Performance difference on single vs multi core
 

I posted this originally on alt.comp.hardware.overclocking.amd

Since there are a number of people over here that know about CPU's
they might be able to enlighten me on what is happening here. I'm not
picking on AMD, I just don't have any Intel's to test it on.


We are seeing some significant performance differences on single vs
multi-core processors on AMD chips.

We have a database application that we are running on XP which also
does regression analysis.

I have asked the manufacturer of the software if they have seen this,
but they have not. They are using Clarion from SoftVelocity to build
the application.

on some systems I know the core others I do not

AMD Athlon Socket 939 (single) 4000+ 102 secs

AMD Athlon 64 X2 (Toledo) (single) 4800+ 145 secs

AMD Athlon 64 (Unknown) (dual) 3500+ 184 secs

AMD Black Edition (Brisbane) (dual) 5000+ 780 secs
(Overclocked to 3GHz)

AMD Athlon 64 (Brisbane) (dual) 5200 +

AMD Quad Phenom 9950 810 secs

We then installed the application on an operating system installed as
guest in VMWare

AMD Quad Phenom 9950 92 secs

Any idea on what is going on here. This really doesn't make sense. I
would expect maybe a little performance hit but 5 to 6 times
different..

The only hint I can see of the issue may be the run made on the
Toledo. While it's clock speed is faster It's performance is still
worse than a 4000+. My understanding is that a Toledo is a dual core
with one core disabled.

Performance difference on single vs multi core
 

wrote:
I posted this originally on alt.comp.hardware.overclocking.amd

Since there are a number of people over here that know about CPU's
they might be able to enlighten me on what is happening here. I'm not
picking on AMD, I just don't have any Intel's to test it on.


We are seeing some significant performance differences on single vs
multi-core processors on AMD chips.

We have a database application that we are running on XP which also
does regression analysis.

I have asked the manufacturer of the software if they have seen this,
but they have not. They are using Clarion from SoftVelocity to build
the application.

on some systems I know the core others I do not

AMD Athlon Socket 939 (single) 4000+ 102 secs

AMD Athlon 64 X2 (Toledo) (single) 4800+ 145 secs

AMD Athlon 64 (Unknown) (dual) 3500+ 184 secs

AMD Black Edition (Brisbane) (dual) 5000+ 780 secs
(Overclocked to 3GHz)

AMD Athlon 64 (Brisbane) (dual) 5200 +

AMD Quad Phenom 9950 810 secs

We then installed the application on an operating system installed as
guest in VMWare

AMD Quad Phenom 9950 92 secs

Any idea on what is going on here. This really doesn't make sense. I
would expect maybe a little performance hit but 5 to 6 times
different..

The only hint I can see of the issue may be the run made on the
Toledo. While it's clock speed is faster It's performance is still
worse than a 4000+. My understanding is that a Toledo is a dual core
with one core disabled.

Start by running a pure compute benchmark. SuperPI runs on a single
core. Select enough digits of PI when running it, so that the
size of the cache does not influence the results. For example,
run 1 million digits and 32 million digits.

http://www.xtremesystems.com/pi/

http://www.xtremesystems.com/pi/super_pi_mod-1.5.zip

AMD processors feature Cool N' Quiet, use a "CPU driver" downloadable
from the AMD site, and in the hardware, have a programmable
multiplier. By making multiplier and voltage changes (FID and VID),
the processor can change from a full power state, to a power saving state,
up to something like 30 times a second.

When there are problems, sometimes an application will end up
"stuttering". For example, Anandtech, in their last series
of movie playback benchmarks, found a difference in playback
smoothness, with CNQ enabled and disabled. So they had to
run benchmarks under both conditions, on the assumption a
****ed off customer would turn off CNQ, to get the smoothness
they wanted. This causes a higher average power dissipation
on the processor (important in the HTPC market).

I don't know anything about which application is best to monitor
the processor state. There is an AMD Power Monitor application
here, and presumably it can tell you the currently used FID and VID.

http://www.amd.com/us-en/Processors/..._15259,00.html

Maybe after you've benchmarked with SuperPI, and compare clock speeds
from the various processors, you'll get a better feeling for where
to look next. Perhaps there are peripheral differences between
boxes ? Faster disks ? More memory for an application level cache ?
And so on.

I have run into the odd person, who has done everything to set up
their AMD processor in terms of CNQ, CPU drivers, Microsoft patches
and the like. And for some reason, the processor stays in "low gear".
Which is why it would be nice to be able to monitor what speed the
thing is running at.

