Environmental Considerations When Measuring Relative Performance of Graphics Cards

In this paper we examine some of the environmentalconditions that have to be considered when comparing the performanceof GPU’s to CPU’s. The range of these considerationsvaries greatly from the differing ages of the hardware used, to theeffects of running the GPU code before the CPU code within thesame binary. The latter of these has some quite surprising effectson the system as a whole. We then go on to test the differenthardware performance at matrix multiplication using both theirbasic linear algebra libraries and hand coded functions. Thisis done while respecting the considerations we have describedearlier in the paper, and addressing a problem that with the useof the Intel MKL library cannot be argued to be unfair to theCPU.