Early Benchmarks Of Linux 4.10 Show Some Improvements & Regressions For Core i7-6800K
This New Year's Eve I finished up some benchmarks of the Linux 4.5 through Linux 4.10 Git kernels on a powerful Core i7 6800K "Broadwell-E" system. I found some improvements with 4.10 Git, but there are also some evident regressions.
I'll have more benchmark results in the New Year as time allows and the 4.10 development settling down, but from the tests I did so far on the Core i7 6800K system there is some concern over what appears to be some definite and noticeable regressions.
The same Intel Core i7 6800K + MSI X99A WORKSTATION + 4 x 4GB DDR4 + Samsung 850 EVO SSD + NVIDIA GeForce GTX TITAN X were used for the duration of these benchmarks.
One of the tests showing an improvement with Linux 4.10 is HPC Challenge's EP-STREAM Triad.
But for many of the scientific OpenMP tests, Linux 4.10 on this box appears to be definitely slower.
With other tests there are some smaller changes in either direction.
FFmpeg appeared to be one of the cases showing a nice improvement.
Stay tuned for more Linux 4.10 benchmarks next week. Find many more of these early benchmark results via this OpenBenchmarking.org result file.
I'll have more benchmark results in the New Year as time allows and the 4.10 development settling down, but from the tests I did so far on the Core i7 6800K system there is some concern over what appears to be some definite and noticeable regressions.
The same Intel Core i7 6800K + MSI X99A WORKSTATION + 4 x 4GB DDR4 + Samsung 850 EVO SSD + NVIDIA GeForce GTX TITAN X were used for the duration of these benchmarks.
One of the tests showing an improvement with Linux 4.10 is HPC Challenge's EP-STREAM Triad.
But for many of the scientific OpenMP tests, Linux 4.10 on this box appears to be definitely slower.
With other tests there are some smaller changes in either direction.
FFmpeg appeared to be one of the cases showing a nice improvement.
Stay tuned for more Linux 4.10 benchmarks next week. Find many more of these early benchmark results via this OpenBenchmarking.org result file.
2 Comments