info
Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
Sabertooth and a AMD 8350 speed issues
8 views
Skip to first unread message
Vince Coen
Nov 4, 2013, 12:48:27 PM11/4/13
to
Hello All!
Having set up above system I am noticing that loads get very high according
to top/htop but the load is NOT spread between all the cores (it is an 8
core) with 16Gb (1600) Corsait Vengeance Pro Ram.
Is there a setting or two that is not correct in the UEFI bios or in the
Mageia v3 set up that I have not done?
I was expecting a very serious improvement over my older system bsaed
around a Intel Dual 2 Core E6600.
Both systems are/were set to stock speeds e.g., no overclocking as a small
performance increase of 10% etc does not justify the increased risk of
CPU/Ram failures.
Thanks for your help,
Vince
Having set up above system I am noticing that loads get very high according
to top/htop but the load is NOT spread between all the cores (it is an 8
core) with 16Gb (1600) Corsait Vengeance Pro Ram.
Is there a setting or two that is not correct in the UEFI bios or in the
Mageia v3 set up that I have not done?
I was expecting a very serious improvement over my older system bsaed
around a Intel Dual 2 Core E6600.
Both systems are/were set to stock speeds e.g., no overclocking as a small
performance increase of 10% etc does not justify the increased risk of
CPU/Ram failures.
Thanks for your help,
Vince
Vince Coen
Nov 4, 2013, 12:49:25 PM11/4/13
to
Hello Jim!
04 Nov 13 16:11, Jim Beard wrote to All:
> The change from single-processor to dual-processor is typically
> dramatic. A change from dual-processor to quad-processor is
> usually barely noticeable unless the machine is a server under
> heavy load and you have the monitoring tools to actually
> recognize the effects of tinkering with settings.
> The change from quad-core to 8-core will be trivial in most
> circumstances. If you have massive amounts of
> massively-cpu-intensive processing to do, and rewrite your
> applications for optimized cpu-sharing, the results can be rewarding.
I must admit I have not done any CPU intensive work but I did run some
backups/restores from the original 3Tb HDD (E6600 system) to the AMD system
with a 1Tb HDD. Two running at the same time and the load went up to 5.0+.
Looking at both top and htop which seems to give differing output regarding
each core at best only cores 1 & 3 had any real load but well below 50% and
cores 2 & 4 had slight load. I could not see what each core was doing (Is
there a tool for that?). It would be a useful function to run a bash script
against a specific core or get the kernel to distribute the load across the
cores but again I cannot see how one does that BUT the kernel should be
doing all this without user input (as it were).
I will be running thw Windows 7 inbuilt benchmark against both boxes to see
what the numbers are however, I suspect that these tests have a bias in one
direction or another and not just core loadings and throughput etc.
Regardless even if the Windows test shows a strong difference between the
two it does not indicate what is happeing (or not) under Linux.
There has to be suitable tools for dealing with these issues under Linux
without me having to run a JCL tool (al la IBM) on the boxes.
Says, Vince wishfully.
Vince
04 Nov 13 16:11, Jim Beard wrote to All:
> The change from single-processor to dual-processor is typically
> dramatic. A change from dual-processor to quad-processor is
> usually barely noticeable unless the machine is a server under
> heavy load and you have the monitoring tools to actually
> recognize the effects of tinkering with settings.
> The change from quad-core to 8-core will be trivial in most
> circumstances. If you have massive amounts of
> massively-cpu-intensive processing to do, and rewrite your
> applications for optimized cpu-sharing, the results can be rewarding.
I must admit I have not done any CPU intensive work but I did run some
backups/restores from the original 3Tb HDD (E6600 system) to the AMD system
with a 1Tb HDD. Two running at the same time and the load went up to 5.0+.
Looking at both top and htop which seems to give differing output regarding
each core at best only cores 1 & 3 had any real load but well below 50% and
cores 2 & 4 had slight load. I could not see what each core was doing (Is
there a tool for that?). It would be a useful function to run a bash script
against a specific core or get the kernel to distribute the load across the
cores but again I cannot see how one does that BUT the kernel should be
doing all this without user input (as it were).
I will be running thw Windows 7 inbuilt benchmark against both boxes to see
what the numbers are however, I suspect that these tests have a bias in one
direction or another and not just core loadings and throughput etc.
Regardless even if the Windows test shows a strong difference between the
two it does not indicate what is happeing (or not) under Linux.
There has to be suitable tools for dealing with these issues under Linux
without me having to run a JCL tool (al la IBM) on the boxes.
Says, Vince wishfully.
Vince
Wes Newell
Nov 5, 2013, 10:41:33 AM11/5/13
to
On Mon, 04 Nov 2013 17:49:25 +0000, Vince Coen wrote:
> There has to be suitable tools for dealing with these issues under Linux
> without me having to run a JCL tool (al la IBM) on the boxes.
Gkrellum will give you real times loads on all cores plus other system
> There has to be suitable tools for dealing with these issues under Linux
> without me having to run a JCL tool (al la IBM) on the boxes.
info. There are cpu frequency panel applets to show the frequency of each
core in real time. If you just want snapshot data, there are many CL
options for that. cat /proc/cpuinfo will give you info on all cores.
cpufreq-aperf cpufreq-info cpufreq-selector cpufreq-set are CL
tools for users for various actions. To find all commnds available to you
type each letter ay CL and then hit the Tab key. Root will have more than
user.
[wes@mythtv ~]$ c
Display all 159 possibilities? (y or n)
0 new messages