optimized for high data transfer volumes. In contrast typical workstations
perform badly when data exceeds their processor caches. Performance of
scientific computations often is wrecked by this deficiency.
The increased graphics functionality, however, allows to shift the
computational load from the CPU to the graphics card. Data can be
represented in images and operations on vectors in graphics operations on
images. Broad access to graphics memory and parallel processing of image
operands thus turns the graphics card into an ultrafast vector coprocessor.
Respecting the precision restrictions a wide area of numerical applications
can benefit from hardware acceleration. The implementations of numerical
solvers for partial differential equations in image processing will
exemplify this approach.