Best possible Netflix streaming device - Page 20

How are you getting the 1080P/5.1 stream on the ROKU 2. I tried 'Ong Bak 2' tonight on my ROKU 2. I didn't see the 5.1 option.

Tomato 10 min average on 'Ong Bak 2' 4.8Mbps - on shows I think are 1080P/5,1 the average is around 5.2Mbps - I tried the Cape Pilot (which I think is 1080P/5.1) and the 10 minute average was 5.2 Mbps see attached

I didn't get 5.1 sound on the Roku--the Roku 2 Netflix player currently has a bug wherein it sometimes does not detect that a title has HD and/or 5.1 sound encodings. The HD thing is no big deal, since you'll get the HD encodings whether it acknowledges that they're there in the GUI or not; 5.1 is an issue, since if it's not there you can't select it (and the Roku 2 makes stereo the default for titles with 5.1 sound ). So the Roku 2 data that I collected yesterday was stereo. According to Netflix, the difference should be 192 Kbps (see this).


Did you snap that at the end of 14 minutes (as precisely as possible)? As I said above, that's what I did to get my results. Since the encoding is variable bit rate, two arbitrary 10 minute segments will have different average bit rates. I selected minutes 5-14 of Ong Bak 2 because it was a long sequence of difficult-to-encode stuff: martial arts combat in the rain, constant fast action with tiny details--raindrops--all over the scene changing from frame-to-frame. Encoders will use as much bandwidth as they're allowed for sequences like that.

The data which I collected yesterday can be seen below, with graphs of data that I collected months back for comparison. (Blow it up fullscreen to see 720p and 1080p side-by-side with the older data beneath the new; sadly, I have no control over scale in Tomato's graphs, so they don't match):

Found some more differences between stream sets. First I'll link again to my thread that shows PC and PS3 are using the same video encodes: http://www.avsforum.com/avs-vb/showthread.php?t=1365060

Months ago, Lady Vengeance was mentioned as having a screwed-up looking encode. At the time I assumed that the PS3 app was doing some sort of zooming/stretching because at any given point you FF/RW to, it will start with full letterbox bars but then jump to screen-filling with only a partial bar on the bottom.

I now see that there are two issues at play: 1) only the lowest quality PC/PS3 stream, 235kbps, is encoded correctly and 2) "Low / SD" is indicated for up to three separate stream levels. I say "up to" because a movie called Final Approach is the only flick I've yet seen with a 375 encode. Most have only 235 and 560 as the lowest levels, both of which are christened "Low / SD".

For both movies, all PC/PS3 streams that are 560+ have the approx. 2.35:1 letterbox bars cropped out and the image stretched to 16:9, though Lady Vengeance leaves some of the bottom bar intact to allow room for mostly-complete subtitles.

Meanwhile, the streams that the Insignia receives are 2.35:1 for all quality levels of Lady Vengeance, and I think all of its Final Approach streams are stretched.

I'll be adding another post to the thread linked above demonstrating all of this pictorially.

The other thing I noticed is that the audio sampling rate for the Insignia streams is 44.1kHz ("CD quality"), while the PS3 puts out 48kHz ("DVD quality"). The PS3 does switch from 48 to 44.1 when playing MP3s so I assume it isn't just upsampling. Unfortunately this isn't something that my Denon AVR reports; I just spotted it while capturing clips to take screengrabs from. Did Netflix's blog ever mention sample rates?

Unless the PS3 data is padded with NULL packets or something, I don't see how that's possible. Here's a side-by-side comparison of PC player bandwidth consumption to PS3, graphs from March for both on the bottom (again, blow it up fullscreen):


The PS3 consumed 20% more bandwidth in those 10 minutes than the PC. If we assume that the PC was sucking down data at 3.6 Mbps (as its diagnostic overlay indicated), then 662.17 second worth of video was consumed in 10 minutes, presumably 62.17 seconds in the buffer unplayed. If we assume that the PS3 was also consuming data at 3.6 Mbps, then the data it brought down in 10 minutes was 796.11 seconds worth, 3:16.11 in the buffer unplayed. I don't believe it. If, instead, we assume that the March PS3 data was at 3.8 Mbps, then it's 682.9 seconds worth, 1 minute, 22.9 seconds unplayed. I we further assume that the new data represents the same 682.9 seconds, then the bit rate was 4.2 Mbps.

Mike- go buy a new WD TV Live and confirm it supports 1080p streaming so I can know whether I should return it or not before my 30days is up.

It looks good at any rate, and I'm able to force 1080/50 out for UK titles which is pretty neat as well.

