Performance of my World Community Grid Tasks - Part 2

Hi everyone!

Last week I did some analyses of the results of the valid results of running World Community Grid on two machines, a laptop running Windows 10 on a Intel i5-6300U CPU @ 2.40GHz and a work station with Ubuntu 16.04 on a Intel i5-3350P CPU @ 3.10GHz.

Both OS versions are 64-bit, and have the same number of CPU threads. The i5-3350P CPU should be more powerful, having a higher clock frequency, 4 cores instead of 2 cores, and 6MB Cache instead of 3MB. According to UserBenchmark.com, the Intel i5-3350P should be 54% faster compared to the i5-6300U looking at the Quad Core Integer Speed, and 37% faster for Quad Core Floating Point Speed.

But how does the two CPUs performs in terms of performance on World Community Grid tasks? Is the OS an important factor? Are there differences in terms of credit points per application? To answer these questions, I repeated the analyse I did last week, now with more valid results (146 valid results in total, 31 from the Windows 10 laptop, plus 115 results from the Ubuntu 16.04 workstation). The results are somehow confirming the observations from last week: despite having a better CPU, the Windows 10 laptop shows worse performance.

But, let's go in the details:

• Looking at the "Elapsed Time vs CPU Time" factor, the Windows 10 system shows a big variation from task to task. In average, the factor is 1.19. Lot of tasks have a factor of 1.02, but there are also many having a value above 1.4.
• The Ubuntu 16.04 machine performs better in terms of "Elapsed Time vs CPU Time", having an average factor of 1.10 but also few tasks having a factor above 1.20.
• The number of Granted BOINC credit points per hour of CPU time is much better on the Ubuntu 16.04 machine with an average of 36.78 points. Also, on this machine there's no significant difference between the two applications (FightAIDS@Home - Phase 2 and Microbiome Immunity Project).
• On the Windows 10 machine the situation is completely different. The average number of credit points per hour of CPU time is 10.83, three times worse compare to the Ubuntu 16.04 work station.
• What is also strange is that on the Windows 10 laptop, all FightAIDS@Home - Phase 2 tasks gets a better average score of 13.55 points per hour of CPU time, compared to an average of 8.28 points for the Microbiome Immunity Project application.