How the Xbox 360 affects PC gamers

bit-tech said:
Oh, and here's a number for you: 48 pipelines. The X850 uses 16, the R520 is mooted to have 32. 48 pipelines is simply bonkers.

Isn't that incorrect? An Xbox360 ALU < R480 pipeline.
 
trinibwoy said:
bit-tech said:
Oh, and here's a number for you: 48 pipelines. The X850 uses 16, the R520 is mooted to have 32. 48 pipelines is simply bonkers.

Isn't that incorrect? An Xbox360 ALU < R480 pipeline.

Yup, MS seems to be pushing the 48 ALUs as 48 pipelines, obviously to try to inflate it's power.
 
trinibwoy said:
bit-tech said:
Oh, and here's a number for you: 48 pipelines. The X850 uses 16, the R520 is mooted to have 32. 48 pipelines is simply bonkers.

Isn't that incorrect? An Xbox360 ALU < R480 pipeline.

Absolutely, it has been mentioned in the article discussion thread. However, it proves to show that the numbers game can often play an important role in marketing.

The old assumption that: 'My card has 256MB of RAM, and yours only has 128MB, so mine must be faster'
 
digitalwanderer said:
ANova said:
Yup, MS seems to be pushing the 48 ALUs as 48 pipelines, obviously to try to inflate it's power.
So do we know yet how many actual pipelines it has? :|

Before Dave chirps in: "you have to reconsider the concept of pipelines..." ;)

Or: it's the end of the world as we know it...
 
Joe DeFuria said:
It has 8 pixel writing pipelines, and 42 ALUs. Zero" pixel shading pipes."
Umh..I think there are 48 ALUs and 8 ROPs. Each ROP retires shaded pixels froma a buffer/fifo, I wouldn't 'attach' an ALU to a pipe.
 
Heh, I was posting a reply to a few posts back and Joe answered me question.

Thanks Dr. DeFurious! ;)

EDITED BITS: And thanks nAo, your reply wasn't there when I was posting either.

Damn, the console related forums are a hopping today here! :oops:
 
Joe DeFuria said:
Um, I specifically DIDN'T attach an ALU to a pipe!
I missread your post then, since you wrote 42 instead of 48 I thought u substracted some ALUs to attach them to pipelines ;)
I'm thinking too much, sorry :)
 
The interesting thing is the whole architecture is load balanced up the ying yang. While traditional pipelines have trouble keeping all the alus busy this one will dynamically distribute load. Load between pipelines and load between vertex and pixel shaders. Loops and texture accesses should have little impact on performance since other threads can be processed while a texture is being accessed.

Sounds like you can use it for other algorithms as well.
 
Back
Top