Also, try running your processor hardware descriptions through the
list here. The "Toledo Single" stops at 4000+.

http://en.wikipedia.org/wiki/List_of...icroprocessors
http://en.wikipedia.org/wiki/List_of...icroprocessors

AMD has their own list of processors, but surprisingly, choose
not to list everything they've ever made. (And I'm not referring
to the one-off lots they ship to OEMs - their server processor
list is far from complete.)

http://www.amdcompare.com/

Paul

Performance difference on single vs multi core
 

On Oct 29, 4:35*pm, Paul wrote:
wrote:
I posted this originally on alt.comp.hardware.overclocking.amd

Since there are a number of people over here that know about CPU's
they might be able to enlighten me on what is happening here. *I'm not
picking on AMD, I just don't have any Intel's to test it on.

We are seeing some significant performance differences on single vs
multi-core processors on AMD chips.

We have a database application that we are running on XP which also
does regression analysis.

I have asked the manufacturer of the software if they have seen this,
but they have not. *They are using Clarion from SoftVelocity to build
the application.

on some systems I know the core others I do not

*AMD Athlon Socket 939 (single) 4000+ * * * 102 secs

*AMD Athlon 64 X2 (Toledo) (single) 4800+ * * * * *145 secs

AMD Athlon 64 (Unknown) (dual) 3500+ * * * * * * * 184 secs

AMD *Black Edition (Brisbane) (dual) 5000+ * * * * 780 secs
(Overclocked to 3GHz)

AMD *Athlon 64 (Brisbane) (dual) 5200 +

AMD Quad Phenom 9950 * * * * * * * * * * * * * * * * * *810 secs

We then installed the application on an operating system installed as
guest in VMWare

AMD *Quad Phenom 9950 * * * * * * * * * * * * * * * * * * *92 secs

Any idea on what is going on here. *This really doesn't make sense. *I
would expect maybe a little performance hit but 5 to 6 times
different..

The only hint I can see of the issue may be the run made on the
Toledo. * While it's clock speed is faster It's performance is still
worse than a 4000+. *My understanding is that a Toledo is a dual core
with one core disabled.

Start by running a pure compute benchmark. SuperPI runs on a single
core. Select enough digits of PI when running it, so that the
size of the cache does not influence the results. For example,
run 1 million digits and 32 million digits.

http://www.xtremesystems.com/pi/

http://www.xtremesystems.com/pi/super_pi_mod-1.5.zip

AMD processors feature Cool N' Quiet, use a "CPU driver" downloadable
from the AMD site, and in the hardware, have a programmable
multiplier. By making multiplier and voltage changes (FID and VID),
the processor can change from a full power state, to a power saving state,
up to something like 30 times a second.

When there are problems, sometimes an application will end up
"stuttering". For example, Anandtech, in their last series
of movie playback benchmarks, found a difference in playback
smoothness, with CNQ enabled and disabled. So they had to
run benchmarks under both conditions, on the assumption a
****ed off customer would turn off CNQ, to get the smoothness
they wanted. This causes a higher average power dissipation
on the processor (important in the HTPC market).

I don't know anything about which application is best to monitor
the processor state. There is an AMD Power Monitor application
here, and presumably it can tell you the currently used FID and VID.

http://www.amd.com/us-en/Processors/...0,,30_182_871_...

Maybe after you've benchmarked with SuperPI, and compare clock speeds
from the various processors, you'll get a better feeling for where
to look next. Perhaps there are peripheral differences between
boxes ? Faster disks ? More memory for an application level cache ?
And so on.

I have run into the odd person, who has done everything to set up
their AMD processor in terms of CNQ, CPU drivers, Microsoft patches
and the like. And for some reason, the processor stays in "low gear".
Which is why it would be nice to be able to monitor what speed the
thing is running at.

Also, try running your processor hardware descriptions through the
list here. The "Toledo Single" stops at 4000+.

http://en.wikipedia.org/wiki/List_of...icroprocessors

AMD has their own list of processors, but surprisingly, choose
not to list everything they've ever made. (And I'm not referring
to the one-off lots they ship to OEMs - their server processor
list is far from complete.)

http://www.amdcompare.com/

* * Paul

Thanks for the reply.
I'm probably going to re-install the 5000+ chip in the system this
weekend and give it another whirl. Right now it's running the phenom
processor with VMWare.
But I would still like to find out what the issue is.