Some quotes about PS3 CPU

[McFly]

I'm more impressed about PS2 emulation on Cell. Is it emulated entirely by software, or only its GS part is executed on Visualizer? I'd thought PS2 backward compatibility would be done by a shrinked EE I/O processor hardware, and PS1 on emulator software.

They could really save cost if they don't need to include the PS2 chip(s). But it's interesting. If they really think that full speed PS2 emulation is possible with Cell, then this chip must be really, really fast. Everything below 512GFlops would actually surprise me now after hearing this. It sounds a lot like they want to break the magical 1TFflop number, no mather what. And those 100 Watts ... I don't think Xenon will be much below with what they try to include.

Fredi

 

[nAo]

Anyway is there any reason why you think this poster is believable?

From the posts he made on that forum I believe is a knowledgeable person, I don't share his vision on many aspects but I can't say he's a crackpot. I'm inclined to believe he's saying the truth about what he saw.
Some of the thing he publicly wrote + other info he said to me fit with other voices I heard from a couple of sources.

CELL design is really wrapped up, maybe even the GPU is complete (but I don't think the GPU will employ anything that is coming from CELL architecture, APUs as we know them are not good at pixel shading). Now Sony with its partners is 'struggling' to improve yields and clock rate.
I expect Sony will employ at least a single PE (1 PU + 8 APUs) in PS3 CPU, but they're working hard to achieve what we called here the 'preferred embodiment' (yes, the 4 PE x 8 APUs thing)
A 4 Ghz CPU is not possible (even with a single PE) in a late 2005 time frame, imho.
A 2 Ghz 4 PEs solution would be nice! But I have the feeling we'll have something like a 2 GHz 2 PEs CPU in the end..

ciao,
Marco

 

[nAo]

Everything below 512GFlops would actually surprise me now after hearing this.

Are you sure?
My expectations were lowered by those quotes..now I expect something below 256 GFlop/s.

It sounds a lot like they want to break the magical 1TFflop number, no mather what.

IMHO...it's not really possible to break the 1 TFlop/s barrier with the kind of power requirements you can have on a games console!

ciao,
Marco

 

[V3]

They should delay the darn thing if they can't achieve at least 1TFLOPS.

 

[nAo]

They should delay the darn thing if they can't achieve at least 1TFLOPS.

Well, they can still achieve it!
What about a 512 GFlop/s CPU + a 512 GFlop/s GPU? 8)

 

[MfA]

Why are they showing this stuff to people who dont grok fused multiply adds?

That guy knows FMACs, I had no time to pick up all the 'interesting' quotes.
Moreover I know you guys here are so smart there is no need to quote basic stuff

Then why does he suppose dual issue or 256 bit wide vectors to reach 9 GFLOPS?

 

[nAo]

Then why does he suppose dual issue or 256 bit wide vectors to reach 9 GFLOPS?

I don't know, you should ask to him I think he formulated the FMAC hypothesis in another post. However I know from some private messages I sent to him he was aware of that (in fact the first time I wrote to him was just to point out the FMAC thing..)

ciao,
Marco

 

[MfA]

Meh.

BTW, a FMAC is something you do on accumulator based architectures ... Im going to assume they go with FMADDs instead.

 

[McFly]

Are you sure?
My expectations were lowered by those quotes..now I expect something below 256 GFlop/s.

BTW, there was news a long time ago of CELL prototypes running at 200-800 MHz, and those were of all sizes ranging from 4 APUs to 128 APUs

128 APUs running at 200MHz would be about 200GFlops. If he's not lying, then we should not expect the CPU to be below 200GFlops as that's the minimum (based on the above quote) they already achieved as a prototype.

Based on that, I still think we can expect a minimum of 512GFlops for the CPU.

Fredi

 

[cybamerc]

Sounds like BS. The guy points out that he's under NDA (conveniently when he asked (?) about an aspect there is no public information on) yet has no problem going into detail about about the current state of the Cell processor.

 

[DeanoC]

The idea that you would see a physical system and then post on a newsgroup is stupid.

To give an idea, when PS2 was at this phase inside Sony offices itself the early dev-kits were locked in seperate rooms and most Sony employees had no access.

And having had dealing with Sony and met almost complete silence on the topic, I find it strange that they invited this guy in, showed him around the system and he still feels o.k. to post on a fan site like PS3Insider.