Well, do we agree that screenshots can accurately demonstrate a frame of the video stream? I haven't found any yet that are different, at any quality level of the titles I've compared. I don't think it's impossible (double negative) that some may differ, so I'll check Ong Bak 2 as well.


I think I understand what you're saying. I don't know what conversions are needed to get from Mb to MiB.

Luckily the PC A/V Stats do show the size of the buffer, in both bytes and seconds, split up into Audio and Video. Does that number match Tomato's report?

Come to think of it, what happens if you pull the internet plug while streaming?

If only it had VUDU and Amazon I'd buy one, to honestly evaluate it as a replacement for the Roku 2--if I liked it, I'd sell the Roku. But I'm not going to buy one just to test it and send it back. Sorry.

If you pack your's up and send it to me with return postage I'll gladly test it and have it back in the mail to you on the following business day (on the same day if I get it in the morning ). I just need it for half an hour to get my data .

Okay, so the PC buffer size is a running tally rather than any sort of full meter of the bytes that have been downloaded.



Doesn't the video buffer size there indicate that it's more like 2.6Mbps than 3.6Mbps?

Incidentally, this has even more bitrate levels I hadn't seen before.

I tried to do that here with 'Ong Bak 2' - but with average of 4.4 Mbps (see attached) I am not sure this is the 1080P stream - it should be as I have 1080P selected on the ROKU 2 and other shows/movies are averaging the 5.2 Mbps netflix say we should be seeing.

I am not sure 'Ong Bak 2' is a good measure.

As I think you have posted in the past Netflix says 'The "Best" setting will use any of the video/audio rates available. Our highest quality files are 4800 kbps (for 1080p HD video) and 384 kbps audio (for 5.1 audio).

10 minutes of 'The Girl Who Kicked the Hornets Nest' seems to be consistent with this - see atached with 5.3 Mbps average.

Both on a ROKU 2XS - both today.

I fired off an email to Dave Zatz and this was his response:

So there you have it. The new WD TV Live is more likely than not 1080p on Netflix. You'd think they'd shout it from the rooftops though. I don't really know what to make of it.

Hmm, Tech of the Hub? I can't find any report by them on the new Live with a brief search. This chart hasn't been updated to include it yet, assuming the "Live" mentioned in the DD+ portion is supposed to be the Live Plus. [EDIT: I see now that the author of Tech of the Hub is the AVS member who said in another thread that it was confirmed; good enough for me I guess.]

I think I've got it, Mike. Pulling the ethernet cable does work to check how big the buffer at the end is. With the PC overlay displayed you can watch it count down to 0 until it finally stops (however, there is 2-3 secs of audio buffer left when the video runs out; who knows how much extra audio the PS3 grabs).

Apart from the different buffer size, the other thing we need to take into account is the audio bitrate. With the PC apparently limited to 64kbps stereo even at 720p, the PS3 with 384kbps 5.1 will suck down 24,000,000 bytes more in a 10-minute period, without the added buffer.

Here are my measurements with dd-wrt, which reports bandwidth in rounded base-10 MBs rather than MiBs. I set the PS3 to Thai stereo and turned off the "CC" track in case that steams some data even though it doesn't appear to do anything. 4:00 to 14:00, and at 14:00 I pulled the plug.

I think that's pretty darn close since my input number isn't precise. The PS3 buffer is more than I expected considering its system RAM is only 256MB.

I am looking at getting a new player for my HT setup since the 360 won't send 5.1. I am reading through this thread but I am impatient so I thought I would post my question. I currently have a Roku XR connected to the TV (old CRT) that my kids use. I realize that it is technically capable of playing 5.1 but can someone verify that it gets this content from netflix? I saw that it was the only old player to get the update to 1080p but I saw nothing concrete about 5.1 content from netflix. I don't recall seeing anything about 5.1 in any of the titles I have played on it, what is a good title to use?

Would I still be better off for sound/picture quality if I bought the Roku 2 XS?

I just realized that everything reported by Tomato is expressed in powers of 2--2^10 for "K", 2^20 for "M" and 2^30 for "G". What an ass (the guy who wrote it)! You have to express rates in powers of 10--the math just gets idiotically ridiculous for no good reason if you don't. Also it just confuses Joe Average--"how can a megawatt be a million watts but a megabyte be more than a million bytes???"

Back to the whiteboard .

No pre-Roku 2 models can do 1080p from Netflix (they might be able to do 1080p from something, just not from Netflix). None of them can do 5.1 from Netflix either, though I know that they can do 5.1 sound from Amazon.

