Re: [lmi] Build time comparison

[Vadim Zeitlin]

On Wed, 14 Oct 2020 11:58:53 +0000 Greg Chicares <gchicares@sbcglobal.net> 
wrote:

GC> It looks like our corporate server is slower than I remember.

 AFAIR it's a VPS, so there might be some overhead due to virtualization
but, mostly, you need to take into account the fact that you share the
server with whatever else is running on the same physical machine and I
guess this is what explains the change in performance, i.e. if you had it
only for your own use before, but now there is something else running on
it, it definitely could explain the difference.

GC> Here's the result of
GC>   $</proc/cpuinfo grep 'model name' |wc -l
GC>   $</proc/cpuinfo grep 'model name' |sort -u
GC> on the server:
GC>   8
GC>   model name      : Intel(R) Xeon(R) Gold 6252 CPU @ 2.10GHz
GC> and on my machine:
GC>   32
GC>   model name      : Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz
GC> The "Gold 6252" has 24 cores according to intel; it looks like they
GC> allocate only eight to us. My PC has 16 physical cores, but I've
GC> enabled hyperthreading, whereas the server has not (according to
GC> "Thread(s) per core" reported by lscpu).

 I don't know if this could be affected by virtualization. Looking at

https://ark.intel.com/content/www/us/en/ark/compare.html?productIds=83356,192447

it definitely seems that Xeon 6252 should be outperforming your E5-2630, as
it has both more cores and higher turbo speed for single-threaded
workloads, which just confirms me in my idea that it must be busy with
something else in parallel.

 In any case I'm afraid there is nothing you can do about it, sorry.
VZ

pgpmJtjBcmu9Q.pgp
Description: PGP signature