Xbox2 = 3x3.5 GHz PPC *confirmed*
Look at THIS: http://bbs.gzeasy.com/index.php?showtopic=149175
The "confirmed" block diagram of the Xbox 2. It will have a tripple core PPC running at 3.5+ GHz, and each core seem to be hyperthreaded (POWER5 derived?). No integrated memory controller though. 84 GFLOPs per chip..
The "confirmed" block diagram of the Xbox 2. It will have a tripple core PPC running at 3.5+ GHz, and each core seem to be hyperthreaded (POWER5 derived?). No integrated memory controller though. 84 GFLOPs per chip..
Comments
Originally posted by Henriok
Look at THIS: http://bbs.gzeasy.com/index.php?showtopic=149175
The "confirmed" block diagram of the Xbox 2.
Those are brutal specs. If it comes out with anything like this next year, desktop PCs and Macs will positively look dated (for a while, at last).
A-bomb simulations on your fingertips...
Originally posted by Henriok
...84 GFLOPs per chip..
Originally posted by PB
The theoretical peak of the 2 GHz G5 is roughly 32 GFLOPS (2 GHz x 2 cores x 4-way SIMD x multiple/add instructions).
Just as a general observation: you folks should get used to this because it is the kind of thing that 65 nm will let the hardware guys do. The current 90 nm 970FX is only 60 million transistors and is considered quite small. 65 nm would allow (very roughly) >100 million in about the same area. Ignoring the L2 cache, the 970FX core is only about 30 million transistors.
Originally posted by Programmer
The theoretical peak of the 2 GHz G5 is roughly 32 GFLOPS (2 GHz x 2 cores x 4-way SIMD x multiple/add instructions).
I bet the 84 Gflops/processor figure is theoretical peak too. Pretty impressive if true, given that the 32 Gflops of the G5 require 2 of them and takes into account the SIMD unit.
65 nm would allow (very roughly) >100 million in about the same area.
Is this going to be the mobile G5 for the Powerbooks? It has been rumored some time ago that Apple and IBM are working on a truly portable 64-bit solution, scheduled for next year. I don't remember if the report said 90 or 65 nm.
Originally posted by Programmer
Just as a general observation: you folks should get used to this because it is the kind of thing that 65 nm will let the hardware guys do. The current 90 nm 970FX is only 60 million transistors and is considered quite small. 65 nm would allow (very roughly) >100 million in about the same area. Ignoring the L2 cache, the 970FX core is only about 30 million transistors.
This trippe core processor would be about 1/3 smaller than one 970 processor. Even if it has three cores and the double the ammount of L2 cache. But not taking in account the added silicon for doing interconnects and the hyperthreading functionality.
Originally posted by Imergingenious
this is a joke. We've been waiting on 90nm for a while, and its proving to have more difficulties with signal crosstalk and such than the engineers anticipated. We won't see 65nm next year. Just like we won't see 3GHz G5's in a few months. It's just not in the cards at IBM. They made big claims when they built the state of the art chip plant at fishkill, NY. These claims are no more than wishful thinking at this point.
I'd never bet against Steve.
Originally posted by Imergingenious
this is a joke. We've been waiting on 90nm for a while, and its proving to have more difficulties with signal crosstalk and such than the engineers anticipated. We won't see 65nm next year. Just like we won't see 3GHz G5's in a few months. It's just not in the cards at IBM. They made big claims when they built the state of the art chip plant at fishkill, NY. These claims are no more than wishful thinking at this point.
90nm is here and only hit snags due to the implementation of new technology. These same issues are not expected at 65nm - which is due to begin fabbing by the end of 2004. You will see 3GHZ+ Power5 derivatives in a few months. I've got a very strong idea that they'll be introduced at WWDC.
Originally posted by cubist
The Xbox2 is supposed to come out next year, correct? I wouldn't be surprised if it had a $400 price tag to start with. However, I don't see how it could even be that cheap with specs like these...
From some googling, I found a comment by the head of Microsoft in Japan saying that he expects the new xbox to be released in 2006. It might change depending on when the offerings by Sony (PS3/PSP) are expected to ship.
That said, 65nm triple-cores don't seem quite as far fetched if put in that time period. These consoles usually have to last quite a long time in the marketplace.
Originally posted by ipodandimac
Just think! Virginia Tech's next supercomputer will be 5 XBox2's
*LOL* this one's very good
oh, and IBL.