Testing processor performance

I have been trying to test cpu performance using Nbench and have found
that I get better results running XP on VMWare system using a quad
core than I do on an XP system running a dual core AM2 5200+ or
5000+ . I clocked the same system with the two different processors
and was only able to obtain results about half of what I was able to
achieve on an XP system running in VMWare on top of a quad core 9850
linus system. For measurement I was most interested in Floating point
and Integer MOPs/sec.
I originally saw the problem when running a database program and found
out it was five times slower on the system running on system A running
directly on dual core vs system B running in VMware on a quad core
9850. The VMware system has only one CPU allocated to it.
When I realized I had a problem, I suspected first the hard drive,
then the MB, then the CPU. All have been changed out, along with the
memory.
Any ideas on why I am seeing this difference? I just feel like there
is something I am not understanding here. I don't really want to
install a quad core into a system when it is only utilizing one CPU

wrote:
> I have been trying to test cpu performance using Nbench and have found
> that I get better results running XP on VMWare system using a quad
> core than I do on an XP system running a dual core AM2 5200+ or
> 5000+ . I clocked the same system with the two different processors
> and was only able to obtain results about half of what I was able to
> achieve on an XP system running in VMWare on top of a quad core 9850
> linus system. For measurement I was most interested in Floating point
> and Integer MOPs/sec.
> I originally saw the problem when running a database program and found
> out it was five times slower on the system running on system A running
> directly on dual core vs system B running in VMware on a quad core
> 9850. The VMware system has only one CPU allocated to it.
> When I realized I had a problem, I suspected first the hard drive,
> then the MB, then the CPU. All have been changed out, along with the
> memory.
> Any ideas on why I am seeing this difference? I just feel like there
> is something I am not understanding here. I don't really want to
> install a quad core into a system when it is only utilizing one CPU

The Phenom has a different cache structure. It has a shared L3. The
dual does not. The memory subsystem is also different. I believe the
Phenom has the ability to run the two channels independently (and is
probably configured that way by default), whereas the dual core would
have them linked together.

If the clock rates were identical, it would suggest your benchmarks
or activities have a high dependence on cache and memory.

Try disabling Cool N' Quiet, forcing both processor to run at full
clock speed all the time.

This thread was a discussion of the things needed to properly install
WinXP on AMD. With SP3 out, I expect "Update 4" is already taken care
of. Leaving the "CPU driver" ? Start reading the thread from back to
front, to get the latest scoop.

http://forum.notebookreview.com/showthread.php?t=60416

(Last post)
http://forum.notebookreview.com/show...=60416&page=67

Processor drivers are available here, for the various processor types.

http://www.amd.com/us-en/Processors/...30_182,00.html

Paul

Paul wrote:
> wrote:
>> I have been trying to test cpu performance using Nbench and have found
>> that I get better results running XP on VMWare system using a quad
>> core than I do on an XP system running a dual core AM2 5200+ or
>> 5000+ . I clocked the same system with the two different processors
>> and was only able to obtain results about half of what I was able to
>> achieve on an XP system running in VMWare on top of a quad core 9850
>> linus system. For measurement I was most interested in Floating point
>> and Integer MOPs/sec.
>> I originally saw the problem when running a database program and found
>> out it was five times slower on the system running on system A running
>> directly on dual core vs system B running in VMware on a quad core
>> 9850. The VMware system has only one CPU allocated to it.
>> When I realized I had a problem, I suspected first the hard drive,
>> then the MB, then the CPU. All have been changed out, along with the
>> memory.
>> Any ideas on why I am seeing this difference? I just feel like there
>> is something I am not understanding here. I don't really want to
>> install a quad core into a system when it is only utilizing one CPU
>
> The Phenom has a different cache structure. It has a shared L3. The
> dual does not. The memory subsystem is also different. I believe the
> Phenom has the ability to run the two channels independently (and is
> probably configured that way by default), whereas the dual core would
> have them linked together.
>
> If the clock rates were identical, it would suggest your benchmarks
> or activities have a high dependence on cache and memory.
>
> Try disabling Cool N' Quiet, forcing both processor to run at full
> clock speed all the time.
>
> This thread was a discussion of the things needed to properly install
> WinXP on AMD. With SP3 out, I expect "Update 4" is already taken care
> of. Leaving the "CPU driver" ? Start reading the thread from back to
> front, to get the latest scoop.
>
> http://forum.notebookreview.com/showthread.php?t=60416
>
> (Last post)
> http://forum.notebookreview.com/show...=60416&page=67
>
> Processor drivers are available here, for the various processor types.
>
> http://www.amd.com/us-en/Processors/...30_182,00.html
>
> Paul

