From: andihartmann@01019freenet.de   
      
   On 06/04/2018 at 06:17 PM intertubes@gmx.net wrote:   
   > On Sunday, June 3, 2018 at 5:00:06 PM UTC+2, Andreas Hartmann wrote:   
   >> On 06/03/2018 at 03:40 PM Darklight wrote:   
   >   
   > snip   
   >>   
   >> Disabling SMT means reducing CPU power by about 50%.   
   >>   
   > snip   
   >>   
   >>   
   >> Regards,   
   >> Andreas   
   >   
   > uh, no.  Not even close.   
      
   You're right - globally spoken is always wrong.   
      
   > Disabling SMT slows things down a bit if you have more than six processes   
   (threads) running in parallel, below that it makes no difference at all.   
   > If you *do* have more than six then the switching out of one process for   
   another takes very slightly longer.   
      
   It depends on the workload. I wouldn't pay for a Ryzen if I don't have   
   the appropriate workload. One workload example is kernel compiling: I   
   measured 22% difference between SMT on or off (make -j8 vs. make -j16).   
   This example is quite more than "a bit" but below "about 50%", too.   
      
   There are certainly workloads where you can't see any difference at all   
   and there may be workloads, too, where disabling SMT even shows advantages.   
      
   Nevertheless, the machine must run stable using default configurations.   
   SMT is enabled by default - therefore the machine must be stable using   
   this default even under high load (-> that's why I bought this machine)   
   and not only after disabling half of the CPUs shown by lscpu and   
   therefore often notably reduced power.   
      
      
   Regards,   
   Andreas   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)