First Cell demo (48 MPEG 2 Videos)

Status
Not open for further replies.
one said:
PC-Engine said:
A single threading 3.5Hz P4 will be able to do at least 12 streams with audio like aaaaa00 mentioned in the other CELL thread.
Unless it's verified that it's scalable linearly to 12 streams with no hiccups in all movie frames, it'd be safe not to assume like that...

BTW, in this Cell demo, there is an OS overhead, just like on Windows. The PPE in Cell runs OS services to secure responsiveness while SPEs are working, which is a big difference from Pentium4.

A 3.5GHz P4 is not a PII running at 3.5GHz...

There's also OS overhead when running PowerDVD in Windows on a PII 300MHz.
 
PC-Engine said:
one said:
PC-Engine said:
A single threading 3.5Hz P4 will be able to do at least 12 streams with audio like aaaaa00 mentioned in the other CELL thread.
Unless it's verified that it's scalable linearly to 12 streams with no hiccups in all movie frames, it'd be safe not to assume like that...

BTW, in this Cell demo, there is an OS overhead, just like on Windows. The PPE in Cell runs OS services to secure responsiveness while SPEs are working, which is a big difference from Pentium4.

A 3.5GHz P4 is not a PII running at 3.5GHz...

There's also OS overhead when running PowerDVD in Windows on a PII 300MHz.
You mean the fat cache on Pentium4 is better than a dedicated processor core like PPE in handling interrupts by OS services?
 
PC-Engine said:
A single threading 3.5Hz P4 will be able to do at least 12 streams with audio
Have you or anyone else actually tried that?

A dual core P4 with HT at 3.5GHz could probably do around 40 DVD streams. ;)
That's just something you pulled straight out of your ass. A P4 system wouldn't have anywhere near the neccessary bandwidth to sustain that many streams. But of course, we know you'll say or do anything to poo-poo cell/ps3/sony. I bet once PS3 launches you'll start claiming how DC is technically superior and how feckin schmenue looks better than any of the PS3 launch titles. Would fit your profile perfectly.
 
Guden Oden said:
PC-Engine said:
A single threading 3.5Hz P4 will be able to do at least 12 streams with audio
Have you or anyone else actually tried that?

A dual core P4 with HT at 3.5GHz could probably do around 40 DVD streams. ;)
That's just something you pulled straight out of your ass. A P4 system wouldn't have anywhere near the neccessary bandwidth to sustain that many streams. But of course, we know you'll say or do anything to poo-poo cell/ps3/sony. I bet once PS3 launches you'll start claiming how DC is technically superior and how feckin schmenue looks better than any of the PS3 launch titles. Would fit your profile perfectly.

I guess you haven't been reading this thread. :LOL:

48MB/s... :oops:
 
Now now kids, give it up.

To stay on topic, is it me or do i feel that the Demo was more to show off how they can get the SPEs to actually work all together. I mean, i feel like it's more of a software show-off than just a hardware show-off...

Or maybe i'm just dreaming.
 
Correction make that 40MB/s for 40 streams...with audio. :p

Anyway it's half software demo and half hardware demo considering CELL will be used in Toshiba's new SED televisions.
 
Yes, as mentioned elsewhere. It was a demo of Toshiba's software platform. I guess it shows an MPG streamer code, and then applying it across multiple Cells without the coder needing to worry about scheduling.

Howver, as it's the only real-world demo of the chip to appear in PS3, everyone's desperate to read PS3 performance out of this. I think we can safely say, based on PC-Engine's comments, that PS3 will be about as powerful as a 3.5 GHz P4, which runs at maybe 20 GFlops? :p
 
Shifty Geezer said:
Yes, as mentioned elsewhere. It was a demo of Toshiba's software platform. I guess it shows an MPG streamer code, and then applying it across multiple Cells without the coder needing to worry about scheduling.

Howver, as it's the only real-world demo of the chip to appear in PS3, everyone's desperate to read PS3 performance out of this. I think we can safely say, based on PC-Engine's comments, that PS3 will be about as powerful as a 3.5 GHz P4, which runs at maybe 20 GFlops? :p

CELL = 26 GFLOPS :p

PII@300MHz = DVD video+audio = 1MB/s :LOL: ;)
 
PC-Engine said:
Correction make that 40MB/s for 40 streams. :p

Anyway it's half software demo and half hardware demo considering CELL will be used in Toshiba's new SED televisions.

