Netflix Streaming Coming to PS3 Next Month

[patsu]

patsu, the point is the needs ARE that simple a huge amount of the time in the BD-J space. You want subtle tweaks, simple menu systems, etc. If you need to load the JRE and then you also need to "install" the code to transform it to native code first, that's considered overhead. It's also considered by some to be overkill.

The overhead is always there as long as you need to deploy it to multiple platform though. It's just that Java has its own way of doing it. Native C has its own overhead to deal with multiplatform deployment.

The menu systems loaded rather quickly after the initial loading. Again, based on the above example, the base VM loading is only sub-second on a PC. It's hard to conclude that it alone contributes to 30 seconds to 1+ minutes of load time on the BR players.

Remember there are complex DRM and graphics heavy UI elements to load from Blu-ray too (during startup). Based on the Samsung player vs PS3 comparison, there may be system design factors that may affect Java loading as well.

 

[Asher]

The overhead is always there as long as you need to deploy it to multiple platform though. It's just that Java has its own way of doing it. Native C has its own overhead to deal with multiplatform deployment.

The menu systems loaded rather quickly after the initial loading. Again, based on the above example, the base VM loading is only sub-second on a PC. It's hard to conclude that it alone contributes to 30 seconds to 1+ minutes of load time on the BR players.

I've never said that alone contributes to the loading time.

The loading time is a mixture of VM loading (which will be significantly slower on the less-capable set top box hardware), JIT compilation (which will be significantly slower on the less-capable set top box hardware), code loading from disc (Java code is not known for being small in size), and initialization logic (Java is not a simple language, BD-J is not a simple spec).

And you are correct that there will always be overhead if you need to deploy on multiple platforms. In HD DVD, the js/DOM runtime needed to be loaded. Incidentally, these discs loaded far, far faster than Bluray discs which may play into my earlier statements that Java is too heavy and overkill for what it's mostly used for on Bluray.

 

[tuna]

I'm perplexed why you keep coming back to this "only on one vendor solution". The Netflix Bluray disc only works on PS3s, an HDi implementation would work on every HD DVD player ever made. To the end user, what's the difference here?

Wasn't every HD DVD player ever made manufactured by Toshiba or used their SW?

 

[bkilian]

You worked only on the 360 HD-DVD peripheral, right? Do you know how this worked on less beefy hardware, like Toshiba's players?

I worked on every HD DVD player from Toshiba, as well as the XBox player. Toshiba's players varied from a 2GHz P4 down to a 300MHz embedded chip. I'm most familiar with the XML and Animation stuff because I helped refine the spec in that area, and I wrote the rendering and animation compliance tests that every player had to pass before being certified. (I also helped with the JS compliance tests, but that wasn't my primary area)

The entire HDi core on the XBox (Rendering, Animation, Javascript etc) ran on a single hardware thread on one core of the 360, and shared time with audio decoding/encoding, disc reading and a bunch of other functions.

On the Toshiba Players, depending on version, it either ran as part of the main player process or on a second core that communicated with memory and the main core through a proprietary Toshiba bus, or directly on the second core of the same chipset Broadcom sells to the Blu-ray companies. I believe the last version may never have shipped as an actual product.

Performance was pretty good though, we never bothered doing a lot of comparison testing on that point. We were mostly concerned with making our code run well, not with how well other code ran.

Netflix disc on other BD 2.0 players? Working? Not Working? So far we have one response of "Tested with no network connection, will update later"

 

[tuna]

The default jre that would be installed is icedtea (http://icedtea.classpath.org/wiki/Main_Page) if you don't specifically configure your system for the Sun JDK.

Is there anything wrong with icedtea? BTW, on fedora it is called OpenJDK.

 

[bkilian]

Wasn't every HD DVD player ever made manufactured by Toshiba or used their SW?

Nope, LG had a dual player with their own code. They never shipped the second version, but it had full compatibility. And there were two PC software solutions that used their own implementations.

It's true that all the Toshiba players and the XBox used the same core interactivity code, but each version was significantly different, due to performance restraints, and the Javascript engine on the XBox is not the same as on the Toshiba players. In fact, the JS engine was different on different versions of the Toshiba players. And by different, I'm not just referring to subtly, I mean they are entirely different codebases, not even written by the same company.

 

[makattack]

Is there anything wrong with icedtea? BTW, on fedora it is called OpenJDK.

Only problem with OpenJDK is that it is a different JVM implementation with bugs unique to itself, different from bugs in the SunJDK.

Applications that I have that work in the SunJDK have different side-effects on OpenJDK.

Unfortunately, these are also commercial applications that cannot legally be modified by the end-users to work on different JDKs, and also come with support contracts that stipulate they will only support problems if it's run in a certified environment, e.g.: RedHat ES4/5, Sun Java 5/6, etc... or even include their own JREs which are usually either Sun, IBM, or Oracle JVMs.

