 |
|  |
 |
|
sprucemoose3311
|
 |
san jose, ca
Jun 2002 time: 21:25
|
|
http://ps2.ign.com/articles/585/585865p1.html
PS3 Power: Details on Cell
Sony, IBM and Toshiba unveil hard specs behind the upcoming PS3 processor.
By Chris Roper
February 7, 2005 - Sony Computer Entertainment, IBM and Toshiba held a press conference in downtown San Francisco this morning to unveil more details on their collaborative Cell processor. While there are still a few details and bits of information that the three companies are keeping mum about for the moment, they did reveal and confirm plenty of specs that had been only rumors and guesses for some time now.
For those of you who like reading numbers more than words, here are some raw specs (which we'll discuss below):
Each Cell processor contains 8 Synergistic Processing Units and a single 64-bit Power Architecture Unit (All are RISC designs with SIMD)
Operates at >4GHz and capable of >256GFLOPS
256KB Local Storage per SPU and 512KB L2 Cache (2.5MB total)
128+ concurrent transactions to memory per processor
High-speed internal element interconnect performing at 96B/cycle
234 million transistors
Prototype die size of 221mm^2
Fabricated with 90nm SOI process technology
Backtracking for just a second to cover those who haven't been following recent events (for example, we've put a man on the moon), the Cell microprocessor is a multi-core processor that's designed to handle a large number of tasks simultaneously. The first known major use of the Cell processor will be Sony's next-generation gaming machine, which we'll refer to as PlayStation 3 for the time being (Sony has yet to announce its official name, but you can bet on PS3).
While there wasn't any mention of PlayStation 3 at the press conference per-se, you can assume that the following specs will be very close to what will be going into the system. While Cell is being called a "supercomputer on a chip", NVIDIA will actually be providing the GPU in Sony's next system, which you can read more about. Knowing that, if Cell is anywhere near as fast as what the rumors and these specs say, we'll have our hands on an extremely fast console by the end of 2006 (if the rumored release time stands true).
Sony, IBM and Toshiba said that they have ran Cell processors at "greater than 4GHz", though they wouldn't name an exact speed. Performance of the processor is said to be in excess of 256GFLOPS (256 billion floating-point operations per second). That's an extremely large number, the sort that you normally throw around when you're talking about arrays of multiple processors or even supercomputers. Measuring FLOPS can be done in a few different ways, but as a comparable number the highest you could get from a fast Pentium 4 these days is somewhere in the upper 20's.
Cell is claimed to have "10X performance for many applications", which is in line with these numbers. The Power Processor Element can handle two threads simultaneously while each SPE can handle one, making for a total of 10 simultaneous threads, in line with the 10X performance number they're claiming. Sony, IBM and Toshiba declined to specify exactly which applications they were talking when they mentioned 10X performance, though we'd assume it's pure number crunching or data compression or encryption, those sorts of things.
Obviously, real-world performance for any processor is entirely dependent on how well any given piece of software is written for it, especially so for multi-processor (and multi-core) machines. When asked about how one would go about programming for Cell or how difficult the task of managing the SPEs would be, an IBM rep stated that there would be multiple programming models, allowing for different approaches to programming for Cell. Each programmer could theoretically pick and choose which method they like to work with and program that way. It was also said that language compilers for Cell (for C, C++, etc.) would be released as open-source so that anyone could learn to tinker with its innards or possibly mold a compiler around a separate programming style for different needs.
NVIDIA has been working on aspects of Sony Computer Entertainment Inc.'s next generation system for the past 2 years
oh and do not forget the blue ray disks that hold 40 gigs of info...
i will be first in line for the ps3
|
|
|  |
 |
|
Asher
|
 |
Calgary, Alberta
Nov 1999 time: 22:25
|
|
quote: It was also said that language compilers for Cell (for C, C++, etc.) would be released as open-source so that anyone could learn to tinker with its innards or possibly mold a compiler around a separate programming style for different needs. |
I can absolutely say this is bullshit -- Cell uses IBM's XL C/C++ compilers, which are very expensive (and very fast). There's absolutely no way they'd be open sourced. They could also be refering to gcc...which is an open source compiler already, but ass-slow compared to XL.
As for the PS3, I'm thinking it'll be a lot like the PS2. Hyped to no end, "supercomputer on a chip", with a harsh reality. The PS3 is going to be an absolute beast to program for, just like the PS2 was, and the first year or two of games will not look that much better than PS2 games (aside from more textures).
PS3 is all hype, just like the PS2 was. People never learn.
|
|
|  |
 |
|
Asher
|
 |
Calgary, Alberta
Nov 1999 time: 22:25
|
|
quote: yes there will be ps3 hype. However, i believe that it will be the best gaming system given the facts and track records of sony, nintendo, and microsoft. Sony did not invest in the RnD and infrastructure to make a 4 gig cpu for nothing, there will be substance behind the hype. Nintendo is gambling big on a "revolution in gaming" which may work if not a lame gimmick. Microsoft runs the risk of reproducing the dreamcast's fate. |
Given the facts, I'd expect Microsoft's console to be the fastest.
Sony once again went for the outlandish system that's massively parallel, which is awesome for building hype (it gives incredible theoretical numbers), but will never come close to achieving that. Until 2 years ago, Sony was going to develop their graphics chip all by themselves. The Xbox scared them, and they went to someone who knew how to do it. 
Sony didn't put that much R&D into Cell, the vast majority of it came from IBM. Cell existed before Sony got involved. Sony offered to subsidize some of the R&D if they could use it in their PS3, and IBM agreed.
The Xbox2 is going to use far more conventional processors, ones that have already proven themselves to be incredibly fast. Current specs are 3 Power5-derived chips (the ones in IBM's hugely expensive/fast servers), rather than many, many very weak chips.
I'm willing to bet the Xbox 2 will overpower the PS3, even though the Xbox 2 will come out first. It just has a lot more going for it, including ease of development. It will be a lot easier to "tap" the resources of the Xbox 2 than the PS3, just like it was for the Xbox to the PS2. And that makes a world of difference.
Games are developed faster. They look better, faster, with less money. Faster development with lower development cost is interesting to 3rd party developers, which is one reason Microsoft got so much support so fast, while Nintendo has the opposite problem.
|
|
|  |
 |
|  |
All times are GMT. The time now is 05:25. Apolyton Time is 00:25. |
top of page
|
|
|
Forum Rules:
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
|
HTML code is ON
vB code is ON
Smilies are ON
[IMG] code is ON
|
|
|
|
|
|