Nehalem Cores vs. Clock Speed Raw CPU Performance

2010-05-13 • SEND FEEDBACK •
Related: Mac Pro

This test measures raw CPU speed with heavily threaded applications (rare beasts).

Since so few applications use multiple CPU cores well, this comparison is relevant only when the application being used has been written (well) to exploit all the cores.

Analysis

In the ideal case (never), the time is cut in half— pure CPU computation that has little in the way of memory bandwidth will run twice as fast when twice as many cores are used, assuming the code is written correctly.

The SHA1 benchmark from MemoryTester scales well, the quad-core model takes 61% longer.

The XCode build test of DiskTester takes 50% longer on the quad-core model. Except for the final link step, this test is CPU-bound, disk speed is not a factor.

Time in seconds, lower is faster.
Click for larger graph.