Consoles the basis for the next big convergence appliance?

[wco81]

http://forbes.com/markets/2004/11/29/1129automarketscan07.html

I think this is the same ML analyst who gave a bullish upgrade for AAPL today because of iPod sales. So his point is that the next great digital entertainment device will be an appliance, like the iPod, rather than a PC.

It could be an extension of the iPod or it could be an extension of the console. Or it could be something entirely new. Well MS is banking on big sales of Windows Media Center this Xmas so we will see if people want PCs as PVRs.

Oh and BTW, IBM and Sony announces today the prototype Cell workstation (as well as talk about the processor in general).

http://forbes.com/technology/feeds/..._ae_0000-421611-.industrytopstories.corp.html

 

[zurich]

Sounds like 2001 all over again

 

[wco81]

Did anyone notice from the second link that STI expects that "a one-rack Cell processor-based workstation will reach a performance of 16 teraflops"?

And the EE article in the long Cell thread notes that Cell workstations are already in the hands of game developers?

 

[zurich]

Did anyone notice from the second link that STI expects that "a one-rack Cell processor-based workstation will reach a performance of 16 teraflops"?

And the EE article in the long Cell thread notes that Cell workstations are already in the hands of game developers?

Yeah I thought that bit about the workstations was a bit strange - however EE did mention that even 1 tflop would be impossible given the expected memory bandwidth (I think they quoted 25gflop?).

 

[Guden Oden]

If most of the data Cell processes is dynamically generated (as is likely to be the case with a CPU with enormous computational abilities), external bandwidth will not be as much of an issue.

One APU does a bit of work, punts the data along to the next and to the next and so on. Hopefully it will only need relatively few external memory accesses, considering there's going to be several levels of on-chip memory too.

 

[Cryect]

Did anyone notice from the second link that STI expects that "a one-rack Cell processor-based workstation will reach a performance of 16 teraflops"?

And the EE article in the long Cell thread notes that Cell workstations are already in the hands of game developers?

Heh, well I doubt the one rack workstations are in the hands of game developers and doubt they even have one made yet. So that 16 TFLOPs number is highly likely to be peak theoretical performance. Doubt you would see it do real well in a SPEC test (hence I doubt we will see any SPEC numbers from them for a while), but that doesn't really matter since SPEC doesn't relate well to what the CELL processors are supposed to do.