My feeling. I mean, they know what Cell does (kinda), they know the performance of the chip, what people are worried about is how to get all those units to work together efficiently, and demos like this show just that.
Cell is Cell, it's a chip, it's a great chip, but we need to see if it can be used efficiently. Guess their job now that the chip is pretty much finished, is to show us that it's not a Saturn+PS2 kind of pain in the ass to code for!
 
That they did'nt try to get to the max should be clear if one SPE was just idle through the whole demo.

Fredi
 
randycat99 said:
That would probably knock it down to 1/4...so, 12 HD streams is still pretty friggen impressive. Anybody try playing those wmv HD samples on the MS website? How many of those can your PC play simultaneously?

WMV9 HD consumes a lot more processing power than MPEG2.

Right now I'm typing this message in while watching an episode of 24 in HD that I captured off the air last week, using my Media Center PC.

1080i ATSC MPEG2 on this single processor non-HT 2.53ghz/533mhz FSB machine, and I'm averaging about 33% CPU load.

This is with all the overhead of the Media Center app, the DirectShow architecture and Windows running.

Like I said before, 48 SD MPEG2 streams for CELL is certainly impressive, but it doesn't seem to be to be orders of magnitude better than current PC tech.
 
Fafalada said:
48MB/s...
That's just the input bandwith.

Output bandwith is around 2 orders of magnitude higher (assuming we perform bilinear downsample for assembling final pic).

In my comparison I've simplified it and taken out the downscaling, but the point still stands.

McFly said:
That they did'nt try to get to the max should be clear if one SPE was just idle through the whole demo.

Fredi

Questionn is why didn't they use that last SPE? It certainly isn't to prove that there is more processing power left. Simple math says using that last SPE would yield 56 streams so why didn't they do it? Makes you think doesn't it? ;)
 
london-boy said:
PC-Engine said:
Fafalada said:
48MB/s...
That's just the input bandwith.

Output bandwith is around 2 orders of magnitude higher (assuming we perform bilinear downsample for assembling final pic).

In my comparison I've simplified it and taken out the downscaling, but the point still stands.

Err not really...

Actually it still stands. The point that a single threading 3.5GHz P4 would be able to decode at least 12 DVD streams with audio.

BTW was this demo done on a CELL computer running Windows? :?
 
PC-Engine said:
Questionn is why didn't they use that last SPE? It certainly isn't to prove that there is more processing power left. Simple math says using that last SPE would yield 56 streams so why didn't they do it? Makes you think doesn't it? ;)
Maybe one SPU is (always?) reserved for DRM functions?
 
rabidrabbit said:
PC-Engine said:
Questionn is why didn't they use that last SPE? It certainly isn't to prove that there is more processing power left. Simple math says using that last SPE would yield 56 streams so why didn't they do it? Makes you think doesn't it? ;)
Maybe one SPU is (always?) reserved for DRM functions?

Maybe the scaling SPU couldn't handle all of the additional streams.
 
PC-Engine said:
Fafalada said:
48MB/s...
That's just the input bandwith.

Output bandwith is around 2 orders of magnitude higher (assuming we perform bilinear downsample for assembling final pic).

In my comparison I've simplified it and taken out the downscaling, but the point still stands.

What is the point? That a P4 could do X amount of streams if we ignore output bandwidth?

PC-Engine said:
Questionn is why didn't they use that last SPE? It certainly isn't to prove that there is more processing power left. Simple math says using that last SPE would yield 56 streams so why didn't they do it? Makes you think doesn't it? ;)

My guess is that the additional streams could not be resized by the other SPE.
 
What is the point? That a P4 could do X amount of streams if we ignore output bandwidth?

What are you talking about? I'm just removing the downscaling part in my example ie multiple streams at native DVD resolutions with audio being output to multiple windows. If you had enough memory on your Windows PC you could just open multiple windows and resize them to thumbnails sizes. I'm running an old PIII600MHz so I might try this actually using two downscaled DVD streams ripped to my HDD. ;)
 
Status
Not open for further replies.
Back
Top