PS3 R&D Funds Helped Build the Xbox 360

Just want to have a nice chat with your beer? Go ahead ;-)

PS3 R&D Funds Helped Build the Xbox 360

Postby billb » 07 Mar 2009, 14:50

PS3 60GB [CECHA01], FW 3.15, YDL 6.2, Samsung T260HD @ 1920x1200
Powermac G4 1.25 GHz x2, 2 GB RAM, YDL 6.2
User avatar
billb
Site Admin
Site Admin
 
Posts: 5522
Joined: 24 May 2007, 20:30
Location: Eastern NC, USA

Re: PS3 R&D Funds Helped Build the Xbox 360

Postby ppietro » 07 Mar 2009, 19:08



BTW - I got this book as a Valentine's Day present from my wife. It's a great book - I've enjoyed reading it.

Not only does it explain why the Xbox 360's performance is so surprisingly good against the Cell - think of a chip with 3 PPEs, but instead of each PPE having an AltiVec like the Cell, they each have a VMX128 - it also explains why Apple switched to Intel.

IBM was using the PPE as the basis of a low power server core - kind of like a "G6". Because of the compressed schedule to get both the Cell and the 360's "PlayBox" Xenon processor done, the decision was made at a major meeting to forgo out-of-order execution. This killed the processor's server performance and the author's boss started having angry meetings with Apple execs. Right after that, the decision to switch to Intel was announced.

This also leads me to some interesting speculations about the Cell and the Xenon.

The Cell has 1 PPE w/AltiVec, plus 6 SPEs. The Xenon has three PPEs, and 3 VMX128s. Each PPE is capable of handling two threads simultaneously. (Like hyperthreading for x86 processors)

If you look at threading and execution units, there's a similarity:
Cell: 2 PPEs + 6 SPEs + 1 AltiVec
Xenon: 6 PPE + 3 VMX128s.

I'm aware that the AltiVec and VMX128s are not independent processors - just saying it's interesting.

The Xenon basically uses a standard triple core processor plus a (powerful) numeric coprocessor, and then works closely with the GPU. (i.e. like a regular PC). The Cell tries to do all of it on chip - and according to the book - was only planning on using a Sony supplied simple framebuffer display chip. The fact that a lot of the games are using the PS3's nVidia RSX GPU for co-processing kinda points out how hard it is to wrap your head around the Cell architecture. This means that many of the first generation PS3 games were probably using the PPE + RSX with only a minimal assist from the SPEs.

Considering the difference in performance between the 360's ATI GPU and the nVidia RSX GPU - the ATI is a more powerful part - that explains why the first generation PS3 games didn't look much different - and in some cases underperform - compared to 360 games.

Cheers,
Paul

P.S. How cool is it that my non-tech wife got me this book? :D
User avatar
ppietro
Site Admin
Site Admin
 
Posts: 4965
Joined: 13 Sep 2007, 22:18

Re: PS3 R&D Funds Helped Build the Xbox 360

Postby billb » 07 Mar 2009, 20:15

ppietro wrote:P.S. How cool is it that my non-tech wife got me this book? :D


Way cool. :D I might want to check that one out myself. The book, I mean. :lol: Thanks for the additional info re: that story -- good stuff.

Re: the Cell and Linux, so far the best thing I've seen released for general purpose use is that SPE scaler tool (and the modified fscaler version). Past that my SPE's stay idle most of the time unless I'm checking out the IBM Raytracer (that gets old after a while) or the mesa-cell driver (any updates on that, I wonder).

EDIT: funny, the yellowdog-board topic on the mesa cell driver comes up second on a Google search for "mesa cell driver"
PS3 60GB [CECHA01], FW 3.15, YDL 6.2, Samsung T260HD @ 1920x1200
Powermac G4 1.25 GHz x2, 2 GB RAM, YDL 6.2
User avatar
billb
Site Admin
Site Admin
 
Posts: 5522
Joined: 24 May 2007, 20:30
Location: Eastern NC, USA


Return to Speaker's Corner

Who is online

Users browsing this forum: No registered users and 35 guests