Aritcle/Analysis: Sony screws up with new PS3 (40 GB)

[Andir]

HappySqurriel said:

Final-Fan said:

HappySqurriel said:

MikeB said:
@ couchmonkey

release a brand-new system with a more traditional architecture (no cell and maybe no Blu-ray)

Hello Gabe!

Yes let's make the PS4 a cutdown PC! Drool!!

Who really wants innovation, advanced features and product diversity when we can all have a cutdown console PC next to our oridinary PCs within our homes?!

LOL

Well, in theory, it would be entirely possible for Sony to have worked on improving the Emotion Engine and released a 4 to 8 Emotion Engine core that was running in the 2GHz to 3GHz range which would give performance in the 25x to 80x range (as compared to the PS2); developers would be familiar with the architecture, there would be tons of well known tricks to get decent performance, backwards compatibility would be 'free' and it would be trivial to improve performance (and upscale with a similar GPU architecture) existing PS2 games.

Seriously, if Sony choose DVD and the processor I described above (with a decent GPU) they could have sold the PS3 for $300 with a minimal loss and demonstrated similar performance to the XBox 360 ... But Sony choose the architecture that was going to ensure poor sales ...

I don't know hardware design ... is this scenario plausible FOR REALS?

From a theoritical perspective, yes ...

The Emotion Engine was built off of a 250nm process back in the day and we are currently working on a 90nm process; this means that the emotion engine takes up (approximately) 12.5% as much space on a processor as it once did. Now, with the necessary infastructure to create a multi-core processor it would be a challenge to build an 8 Emotion Engine core processor but it would represent the upper edge of our capability (with a 90nm process).

The 2GHz to 3GHz range is just a rough guestimate based on what we have seen similar processors do in recent times ...

I (personally) don't see what would be overly challenging about producing a 6 core emotion engine at 2GHz or a 4 core emotion engine at 3Ghz.

The Cell is designed for scalability, the EE was not. The Cell was designed from the ground up to work with other Cell processors and share workload. It was also designed to be able to scale up the EIB to accept more SPUs to handle more tasks. The PS4 could have a dual/quad core Cell with 32 SPUs each if they wanted to.

You would rather have Sony build up a faster and more complicated EE chip than design a chip that could expand into the PS4 via the route you have claimed the EE should have gone? I think your grasping at straws here. Computing is going in a very interesting direction right now. Many of the chip manufactures producing chips to compete in the future, including Intel, are not including massive numbers of general purpose cores, but are looking to have specialized cores. The near future Intel roadmap include 8 core processors, but even they agree processors will not always head this way. IBM just happens to be the first to drop with the Cell.

The EE is/was harder to program for than the Cell. Many devs have stated this (I don't have the links right now... but I guarantee it was said. I think it was one of the SE guys.) It has to do with hand writing assembly code and setting stack flags to get the best performance out of it. I'd argue to say that with the help of the IBM compiler, the Cell is light years ahead of the EE in ease of programming.

 

Edit: Also, simply making an 8 core EE would not solve the problems developers today have. Thinking in parallel.