More hits to the heart for the 4 PE's CPU fans (CELL)

Few more details about the CELL processor continues to dribble out. Unfortunately nothing about the programming model and programming environments that IBM/Sony will promote/recommend.

If I have time, I'll write up some of this as a "part 2" to the CELL article, although that may not happen. Regardless, there are a few new details that I believe are "new" as far as the CELL processor goes, or I simply missed the discussion at ISSCC.

1. The CELL processor will natively support glueless 2 way SMP. N Way SMP for N > 2 needs a coherent switch. Glueless SMP ring is not possible

2. SPE dynamically reconfigurable as secure processor.

3. Locking caches (via replacement Management tables)

4. DMA is fully Power Arch Protect/x-late

5. Direct programmer control DMA/DMA-list

http://realworldtech.com/forums/ind...&Thread=1&entryID=46582&roomID=13

A 4-way PE's system gets more complicated: I do not recall a coherent switch in the original patent by Suzuoki Masakazu.
 
Isn't it because it has only one interface in the processor? Still, the SMP by 2PE + 2PE is OK without a bridge, if your power consumption budget is limitless, that is.
 
1107869831_622.jpg



Dual-Cell or not Dual-Cell?
 
You asked that already on another thread and it was answered. Can't remember to be honest but it was something about "control".. errr i really can't remember, it was nothing major though.
 
london-boy said:
You asked that already on another thread and it was answered. Can't remember to be honest but it was something about "control".. errr i really can't remember, it was nothing major though.

i never see that on another wafer (> google image > wafer)
 
It's not part of Cell. This was answered already, i think it had to do with testing and stuff. Obviously there is a lot of testing to be done at this stage, maybe that's the reason these early wafers have those extra circuits.
 
Back
Top