>>44
So just to clarify, you genuinely think that lack of precision, rather than speed, is a serious issue in most computing? If only we had 128bit floating point operations done in one clock cycle, we'd have realtime raytracing or something? Of course a 2 bit processor would be awful: because 2 bits is clearly not sufficiently precise for any real number calculations. 64 bits is plenty for all non-scientific purposes. Hell, 32 bits is most likely enough for all non-scientific purposes. You have a 10-year-old's understanding of this matter if you truly believe that 128-bit or 256-bit or 1024-bit computers are going to be radically faster than 32- or 64-bit computers are.
>>45
And how is this supposed 128bit CPU going to compute 80,000 "raytraces" (I assume you mean ray-triangle intersections; the definition of a "raytrace" would clearly depend on what you are ray tracing) in the same time it takes to calculate 1+1? At best a 128-bit CPU could perform operations on 4-vectors of 32 bit floating point numbers (16 bits is generally considered insufficient for graphics purposes nowadays); this would hypothetically offer a 4x speed increase. However, a 4 core 32 bit CPU would offer that same 4x speed increase and be more flexible at the same time.