KK slips CELL performance at decoding HD streams.

[one]

http://pc.watch.impress.co.jp/docs/2005/0521/e304.htm

According to Kutaragi, a full HD MPEG2 stream video results in 70% utilization of single APU, which is enough to run additional audio filter.

So CELL as seen in PS3 can decode 6 HD MPEG2 streams? That's not very impressive at all considering a 3GHz P4 can manage 3 HD MPEG2 streams. :?

Where did you find that quote? I can't find it in the linked page.

 

[London Geezer]

Besides, KK saying "We're targetting PS3 to be 1000x what PS2 was" was never really clear.
1000x what? floating point? polygons? whatever...

does it matter? is PS3 1000x PS2 at whatever metric you want to use?

The difference is that years ago KK came out saying they were targetting 1000x total performance.

Today, he has the PS3 in his hands and claims that it is around 30x the power of PS2.


Besides, i'd love to see how long it takes for PS2 to render a frame like the subsurface scattering demo showed at the conference, at 1080i.

 

[PC-Engine]

http://pc.watch.impress.co.jp/docs/2005/0521/e304.htm

According to Kutaragi, a full HD MPEG2 stream video results in 70% utilization of single APU, which is enough to run additional audio filter.

So CELL as seen in PS3 can decode 6 HD MPEG2 streams? That's not very impressive at all considering a 3GHz P4 can manage 3 HD MPEG2 streams. :?

Where did you find that quote? I can't find it in the linked page.

The exact quote is not in that article. That quote is from here which describes what KK said.

http://www.opa-ages.com/?bomba=article&id=32548

 

[hey69]

sony said in the E3 presentation CeLL inside ps3 can decode 12HD streams AT ONCE.

so according to pc-engine , a 3GHz P4 can manage 3 HD MPEG2 streams

so that makes 1 gameconsole has THE POWERZZ of a 12GHZ PENTIUYM4

if thats not impressive i dont know what it is.

and personaly i dont give a shit about this
we need a nextgen benchmark

 

[Qroach]

The difference is that years ago KK came out saying they were targetting 1000x total performance.

Today, he has the PS3 in his hands and claims that it is around 30x the power of PS2.

It's more like he claimed they would reach 1000X the power of PS2, but spin it anyway you like. LB, you need to look up the word "claim" in a dictionary

 

[London Geezer]

The difference is that years ago KK came out saying they were targetting 1000x total performance.

Today, he has the PS3 in his hands and claims that it is around 30x the power of PS2.

It's more like he claimed they would reach 1000X the power of PS2, but spin it anyway you like.

You conveniently missed the last part of my post.

How long would it take a PS2 to render a frame in a game that uses the PS3 as much as a dev can possibly get out of the system?
That is, if the game data could even be fitten on PS2, which is not the case.
With all the per-pixel lighting, the numbers of polygons on screen, and all the fancy effects PS3 will be pushing out in realtime.
That would give us a real number, but obviously it's not possible.

 

[Qroach]

Actually you missed the last part i added to my post as well. I'm not getting into an argument about what 1000x means or how the PS2 could render a scene created for the PS3. I really don't care.

Like I said above, you need to look up the word "claim" in a dictionary.

 

[one]

The exact quote is not in that article. That quote is from here which describes what KK said.

http://www.opa-ages.com/?bomba=article&id=32548

I see, it's from the comment by the interviewer (Honda) that SPE can decode full MPEG2 HD stream with 60-70% load. It's probably guessed from what Phil Harrison said in the SCE conference at E3 ("Cell can decode 12 HD streams"). It means one SPE uses 60% of its power for a single HDTV stream.

 

[Heretic]

I believe the claim was that Cell would have 1000 times the graphical processing power of current PCs (at the time). Not a huge difference but it does make the stated goal seem slightly less far fetched.

 

[JF_Aidan_Pryde]

I believe the claim was that Cell would have 1000 times the graphical processing power of current PCs (at the time). Not a huge difference but it does make the stated goal seem slightly less far fetched.

No. There was a clear diagram with PS2 then (X1000) to PS3.

 

[Qroach]

I believe the claim was that Cell would have 1000 times the graphical processing power of current PCs (at the time). Not a huge difference but it does make the stated goal seem slightly less far fetched.

I don't think so, I believe it was 1000x PS2. Even still if it was PC it's very far fetched.

 

[London Geezer]

Actually you missed the last part i added to my post as well. I'm not getting into an argument about what 1000x means or how the PS2 could render a scene created for the PS3. I really don't care.

You seem to care a lot actually...

 

[Qroach]

you're misunderstanding. When i said "I don't care' and "I'm not getting into an argument about 1000x means". I just mean I'm not about to get into an argument that the 1000x claim means floating point power or texturing ability or polygon crunching power, etc. That's just a waste of time and grasping at straws.

 

[one]

Anyway it's in the section I omitted from the translation.
It's in this part

(Honda's comment: the same kind of upconverting a movie was demonstrated by Intel but they explained it requires a future many-core CPU. CELL will be able to do it and other wonderful things on a home network)

of my translation and it's no way a comment by Kutaragi. There Honda speculates and explains what you can do with Cell-embedded Blu-ray recorder and Cell Storage, including ripening, for readers.

The literal translation of that sentence in Honda's speculation is like this

...
This comment (by Kutaragi) is very interesting.
Sony has a plan to release a BD recorder in which Cell is embedded in the future. An SPE in Cell can decode full-HD MPEG2 with 60-70% load. Probably power of an SPE will be surplus even to full-HD H.264. Of course, you can use it for decoding sounds, and farther than that, you'll be able to do some video and audio filters on Cell.
...

As you know the Toshiba video decoding demo didn't use the 8th SPE either, and considering that Toshiba also is supposed to use mass-produced Cell instead of a full 8 SPE working Cell, I guess it was also done with 3.2Ghz Cell.

 

[London Geezer]

Isn't a BR player with Cell... a PS3?

 

[Mulciber]

one would hope that for something like that they would have a different asic with say 1 or 2 SPEs with half the memory on each of them, a castrated memory interface and possibly smaller L2 cache for the PPE.

such a chip would be reletively small, especially on .65 micron, and would still be quite powerful.

 

[London Geezer]

one would hope that for something like that they would have a different asic with say 1 or 2 SPEs with half the memory on each of them, a castrated memory interface and possibly smaller L2 cache for the PPE.

such a chip would be reletively small, especially on .65 micron, and would still be quite powerful.

Why? They're gonna make millions of PS3 Cell chips, might as well keep making those and stick them in BR players. Having different chips will be more difficult than having only one chip to be used for everything.

 

[Bohdy]

Who cares about Mpeg2 -- how many AVC streams can it decode?

 

[-tkf-]

That's not very impressive at all considering a 3GHz P4 can manage 3 HD MPEG2 streams. :?

Any pointers to a test or something like that where a 3Ghz P4 actually decodes a MPEG2 HD stream and only uses 33%?

 

[jonnyp]

http://pc.watch.impress.co.jp/docs/2005/0521/e304.htm

According to Kutaragi, a full HD MPEG2 stream video results in 70% utilization of single APU, which is enough to run additional audio filter.

So CELL as seen in PS3 can decode 6 HD MPEG2 streams? That's not very impressive at all considering a 3GHz P4 can manage 3 HD MPEG2 streams. :?