Yes. It can play 1080p encodings from Netflix HD titles and 5.1 sound from Netflix titles which have that, as well as closed-captions from titles which have that.

If you want to see which movies have 5.1 sound (last I looked, all but one of them was HD), try tvandmoviesnow.com--it's new and unpolished (and probably slightly buggy), but it can search Netflix and filter for 5.1 sound. Set the filters and leave the search term field blank and it will show you all titles which match your filters. For example, you can see a list of movies on Netflix with 5.1 sound here.

That's what I was thinking until I saw this post:

http://www.avsforum.com/avs-vb/showt...0#post20804870

I was just wanting some real-world confirmation.

It would be great if the XR would do 5.1 as it has component video and optical audio. I am out of HDMI inputs on my denon and would have to shift things around with the Roku 2.

From this post on the Roku forums (posted 22 July):

There's a thread on those forums (here) for discussion of that announcement; no doubt just a ton of griping.

You can get an HDMI switch--I use one (my BD player and Roku share one of the four HDMI inputs on my AVR). I bought it some years back before my recent AVR upgrade, and used to use it to share one port between 3 devices--I only had four HDMI devices connected to 2 inputs then.

I'm truly confused. I should be able to get million-bytes-per-second by dividing million-bytes by number of seconds, right? If you're right and the PC has 239 seconds of buffered, unplayed content at the end of 10 minutes, then the amount of data brought down in 10 minutes of streaming represents 600 + 239 = 839 seconds. For the PS3, 10 minutes consumption would equal 729 seconds of data.

PC-: 284.17 MiB = 2383.79 million bits / 839 secs = 2.84 Mbps - .064 = 2.776 Mbps
PS3: 341.65 MiB = 2865.96 million bits / 729 secs = 3.93 Mbps - .384 = 3.546 Mbps So your estimated buffer size works to produce roughly the result you expect for my PS3 numbers but not for the PC ones. It was never going to explain why I get so much less data for the PC than for the PS3--that the PC had more data buffered would only make things worse. If I try to make 3664 Kbps

Remember, I reran this test 3 times on Thursday on the PS3 and a couple of times on the PC with essentially identical results each time--the PC numbers are almost identical to the ones I got back in March.

Just for grins and chuckles, I'm going to run the PS3 data over with stereo sound selected (and cc explicitly turned off, just in case).

I don't know that I believe that the numbers that Netflix has been giving (the old numbers of 2.6 Mbps, 3.8 Mbps and the new ones like 3.6 Mbps) aren't an aggregate of video and audio. If they are, it's really assinine of them not to mention the bandwidth of the audio component. 3.664 Mbps is reasonable close to 3.6, but if you say that I'm getting 3.8 Mbps and then deliver 3.992 (+.192) or 4.184 (+.384) that's significantly different.

This discussion, while really interesting, has become a massive rathole with little to do with "Best possible Netflix streaming device". I've PM'd the mods asking that they relocate these posts to the "Netflix streaming quality" thread.

If I understand correctly, you're dividing your streaming results using my buffer size. I'm sure the PC player varies its buffer based on the hardware it's running on so you should try the pull-the-plug test yourself if possible.


I suppose, though I do think determining the quality on each device is relevant.

You're sure? What exactly would be the purpose of that? Why would you use a larger buffer with one machine than with another?

Not sure as in "I have tested this myself", no. I don't really have another machine that's fast enough to try it. In terms of percentage of system RAM it would make sense to use a smaller buffer on lower specs.

Though far from high-spec, my system is a 3 y/o Core 2 Quad desktop (Q8300 at 2.5 GHz) with 6 GB of RAM with a nominally 25 Mbps connection to the network (who knows how much bandwidth I can typically get to local Netflix servers). It seems much more than sufficient to handling a 3.6 Mbps stream.

Isn't that the common use of MByte - the reasons go back into semiconductor memory sizes which are typically powers of 2. So a 1KByte memory is 1024 Bytes - typically organized 32 x 32. Likewise 1Mbyte of RAM is 1048576 Bytes similar for GBytes etc.

(don't me started on Billions - a UK Billion is one million million a US billion is 1000 million - that way the US can have many more billions....)

I am pretty sure when Netflix says 'Our highest quality files are 4800 kbps (for 1080p HD video) and 384 kbps audio (for 5.1 audio)' they are talking about 'digital' kbps

I still don't understand when you are seeing very different rates for 'Ong Bak 2' on the ROKU 2 than I am seeing.....

