skip to content
Benchtype:
Memory
 
Memory footprint:
Low to high
 
Description:
Memory bench is the classic way to mensure bandwidth of an memory subsystem.
 
It is improved by using continuous memoryblocks, sized in power of 2 from 16MByte up to 512MByte,
starting either writing to or reading from it. To enable high-precision memory performance measurement,
they both internally work with multiple passes and averages calculations per run.
 
Further, the main goal was to minimize (CPU) cache pollution on memory reads and to eliminate it (almost completely) on
memory writes. In addition to this, Memorybench operate with an aggressive data prefetching algorithm.
This all will deliver almost the maximum theoretical bandwidth while reading and writing.
 
Reached results (score):
Is given in Gigabyte per second (GB/sec.) , it is the arithmetic average between: "Average (Avg.) read" and "Average (Avg.) write".
 
Further given informations:
Benchtime (Benched in) shows the time needed to run memorybench.
Displaytime (Displayed in), shows the time needed to show the current resultset.