Q since it worked well enough for me under GBPVR, but if I want to use NPVR maybe I should look for a device with hardware encoding. There's not much selection of tuner hardware available in my part of the world, which is how I ended up with the 950Q in the first place. But that's my problem, not yours!
Cheers!
Glen