Details trickle out on CELL processor...

V3 said:
When people said stream processor, they meant the processor cannot create additional data, data are just passing through and gets processed. Example, would be a vertex shader.

I think APUs will be able to create local data for themselves... EE's VUs could after-all :).
 
Jaws said:
From EEtimes,

And each processing element is connected to its neighbors in the cell by high-speed "highways." Designed by Rambus Inc. with a team from Stanford University, these highways — or parallel bundles of serial I/O links — operate at 6.4 GHz per link.

Redwood between Cell PE's...You know there might not even be a BE? ...Is this PE>PE 128bit or 1024bit? My memories going badly! :p

The PE<->PE on a single chip bus used to be a 1,024 bits bus.
 
I think APUs will be able to create local data for themselves... EE's VUs could after-all

Not if they're stream processors. EE's VUs are not stream processor. Only EETimes and PCweb refer them as stream processors, gotta wait for more detail to confirm it though.
 
I don't think a stream processor should stick to that, imho restricted, definition.
 
Panajev2001a said:
V3 said:
When people said stream processor, they meant the processor cannot create additional data, data are just passing through and gets processed. Example, would be a vertex shader.

I think APUs will be able to create local data for themselves... EE's VUs could after-all :).

Heh, agreed otherwise there is no way in hell you can reach anywhere near peak performance. Example being raytracing with pixel shading where only a small amount of data is inputted and outputted but a lot of intermediary data being created throughout all the operations (but its just temporary data of course).

The problem I see with Cell processors is they will only be good for situations where you don't need a lot of intermediary data due to low levels of cache but luckily for 3d graphics you don't need lots of cache.

But, this means it won't be a good supercomputer processor :( (ex a BlueGene chip is 2 cores but has 4 MB of L3 cache (EDRAM) shared between them).
 
V3 said:
I think APUs will be able to create local data for themselves... EE's VUs could after-all

Not if they're stream processors. EE's VUs are not stream processor. Only EETimes and PCweb refer them as stream processors, gotta wait for more detail to confirm it though.

We will see how loose is this definition of "stream" processor or not.

It would kind of make sense with how the APUs should operate with the Apulets, but still I do not think that they have not thought about things like real-time geometry tessellation (especially given how the flexibility of the EE's VUs was so important to Sony/SCE and Toshiba).

I do not see a Primitive Processor in there and I do not see yet a big VS in the GPU... with the CPU doing all the VS work, going back to VS 2.0 capbilities when PlayStation 2 had better capabilities would not seem like a big win.
 
EEtimes,

Workstations fitted with the Cell architecture — a $2 billion endeavor — are already in the hands of game developers.

So why haven't you guys got these yet? Or are they on their way? :p
 
Maybe thats a typo and they meant

A workstation fitted with the Cell architecture — a $2 billion endeavor — is already in the hands of a game developer.

Edit: whoops needed to correct an "are" with "is" :D
 
Cryect said:
Maybe thats a typo and they meant

A workstation fitted with the Cell architecture — a $2 billion endeavor — are already in the hands of a game developer.

Hmm...1st/2nd parties prolly have them with watertight NDA's....except Tecmo! :)
 
http://finance.yahoo.com/mp#rmbs

10:23AM Mike Tarsala's TechWatch Alert -- RMBS (RMBS) 24.34 +2.17: Analysts we reached this morning say that Rambus is very likely to benefit from the Cell processor announced today by Sony (SNE), IBM (IBM) and Toshiba (TOSBF). Of note, Rambus announced back in January 2003 that it had licensed its XDR memory and its Redwood high-speed parallel interface (its Yellowstone and Redwood technologies) to Sony and Toshiba. While IBM was NOT MENTIONED in that 2003 release, Big Blue is known for having a strict approval process when its name appears in other company's press releases. Another company analysts have mentioned as possible benefactor from the Cell processor, which is expected to go toe-to-toe with Intel in the home entertainment (specifically the multimedia living room) could be graphics-processor maker Nvidia (NVDA) -- although none of the analysts offered any specific proof. More details on the Cell processor are expected to be announced at the International Solid State Circuits Conference, which is scheduled to begin on Feb. 6 in San Francisco.
 
V3 said:
I think APUs will be able to create local data for themselves... EE's VUs could after-all

Not if they're stream processors. EE's VUs are not stream processor. Only EETimes and PCweb refer them as stream processors, gotta wait for more detail to confirm it though.
What is so special about stream processors, other than the fact that they only have very little internal memory (I take it that a DSP ~ stream processor?)
 
Squeak said:
V3 said:
I think APUs will be able to create local data for themselves... EE's VUs could after-all

Not if they're stream processors. EE's VUs are not stream processor. Only EETimes and PCweb refer them as stream processors, gotta wait for more detail to confirm it though.
What is so special about stream processors, other than the fact that they only have very little internal memory (I take it that a DSP ~ stream processor?)

Basically cheap for their performance, a DSP is an example of a fixed function stream processor where everything has been prespecified. Stream processors can be setup to allow programming but are often fairly limited. Basically stream processors are good if you need a specified processing rate for a fixed maximum amount of data bandwidth (example a router).

Really the CELL chip is prolly somewhere between a stream processor and a general CPU. The use of calling it a stream processor is prolly more to differentiate from other designs which are also taking parallel approaches to getting more performance.

Major thing I would say in stream processor is preferrably there is very little instruction setup necessary where each unit is placed into a desired operation then data is streamed in to an initial set of APU's and they pass it on to other APU's creating a pipeline. Not sure if the CELL will allow you to tie APU's to other APU's though.
 
What is so special about stream processors, other than the fact that they only have very little internal memory (I take it that a DSP ~ stream processor?)

Stream processor typically can't create or destroy data.
 
We will see how loose is this definition of "stream" processor or not.

Yes. The title of the article is streaming processing, so it may be loose.

It would kind of make sense with how the APUs should operate with the Apulets, but still I do not think that they have not thought about things like real-time geometry tessellation (especially given how the flexibility of the EE's VUs was so important to Sony/SCE and Toshiba).

Well, with all that bandwidth and processing resources, it can still tessellate, and function like typicall processor.

I do not see a Primitive Processor in there and I do not see yet a big VS in the GPU... with the CPU doing all the VS work, going back to VS 2.0 capbilities when PlayStation 2 had better capabilities would not seem like a big win.

Its not like the PE is a stream processor, only those APUs are the one in question. Even if it turns out to be so, it still has that Power core in there and if they clock that fast as well, there are plenty of processing resources there.
 
Back
Top