 

[nAo]

He wrote he had 2 interviews (not in the same day). First time he signed a NDA..the second time he didn't.
It seems very weird to me too that they (Sony) let him posts such things ..but that's not the first time I know for sure Sony (in the high ranks..) guys let precious info spread out

Moreove a lot of people here are getting excited on the 16 PEs prototype..but even if it's true it's obviously a research thing..nothing more. Do you seriously believe that a 16 PEs chip (128 APUs!!) (IF it does exist at all..) will make it in the final PS3 CPU? c'mon guys!

 

[nAo]

BTW, a FMAC is something you do on accumulator based architectures ... Im going to assume they go with FMADDs instead.

Sure. As you can see..too much PS2 VU coding makes everything in your life looking like an accumulator based architecture
Marco..you got a PM..

 

[Guden Oden]

I don't understand why people here seem to think 100W is such a problematic figure for a microprocessor. It's not a particulary high amount of power, chances are you have 100W light bulbs on sale in your local supermarket, if not burning brightly inside your own residence.

There's 500W halogen lamps readily available for both indoor and outdoor lighting purposes too by the way.

A modern heatpipe-based cooler + large, low-RPM fan easily deals with 100W power dissipation. Check out the new iWill dual Opteron SFF system, it cools mobo, GFX card (up to GF 6800 Ultra), 3 harddrives + optical, 2x90W Opterons + PSU with 2 80mm fans. That's it.

Now, who knows how much power next-gen consoles will REALLY draw, I'm just saying 100W isn't some sort of impenetrable sonic barrier or somesuch.

 

[zidane1strife]

I hope they at least give us the prefered-embodiment, that's the least they could do. If they get lucky and achieve 45nm in 2k5(this would be so cool, just imagine how the guys at intel would react, hehehe !!!), with the x-architecture, their capacitor-less e-dram, and other techs, I'd bet they could achieve much more.

 

[MrSingh]

The idea that you would see a physical system and then post on a newsgroup is stupid.

To give an idea, when PS2 was at this phase inside Sony offices itself the early dev-kits were locked in seperate rooms and most Sony employees had no access.

And having had dealing with Sony and met almost complete silence on the topic, I find it strange that they invited this guy in, showed him around the system and he still feels o.k. to post on a fan site like PS3Insider.

indeed, PS3 is still off limits to most Sony employees right now.

early specs are available to some PS3 internal software teams, but they aren't final. oh.. the initial post is way off.

 

[one]

So, as long as we won't see a C&D order from SCE to the poor ps3insider forum in a week, it's a hoax. If mysterious power from above wipes out all his comments and membership from the board, then 8) 8) 8)

 

[MrWibble]

So, as long as we won't see a C&D order from SCE to the poor ps3insider forum in a week, it's a hoax. If mysterious power from above wipes out all his comments and membership from the board, then 8) 8) 8)

Of course now you've said that, they might play a double-bluff to prevent assumptions, and leave true information on the site while closing down sites with the least accurate rumours...

 

[passerby]

The idea that you would see a physical system and then post on a newsgroup is stupid.

To give an idea, when PS2 was at this phase inside Sony offices itself the early dev-kits were locked in seperate rooms and most Sony employees had no access.

And having had dealing with Sony and met almost complete silence on the topic, I find it strange that they invited this guy in, showed him around the system and he still feels o.k. to post on a fan site like PS3Insider.

indeed, PS3 is still off limits to most Sony employees right now.

early specs are available to some PS3 internal software teams, but they aren't final. oh.. the initial post is way off.

Basically I was wondering about the poster due to all the stated reasons.

'PS3 premiere' in end of FY04 = Jan-March or 2005 according to Sony schedule. Just wait a little longer...

 

[j^aws]

So, as long as we won't see a C&D order from SCE to the poor ps3insider forum in a week, it's a hoax. If mysterious power from above wipes out all his comments and membership from the board, then 8) 8) 8)

Of course now you've said that, they might play a double-bluff to prevent assumptions, and leave true information on the site while closing down sites with the least accurate rumours...

Hehehe...that means B3D is going down first!

Anyway, I've exchanged a few posts with cpiasminc at PS3I and he's extremely knowledgable. IIRC, he's a PS2 dev who's in-between jobs (I guess this is where the SCEA interviews come in).

A 2006 launch has 45nm written all over it from Sony, IMHO, and we should see the 'pre-ordered embodiment' BE/CPU > 2Ghz. 8)

Btw, cpiasminc, if you're reading this, I must catchup on that GI blogg of yours!