At any rate... if I work on a product that embeds the IBM JRE, obviously, I am forced to use the IBM JDK when developing extensions, etc for that product. If I'm working on a product that includes the Sun JRE, obviously, I use the Sun JDK... I have just about every JDK/JRE installed under the Sun (sorry for the word play) on my work place workstation, which has a based OS of Gentoo -- my obvious personal choice, and virtual machines (under KVM) with RHES 4, RHES 5, Solaris x86 -- obviously a heterogeneous nightmare, and all in the name of developing and testing for a single, inter-operable Java environment.

I used to use and like the IBM Jikes java compiler, but it can produce some strange byte-code.

 

[Arwin]

The curse of web-based technologies, in other words.

 

[wco81]

Anyone know of any data indicating how widely these interactive features are used?

Maybe the studios think these will make people buy discs rather than pirate them but why there was such contention between companies back then (and why there's such an argument in this thread) about the technology used seems unfathomable if as I suspect, few people bother to really delve into these things.

 

[NavNucST3]

...

Netflix disc on other BD 2.0 players? Working? Not Working? So far we have one response of "Tested with no network connection, will update later"

Never spits back any errors on my sons BD-1500, goes through each step, seems to connect to Netflix, (you can see the bright red background with Netflix on it for about 1/3 of a second) and then just gives me a black screen. His player has one firmware update to run after that I can try again..."later". I have a 2GB mini-sd for persistent storage on the back of his player.

 

[patsu]

I've never said that alone contributes to the loading time.

The loading time is a mixture of VM loading (which will be significantly slower on the less-capable set top box hardware), JIT compilation (which will be significantly slower on the less-capable set top box hardware), code loading from disc (Java code is not known for being small in size), and initialization logic (Java is not a simple language, BD-J is not a simple spec).

I think you forgot to include the time to load the actual movie resources, and also the time for the Java program to process a BD-J movie.

According to the blu-ray.com benchmarks, the basic movie loading -- without Java -- contributed to the bulk of the loading time. The 500MHz MIPS 24Kc Samsung player took 30 or so time units (About 58%) to load the basic movie, and the other 21 time units (42%) to load _and_ "decode" the BD-J movie. On the 3.2GHz PS3, it took 30 time units (86%) to load the movie resources, and 5 time units (14%) to load and run the BD-J encoded movies. The standalone players halved their load time since the initial models. Let's see if they can improve their times even more.

To decode a BD-J movie, the Java program needs to load itself, perform security operations, decode UI images (for the main menu), and run any time synchronization logic.

And you are correct that there will always be overhead if you need to deploy on multiple platforms. In HD DVD, the js/DOM runtime needed to be loaded. Incidentally, these discs loaded far, far faster than Bluray discs which may play into my earlier statements that Java is too heavy and overkill for what it's mostly used for on Bluray.

Different hardware though. The HD DVD player someone tore down had a (desktop) Pentium 4 2.4GHz CPU and whatever graphics processor(s) to accelerate HDi graphics operation and render movie (as alluded to by bklilian earlier, also mentioned in the tear down). Like the PS3, the early HD DVD players were believed to be sold at a loss. If so, they should be better spec'ed than regular standalones.

Not sure what security measures HD-DVD used besides AACS. May not be an apple to apple comparison.

Incidentally, check out http://hdcookbook.com/ for more references on BD-J

 

[patsu]

Anyone know of any data indicating how widely these interactive features are used?

Maybe the studios think these will make people buy discs rather than pirate them but why there was such contention between companies back then (and why there's such an argument in this thread) about the technology used seems unfathomable if as I suspect, few people bother to really delve into these things.

The only one I found is vague:
http://www.thedigitalbits.com/mytwocentsa173.html#110509

All studios are going to continue trying new BD-Live applications and technologies. While these have less appeal to early adopters of Blu-ray (who tend to be more movie-content interested), as Blu-ray moves into the mass market these features will have more consumer appeal.

Deluxe reports that 50 million BD discs will be connected to the Net via BD-Live by the end of 2009.

Sony reports that once people connect to the Net via BD-Live, 50% of them come back and reconnect at least once.

In any case, that statistics has just been skewed by NetFlix's BD-Live instant streaming.

 

[bkilian]

Different hardware though. The HD DVD player someone tore down had a Pentium 4 2.4GHz CPU and whatever graphics processor(s) to accelerate HDi graphics operation and render movie (as alluded to by bklilian earlier, also mentioned in the tear down)

No, the A1 only had a single processor, the P4. It was also the slowest of all the HD DVD players in loading time. This was due to the AACS negotiation on the drive I believe. The A3 loaded in 15 or 20 seconds(from disc insertion), and it had the slowest processor of all the players.

Most of that time is in setting up the video, as mentioned before. Application load time is a tiny fraction of total load time.


It's interesting hearing that the BD disc doesn't seem to work in other players (so far). So much for "write once, run anywhere". I wonder what the disc is doing that is specific to the PS3?

 

[patsu]

Ah, might want to correct them ?
http://www.informationweek.com/news/global-cio/showArticle.jhtml?articleID=189601107

The Toshiba HD-A1 is basically a combination of a low-end PC and a high-end DVD player," said Andrew Rassweiler, teardown services manager and senior analyst for iSuppli, in a statement.

