The PSP CPU is not based on EE or GS or PSOne CPU: link...

If it's not based on PS2 than it has to be some form of Cell, hell the patent actually OUTLAYED a Cell configuration for a hand held.
 
that would be a funny situation...for a year or two their portable system would be more powerful than their home system!
 
that would be a funny situation...for a year or two their portable system would be more powerful than their home system!

Not neccesarilly, the whole point of Cell is scalability. Cell doesn't have to mean faster then EE.
 
Cell doesnt make much sense if you arent using multiple processing units ... you just dont want that kind of architectural overhead.
 
MfA said:
Cell doesnt make much sense if you arent using multiple processing units ... you just dont want that kind of architectural overhead.

We can use two APUs and put a Pixel Engine, Image Cache, 1 Display Controller, 1 PU, 1 DMAC and some e-DRAM...

Or maybe even a single APU...

The point behind Cell seemed to be as a technology that can scale from 1 Cell configuration ( 1 PE ) to multiple ones, that can scale from a PDA-like device to a game console or a bigger machine like a server or a workstation...

Plus this would provide a very similar coding environment to PlayStation 3 and would enable easy inter-operation between PSP and PlayStation 3.
 
One size doesnt fit all, no matter how well it is designed.

I dont see how compatibility to Cell is of any import anyway ... compatibility with PS1/2 code is much nicer, given where most ports will be coming from.
 
So you are basically saying that Cell cannot scale to PDAs ?

I disagree on that...

Plus they mention it is a brand new architecture and that doesn't make it any easier on porting "PSone or PlayStation 2 code"...
 
MfA said:
One size doesnt fit all, no matter how well it is designed.

But it isn't one size...?

And do you really think people will want to buy, and play all the old PSone games they already played to tatters? Not to mention all the PS2 games they’ve just grown tired of? It would be worse than SNES to GBA ports are now! :(
Of course some games are bound to be ported, but I don’t think that’s enough reason to compromise the architecture of a new machine.
 
Mfa, but Cell is not a Desktop or Server chip... this is like saying that MIPS is a dektop architecture or a server architecture... Cell is an architecture, quite clearly in the patent there is the feeling of having the same architecture in different product families including PDAs, HDTVs and game consoles...

Want to have a workstation configuration ? Add more APUs and more PEs...

You want to have a PDA have like only 1 PU, 1 APU, the DMAC and some Pixel rendering HW ( Pixel Engine, Image Cache, etc... ) and set it at like 150 MHz and you should be all set...

I do not think the APUs take so much die area at 90 nm and neither would the Pixel Engine and the Image cache... and the e-DRAM would be shared...

Do you think of that as a Desktop configuration ?


Plus, if compatibility with "code" from PSOne and PlayStation 2 were that fundamental then why are they going with a brand new architecture which is not related to PSOne or PlayStation 2 ?

( not trying to provoke you, sorry if this is the feeling you get )
 
Makes sense to me. Cell is about 'connectivity'.. PSP comes out 2k4, PS3 comes out 2k5, why not have both Sony systems using the same architecture?
 
Panajev2001a said:
You want to have a PDA have like only 1 PU, 1 APU, the DMAC and some Pixel rendering HW ( Pixel Engine, Image Cache, etc... ) and set it at like 150 MHz and you should be all set...

I do not think the APUs take so much die area at 90 nm and neither would the Pixel Engine and the Image cache... and the e-DRAM would be shared...

Do you think of that as a Desktop configuration ?

I think it is an architecture designed for massive parallelism, with all the necessary trappings that go with it, which would only be used for relatively straightforward pipelined computing ala the PS2 with a couple of processing units ... hence overengineered, so yes.

Plus, if compatibility with "code" from PSOne and PlayStation 2 were that fundamental then why are they going with a brand new architecture which is not related to PSOne or PlayStation 2 ?

Most likely they cant fit in the PS2 chipset, and the PS1 is too antique. With a MIPS instruction set providing an emulator for porting purposes for PS1 games wouldnt be that hard though.
 
I think it is an architecture designed for massive parallelism, with all the necessary trappings that go with it, which would only be used for relatively straightforward pipelined computing ala the PS2 with a couple of processing units ... hence overengineered, so yes.

While using the inherent parallelism present in Multi-tasking environments and multimedia applications which ar epart of the PDA world as well as the HDTV world and the game console world ( and more... ) is something Cell tries to do this doesn't mean that the architecture is only limited to Desktops and High-end workstations... to me it sounds like saying that IA-64 is inherently a Workstation and Server architecture which will never fit the Desktop market...

ARM processors, Strong ARM, embedded MIPS, they all have enough horse-power to be fast enough in more "serial" tasks, but they are not very well geared inheerently towards hevvy multi-media applications and heavvy multi-tasking, while Cell would be able to take care of such things quite nicely...

I know you disagree on the Pervasive Computing approach ( and you have your reasons ) and the importance of the role of Cell for Sony and Toshiba in this regard... ( the patent clearly shows that they plan for the architecture to scale from PDA-like devices upwards [in terms of complexity and work load]... )

Limiting Cell to Desktops and servers would seriously limit the reach of Cell...

They specifically mentioned the ease of migration from computer to computer for Apulets, they mentioned how sharing a common ISA ( you cna execute a normal Apulet on any APU ) was going to be beneficial in inter-operating all these Cell based device and how much easier data sharing would be...

I believe Sony, IBM and Toshiba want to see Cell spread around in as many markets as possible ( for Sony [and also Toshiba] it would save them some money, all the Cell R&D budget spent so far would pay off as it would keep the fruits of that work inside Sony helping all its sub-divisions [which could all benefit from using Cell in one way or the other] ) to further create demand for Cell and elated applications...

The chip should use a MIPS core so we would not need to do much emulation work there ( for compatibility with the the ISA used by PSOne's CPU core... ), but I understand your point even if I still think that in PSP or PSP 2.0 Cell would make sense...
 
Cell is still limited as a proprietary platform. If Cell is only in Toshiba/Sony devices, then it's only pervasive as long as you're using Toshiba/Sony equipment. Java has the potential to make Cell a truely pervasive platform, however I'm not sure how well a design like Cell would handle Java.
 
It should handle it pretty well.. several features Java would need are available in Hardware ( like memory sandboxes ) and there would be less need of VM slowdowns as Cell to Cell communication would not require them much... Jini, IIRC its features and operation should live well with the idea of Apulets...


Also, you forget IBM Hardware... Sony and Toshiba have quite large presences in the consumer electronics market so that should be a good start... Cell can be licensed, nothing prohibites that...
 
Josiah said:
Cell is still limited as a proprietary platform. If Cell is only in Toshiba/Sony devices, then it's only pervasive as long as you're using Toshiba/Sony equipment. Java has the potential to make Cell a truely pervasive platform, however I'm not sure how well a design like Cell would handle Java.

Many of the engineers working on cell have previously worked on BOA and DAISY which have achieved the compatability you describe above (PowerPC, x86, Java) and dynamically optimize on a VLIW or more recently with BOA and it's derivatives an architecture that is "inheriently lower complexity." FYI, DAISY R&D which did this (BOA was a more recent offshoot) was done in the mid-1980's.

So, I allways laugh when people on here talk about programming challenges and how Sony is going to piss off developers, et al. While it's entirely possible, they (nor do I fully) have the ability to even state this with how little they know.

Yet, I question is Java per say will be used, or x86, PPC, et al. But, this has been done by them.
 
Back
Top