Maybe it is a rat hole - but I did purchase the ROKU 2 to be my best Netflix player - it's plusses are 1080P/5.1 and it's minusses are incorrect black levels (will be fixed soon) and no 24P out (a big issue for me).

It doesn't make any sense at all to use "digital" numbers (gibi, mebi, kibi) because they're only interesting to computer programmers. Also, when you say "giga", "mega" and "kilo", the common man thinks that you're talking about billions, millions and thousands. It become particularly ridiculous when you're talking about rates. When it's said that the maximum payload of a 6 MHz ATSC channels is 19.39 Mbps, they mean that it's 19,390,000 bits per second. Read the Data Rate Units Wikipedia article.

I'm pretty certain that they're talking about thousands of bits. The math just becomes so much easier. 4800 Kbps = 4.8 Mbps; if we're working in powers of 2, 4800 Kbps = 4.6875 Mbps--ugh.

I'm 53 and I've been programming computers since I was 16 (it was really hard to find the opportunity to learn to program back then, affordable home PCs being some years off). I've spent years with the confusion between mega and kilo meaning 10^6th and 10^3rd or 2^20th and 2^10th. We seem to have largely settled for the base 10 definition and I'm happy with that. If a manufacturer tells you that a disk drive is 250 GB in size, it means that it's 250,000,000 bytes (usually a little less or a little more, since computer disc storage is usually formatted in 512 byte sectors. Internal memory is probably the last hold out--even SD card sizes are quoted in power of 10 units. If you go to manufacturers' pages about disk drives or SD cards, you'll find this footnote all of the place:

1 megabyte(MB) = 1 million bytes.
1 gigabyte (GB) = 1 billion bytes.

I settled on Ong Bak 2 some months back. I wanted something with a 10 minute long heavy-action sequence near the beginning so the 10 minute average would be at or near the maximum average bit rate. I wanted something with a 1080p encoding on all services offering such. I wanted something available on as many services covered by the devices I had as possible: Ong Bak 2 is on Netflix, Amazon, VUDU and Zune--I rented it on Amazon and Zune and tested both.

I've tested various devices with Ong Bak 2 dozens of times and consistently got the same result for each device and results which were consistent between devices--nearly the same result for 720p on the Roku as on the BDT110 as on the PS3 as on the Xbox as on the TiVo Series3, etc (until Thursday, when Roku 2 and PS3 diverged). The Roku 2 and PS3 get the same results for 1080p Netflix.

The only thing that I can think of is that bandwidth on your connection to Netflix's servers was not steady or somehow constrained--your curves are strangely flat.

I ran this test on the PC and got a 239-243 video buffer size as well. I cannot explain why my PC numbers are what they are. The buffering and playing bit rates for 64/3600, rock solid.

OK it looks like data comms uses decimal while I can assure you that the semiconductor industry uses binary. So you are correct here it should be decimal.

So lets see if I understand - you are now measuring Ong Bak 2 1080P/stereo on a ROKU 2 on Tomato at 6.217 Mibps. Which is actually about 6.5Mbps.

This does indeed seem significantly higher than Netflix's comment 'Our highest quality files are 4800 kbps (for 1080p HD video) and 384 kbps audio (for 5.1 audio)' especially as Ong Bak 2 is only stereo. Per your post above 5.1 would add another 192 kbps - presumably taking it up to about 6.7Mbps for 1080P with 5.1.

Or am I misunderstanding?

I just tried Ong Bak 2 again - 10 minutes starting at minute 5 - on my ROKU 2.

Tomato reported ~4900kibps - which seems to be pretty close to the number Netflix would indicate for a 1080P stereo stream (but a little higher than I saw last night).

Very strange.....

As msgohan and I were discussing, that bit rate is higher than the actual bit rate because of "undigested" content in the buffer. On Thursday, I got 4752 Kibps for 720p on the Roku and 4766 Kibps for 720 on the PS3; at 1080p I got 6217 Kibps on the Roku and 6282 Kibps on the PS3, as you can see in those graphs back in this post.

What's the rated speed of your network connection?

7.1 Mbps (probably decimal )

Realistically somewhere above 6 Mbps.

If the various playback devices are buffering ahead - then probably the only real measure is too watch a whole movie and see how much was downloaded.

I am seeing the same on my iPod - Radio Paradise is downloads at about a constant 128kbps - Pandora downloads a burst - stops - another burst - etc.

Good to see Top Gear in your playlist and great to see Netflix has recent shows in HD - I remember seeing Clarkson on the first shows in the UKs in the early 80's