I agree with Shifty and XBD. CE goods may [have to be] the market where it catches on, initially at least.
Two thing we tend to overlook when discussing CELL's power in regards to retail markets:
1) It takes time and money to write applications to take advantage of the architecture; adding in hardware startup costs moving to the CELL architecture is an expensive move and a big risk with a lot of risk and possible upside.
2) Intel and AMD are not sitting still. Conroe is coming this Summer as well, and the peak FLOPs on Conroe are not too bad, and the chip is relatively cheap (~$300 for 2.4GHz, less than $600 for 2.66GHz). Quad cores are also around the corner. For the consumer space that means no need to get new software (just keep using the tens of thousands of back library items and allow the new high-demand computing stuff trickle in); for IT it means you can use your older software and have access to relatively cheap machines for server farms. AMD looks to have a slow next 12 months, but they are moving forward on ZRAM and on a FPU coprocessor using Clearspeed. So in the next couple years you may be able to pick up a quad core CPU where half the processors are FPUs. May not be idea for all situations, but this offers some flexibility and software compatibility.
Personally I have thousands of dollars of software. If I have to choose between a quad core x86 CPU where all my software works and 99.99% of new software will work on and I get a significant boost in performance or choose Cell with no compatibility with past software with very few new titles and limited selection, I am going to choose the x86. As an IT person unless time & cost come out on CELLs side and the risk is low *and* it is well established that my task will absolutely work well on CELL and can be deployed within the time frame of the project, then I could look at it. But then again people tend to be lazy so even that is not a guarantee. It is also nice on the consumer sector to have AMD and Intel competing and using the same memory platforms and reaching a huge market (over 200M PCs a year). This creates a very price competitive market.
It seems to me that if CELL is going to get a foothold, if it can outside the PS3, is going to be in specialized server markets (which seem to be very small for CELLs unique structure and needs right now) and in CE goods. TVs, HD Media drives, etc
In which case it is up to Sony, Toshiba, Samsung, etc to incorperate CELL into their own products.
And the wild card: NV. I always thought that using CELL for a PPU would have been a good idea seeing as AGEIA and Havok already are developing libraries for it. Maybe NV will take a stab at this? Or something more wild...
Two thing we tend to overlook when discussing CELL's power in regards to retail markets:
1) It takes time and money to write applications to take advantage of the architecture; adding in hardware startup costs moving to the CELL architecture is an expensive move and a big risk with a lot of risk and possible upside.
2) Intel and AMD are not sitting still. Conroe is coming this Summer as well, and the peak FLOPs on Conroe are not too bad, and the chip is relatively cheap (~$300 for 2.4GHz, less than $600 for 2.66GHz). Quad cores are also around the corner. For the consumer space that means no need to get new software (just keep using the tens of thousands of back library items and allow the new high-demand computing stuff trickle in); for IT it means you can use your older software and have access to relatively cheap machines for server farms. AMD looks to have a slow next 12 months, but they are moving forward on ZRAM and on a FPU coprocessor using Clearspeed. So in the next couple years you may be able to pick up a quad core CPU where half the processors are FPUs. May not be idea for all situations, but this offers some flexibility and software compatibility.
Personally I have thousands of dollars of software. If I have to choose between a quad core x86 CPU where all my software works and 99.99% of new software will work on and I get a significant boost in performance or choose Cell with no compatibility with past software with very few new titles and limited selection, I am going to choose the x86. As an IT person unless time & cost come out on CELLs side and the risk is low *and* it is well established that my task will absolutely work well on CELL and can be deployed within the time frame of the project, then I could look at it. But then again people tend to be lazy so even that is not a guarantee. It is also nice on the consumer sector to have AMD and Intel competing and using the same memory platforms and reaching a huge market (over 200M PCs a year). This creates a very price competitive market.
It seems to me that if CELL is going to get a foothold, if it can outside the PS3, is going to be in specialized server markets (which seem to be very small for CELLs unique structure and needs right now) and in CE goods. TVs, HD Media drives, etc
In which case it is up to Sony, Toshiba, Samsung, etc to incorperate CELL into their own products.
And the wild card: NV. I always thought that using CELL for a PPU would have been a good idea seeing as AGEIA and Havok already are developing libraries for it. Maybe NV will take a stab at this? Or something more wild...