The HD-A1 utilizes a general-purpose microprocessor instead of more cost-effective application specific standard product (ASSP) semiconductors typically used in consumer-electronics products, iSuppli said. The HD-A1 also employs an Intel Corp. Pentium 4 as the main microprocessor, as well as Broadcom Corp.'s BCM7411 for high-definition video decoding and four ADSP-2126x SHARC programmable DSPs from Analog Devices Inc., according to iSuppli, which estimated that the total cost of these chips is $137.

The HD-A1 also uses $125 worth of memory, including a 1-gigabyte dual inline memory module (DIMM) from Hynix Semiconductor Inc., three other types of DRAM, a 256-megabyte flash memory disk from M-Systems and 32 megabytes of MirrorBit flash memory from Spansion, iSuppli said. Adding the memory chips bring the total cost of ICs used in the HD-A1 to roughly $247 per unit, according to the firm.

You mentioned HDi supported hardware acceleration ? Which operations were supported ? What processor and clock speed were in the A3 ?

EDIT:
According to here: http://www.avsforum.com/avs-vb/showthread.php?t=936118, A2 is the fastest; A3 and A30 became slower than A1 though.

--until disc is read and "hd-dvd logo" appears on screen:
A2 (firmware 2.5): 32 sec.
A3 & A30 (firmware 1.0): 50 sec.
A3 & A30 (firmware 1.1): 33 sec.

--Additional time until first screen of disc is played:
A2 (firmware 2.5): 16 sec.
A3 & A30 (firmware 1.0): 25 sec.
A3 & A30 (firmware 1.1): 30 sec.

----UPDATE:
The 2.7 A2 firmware speeds boot time about 2-3 seconds from v.2.5.
The 1.3 A3/30 firmware speeds boot time about 5 seconds from v.1.1.

 

[eastmen]

Ah, might want to correct them ?
http://www.informationweek.com/news/global-cio/showArticle.jhtml?articleID=189601107



You mentioned HDi supported hardware acceleration ? Which operations were supported ?

Why 1Gb memory ? The BD-2.0 players mostly just use SD cards to satisfy the 1Gb persistent memory requirement.

I like how it took Bluray a few revisions to finally catch up in features with hd dvd.

Would have been nice to see a new revision to hd dvd with 3 layer discs and other features introduced.

I'm sure we can look back in 5 years and see hwo awsome BD 4.0 spec players worked.

 

[sethk]

I got my Netflix PS3 disc on Friday and played with it a little this weekend. The browsing is quite sluggish and the player is just OK. The Xbox 360 version is a lot smoother by comparison, and loads faster and doesn't require a disc. Don't think I'll be using the PS3 for Netflix.

 

[eastmen]

I got my Netflix PS3 disc on Friday and played with it a little this weekend. The browsing is quite sluggish and the player is just OK. The Xbox 360 version is a lot smoother by comparison, and loads faster and doesn't require a disc. Don't think I'll be using the PS3 for Netflix.

i'm sure the next revision will be better. Remember how bad the original was on the 360

 

[deathindustrial]

It's interesting hearing that the BD disc doesn't seem to work in other players (so far). So much for "write once, run anywhere". I wonder what the disc is doing that is specific to the PS3?

I thought this was strictly a licensing issue as opposed to a technical limitation? Unfortunately I am having a hard time finding an implied reference to that effect but was sure I saw something along those lines in one of the Netflix interviews. . .

 

[bkilian]

Ah, might want to correct them ?
http://www.informationweek.com/news/global-cio/showArticle.jhtml?articleID=189601107



You mentioned HDi supported hardware acceleration ? Which operations were supported ? What processor and clock speed were in the A3 ?

EDIT:
According to here: http://www.avsforum.com/avs-vb/showthread.php?t=936118, A2 is the fastest; A3 and A30 became slower than A1 though.

I don't have to correct them, their statement is correct, you're just misinterpreting it. The P4 is the only CPU. The video processor is a single stream hardware H.264/MPEG2 decoder. Secondary streams (PiP) were software decoded on the P4.

As far as hardware acceleration, It was the usual 2D primitives you used to find on the GPUs in the early 90s.

CPUs, it's been a while, but the A1 was a 2.4GHz P4, the A2 series was a NEC EMMA3 with a Celeron 900 slave CPU, and the A3 series used the EMMA3 with a NEC 5701 MIPS slave CPU. I think it ran at around 300MHz.

Yah, looks like the A2 was the fastest, but I think that was again mainly due to drive firmware. I think the A3 drives were quite a bit cheaper than the A2 and XBox drives. The A1 was still definately the slowest though, 45 seconds easily from disc insertion to first video.

 

[bkilian]

I got my Netflix PS3 disc on Friday and played with it a little this weekend. The browsing is quite sluggish and the player is just OK. The Xbox 360 version is a lot smoother by comparison, and loads faster and doesn't require a disc. Don't think I'll be using the PS3 for Netflix.

Don't sell them short, it's an amazing accomplishment considering that it's just a BD-J app. The XBox app is a native app (I think) with all the advantages associated with that. The PS3 version is written in a VM based language, in a sandbox designed for providing small interactivity apps and menus for disc based movies.