System Benchmarks
Our system benchmarks will complete the low-level picture. The CPU is the main performance limiter with these test, so with an average of 15% CPU load generated by the second MagicTwin station, we'll see how close our tests in this section get to the other 85% of available CPU performance.
We're 88% 'efficient' here, compared to the usual zero-load scenario that occurs when the test is normally run. Not too shabby and it shows we're not completely CPU limited with Kribibench.
81% this time, slightly less impressive than Kribibench but otherwise close to what we'd expect given the extra load on the system.
It takes 18% longer to complete the LAME test with the artificial load. With LAME hitting the processor hard, we can probably be thankful of HyperThreading in this case. It'll be interesting to see how Jetway's AMD-base MagicTwin systems fare in these tests.
With Quake3 loving CPU and memory bandwidth, it's not too surprising to see a near 22% deficit here, since we're not entirely CPU limited in its case.