Reputation: 789
sysbench version: 1.0.7
OS: macOS 10.11.6
No matter where I ran sysbench cpu run
I get very similar results like the following.
sysbench 1.0.7 (using bundled LuaJIT 2.1.0-beta2)
Running the test with following options:
Number of threads: 1
Initializing random number generator from current time
Prime numbers limit: 10000
Initializing worker threads...
Threads started!
General statistics:
total time: 10.0005s
total number of events: 9083
Latency (ms):
min: 0.96
avg: 1.10
max: 7.18
95th percentile: 1.34
sum: 9995.18
Threads fairness:
events (avg/stddev): 9083.0000/0.00
execution time (avg/stddev): 9.9952/0.00
I read some blog posts and all say that I should look at total time
but it's always 10 sec in different platform/env. I also get the very similar result with very small prime number list e.g. --cpu-max-prime=100
. I also run with --time=0
and the benchmark never finishes.
My guess is the total time
matches the value specified with --time
option but then I don't know what's the right command to use.
Thanks in advance
Upvotes: 1
Views: 5785
Reputation: 1
The important value is « number of évents » I go from 2k on a pi3b to 45k in a framework 13" amd ryzen
Upvotes: 0
Reputation: 1
Always read man sysbench for different parameters you needed. Use example from bottom, to have a perspective of cpu performance
sysbench --time=60 --resoults-interval=10 --threads=8 cpu run
Upvotes: 0
Reputation: 1
You can set --max-time to indicate a different max time (in seconds) for the test. Default is 10 seconds. You can see the full manual here: http://imysql.com/wp-content/uploads/2014/10/sysbench-manual.pdf
Upvotes: 0
Reputation: 1
Take --cpu-max-prime big enough, for example let it be 20000. As for sysbench, it looks like it always run 10 sec now, so you should see "total number of events" value (high value means better performance). To get correct total performance of your server CPU, you should put --num-threads=[number of hyperthreads from cpuinfo] too.
Upvotes: 0