Intel Xeon E5-2680v4 vs E5-2670v3

will be a bit slower but do you really need or will use that extra 2GB ram ?

 

Linux uses memory different from Windows in how it reports free and available memory and used memory. For an explanation see Help! Linux ate my RAM!

Check your output for SSH commands

Code (Text):

free -m

Code (Text):

sar -r

Are you on dedicated server or VPS ? If VPS is it OpenVZ or KVM/XEN based ?

If on OpenVZ VPS then memory is reported differently in sar -r output as virtual memory (VSZ) is also reported as used memory while on non-OpenVZ VPS or dedicated, virtual memory is not reported as used memory. CentOS 7 sees to report correctly memory in free command on OpenVZ but CentOS 6 would report like sar output where VSZ memory is used but available. See Difference between VSZ vs RSS memory usage - LinuxConfig.org

You can see breakdown of your memory usage percentage, VSZ and RSS (actual memory usage) per process using command

Code (Text):

ps aufxw

use small sample of output for nginx related processes

Code (Text):

USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND

root       897  0.0  0.7 645924 30376 ?        Ss   Apr08   0:00 nginx: master process /usr/local/sbin/nginx -c /usr/local/nginx/conf/nginx.conf
nginx      898  0.0  1.3 674596 53880 ?        S<   Apr08   0:06  \_ nginx: worker process
nginx      900  0.0  1.4 674596 54372 ?        S<   Apr08   0:07  \_ nginx: worker process
nginx      902  0.0  0.8 645924 31088 ?        S    Apr08   0:16  \_ nginx: cache manager process

If you want to look at per process id memory usage stats use pidstat i.e. for nginx worker process pid = 900 run pidstat every 1 sec for 5 runs and sed is just to replace your main hostname with word hostname so mask your domain for posting on public places

Code (Text):

pidstat -durh -p 900 1 5 | sed -e "s|$(hostname)|hostname|g"

example output for pid = 900 for nginx worker process showing %MEM used = 1.4% with RSS real memory usage at 54372 KB and virtual memory VSZ at 674596 KB

Code (Text):

pidstat -durh -p 900 1 5 | sed -e "s|$(hostname)|hostname|g"
Linux 3.10.0-693.21.1.el7.x86_64 (hostname)     04/09/2018      _x86_64_        (2 CPU)

#      Time   UID       PID    %usr %system  %guest    %CPU   CPU  minflt/s  majflt/s     VSZ    RSS   %MEM   kB_rd/s   kB_wr/s kB_ccwr/s  Command
 1523258123  1000       900    0.00    0.00    0.00    0.00     1      0.00      0.00  674596  54372   1.40      0.00      0.00      0.00  nginx

#      Time   UID       PID    %usr %system  %guest    %CPU   CPU  minflt/s  majflt/s     VSZ    RSS   %MEM   kB_rd/s   kB_wr/s kB_ccwr/s  Command
 1523258124  1000       900    0.00    0.00    0.00    0.00     1      0.00      0.00  674596  54372   1.40      0.00      0.00      0.00  nginx

#      Time   UID       PID    %usr %system  %guest    %CPU   CPU  minflt/s  majflt/s     VSZ    RSS   %MEM   kB_rd/s   kB_wr/s kB_ccwr/s  Command
 1523258125  1000       900    0.00    0.00    0.00    0.00     1      0.00      0.00  674596  54372   1.40      0.00      0.00      0.00  nginx

#      Time   UID       PID    %usr %system  %guest    %CPU   CPU  minflt/s  majflt/s     VSZ    RSS   %MEM   kB_rd/s   kB_wr/s kB_ccwr/s  Command
 1523258126  1000       900    0.00    0.00    0.00    0.00     1      0.00      0.00  674596  54372   1.40      0.00      0.00      0.00  nginx

#      Time   UID       PID    %usr %system  %guest    %CPU   CPU  minflt/s  majflt/s     VSZ    RSS   %MEM   kB_rd/s   kB_wr/s kB_ccwr/s  Command
 1523258127  1000       900    0.00    0.00    0.00    0.00     1      0.00      0.00  674596  54372   1.40      0.00      0.00      0.00  nginx

So go by free output for memory usage if on CentOS 7.

 

You have 2277 out of 5967MB available memory plenty it seems