The cell can't help out with graphics as it has no graphics functions onboard, how is that possible?
At first, I wondered what one of the Dev's was on about when he said a design that would work well on the PS3 would also benefit the 360..
When I say Asymmetrical, I am referring to the whole system..
Traditionally, you code you game engine, and it effectively does a load of processing on the CPU, throwing data out finally to the graphics card (I know this is quite simplistic, but bear with me)..
This 'scales' well to multiprocessor systems without too much pain, you can split your code up into chunks that can be run in parallel, and 'farm' them out to the other processors. You haven't had to re-write too much, just re-organised to some degree, and decided which bits of code to run on which processor...
This approach works well on the 360, becuase all processors are identical, you have 6 threads over 3 CPU's and all can handle any of the individual bits of code in the same way they did before...
This doesn't work on the PS3, because the PPE and SPE's are very different, you can't just split your traditional code into sections and farm them out, the SPE's are that good at normal integer computation. You have to redesign your code into 'normal' and 'data intensive' and farm the nnumber crunching elements out to the SPE's and the other stuff to the PPE, and getting the throughput in your design is a real headache.
The problem with both of these is that you are increasing the amount CPU resource you have, but with no further thought, all you end up doing is generating more data for your weaker components, such as your GPU then it can handle..
What developers 'say' they are now doing is approaching it from a different angle, they are looking at the system from a whole, and using the multiprocessor environment in a much more efficient way, they know that the DATA IO on the consoles (data from BR/DVD/HDD and data out to the graphics card) are bottlenecks, and so use the 'spare' CPU power to augment/help out where these bottlenecks are. A good example of this, is developers using an SPE to do geometric work, such as pre-culling, taking this from the RSX improves the overall graphics throughput..
Since both consoles are CPU 'heavy', which means they can generate much more data then the GPU's can handle, both can benefit from taking 'chunks' of the CPU resource, and adding these to the GPU to provide a synergistic effect.
It's all IMO, and based on reading around on Beyond 3D and the odd article here and there, I am no developer so it may be all rubbish...