Given enough memory bandwidth, this chip may be a floating-point monster.

One problem with the design was the limited memory bandwidth as a result of having only one load-store unit.

Thus, good system performance requires very high memory bandwidth.

This provides memory bandwidth of 1.60 gigabytes per second.

This additional channel however increases the load on the graphics system's memory bandwidth.

It also can be used to more efficiently use memory bandwidth within a memory chip.

Computationally I can see that being true, but it'll likely be starved for memory bandwidth.

These are intended to provide insight into the memory bandwidth that a system should sustain on various classes of real applications.

The memory bandwidth is shared between the graphics adapter and the rest of the system.

This helps only to the extent that memory bandwidth is not the bottleneck in performance.