Dear All,
I want generate a precise clock in μs resolution. Therefore I run busy wait loop in isolated core to get system time stamp, check in interval then generate clock in μs resolution.
In desktop(i7-4770), this program can run with low latency (around 5-10 μs variation). However, when I put this program into Xeon(E5-2698), the latency will increase to 500-1000 μs. It is beyond our expectation that Xeon should have better performance including latency.
Condition:
Desktop: ubuntu LTS 14.04 with 3.17 kernel/3.4GHz/8G memory
Xeon: ubuntu LTS 14.04 with 3.17 kernel/3.3GHz(turbo boost to increase 2.3GHz to 3.3GHz)/32G Memory/HT disabled
We assume there should be OS jitter to cause the high latency but our question is why i7-4770 does not have the same high latency with the same linux distribution and kernel.
Does any expert give suggestion to lower the latency in Xeon?
Best Regards,
Kevin