I should get my terminology right.

http://www.xbitlabs.com/articles/cpu...-phenom_3.html

Ganged versus unganged.

"The second peculiarity of the integrated memory controller
is the fact that it is implemented as two independent 64-bit
controllers rather than a single 128-bit interface. This
allows the memory of Phenom platforms to work in two
modes: ganged and unganged. The first one is analogous to
the usual 128-bit dual-channel mode. The second one implies
that the CPU can deal with two independent 64-bit memory controllers
thus processing two memory requests simultaneously, which is good
for multi-threaded environments."

Linkage, refers to the relationship between core and memory clocks

"Phenom memory controller is also interesting because it runs at
a different frequency than the CPU clock: it uses its own clock
frequency multiplier and runs at 2GHz on all currently available
CPU models. They did it this way to ensure that the actual memory
frequency always matches the desired one, which was not always the
case with previous generation AMD processors. They frequency was
obtained by applying integer dividers to the CPU clock speed value.
So, the Phenom memory controller may set the DDR2 SDRAM frequency
at exactly 400, 533, 667, 800 or 1066MHz at any clock speeds."

HTH,
Paul

On Sep 29, 1:41 pm, Paul <nos...@needed.com> wrote:
> Paul wrote:
> > phwashing...@tx.rr.com wrote:
> >> I have been trying to test cpu performance using Nbench and have found
> >> that I get better results running XP on VMWare system using a quad
> >> core than I do on an XP system running a dual core AM2 5200+ or
> >> 5000+ . I clocked the same system with the two different processors
> >> and was only able to obtain results about half of what I was able to
> >> achieve on an XP system running in VMWare on top of a quad core 9850
> >> linus system. For measurement I was most interested in Floating point
> >> and Integer MOPs/sec.
> >> I originally saw the problem when running a database program and found
> >> out it was five times slower on the system running on system A running
> >> directly on dual core vs system B running in VMware on a quad core
> >> 9850. The VMware system has only one CPU allocated to it.
> >> When I realized I had a problem, I suspected first the hard drive,
> >> then the MB, then the CPU. All have been changed out, along with the
> >> memory.
> >> Any ideas on why I am seeing this difference? I just feel like there
> >> is something I am not understanding here. I don't really want to
> >> install a quad core into a system when it is only utilizing one CPU
>
> > The Phenom has a different cache structure. It has a shared L3. The
> > dual does not. The memory subsystem is also different. I believe the
> > Phenom has the ability to run the two channels independently (and is
> > probably configured that way by default), whereas the dual core would
> > have them linked together.
>
> > If the clock rates were identical, it would suggest your benchmarks
> > or activities have a high dependence on cache and memory.
>
> > Try disabling Cool N' Quiet, forcing both processor to run at full
> > clock speed all the time.
>
> > This thread was a discussion of the things needed to properly install
> > WinXP on AMD. With SP3 out, I expect "Update 4" is already taken care
> > of. Leaving the "CPU driver" ? Start reading the thread from back to
> > front, to get the latest scoop.
>
> >http://forum.notebookreview.com/showthread.php?t=60416
>
> > (Last post)
> >http://forum.notebookreview.com/show...=60416&page=67
>
> > Processor drivers are available here, for the various processor types.
>
> >http://www.amd.com/us-en/Processors/...30_182,00.html
>
> > Paul
>
> I should get my terminology right.
>
> http://www.xbitlabs.com/articles/cpu...-phenom_3.html
>
> Ganged versus unganged.
>
> "The second peculiarity of the integrated memory controller
> is the fact that it is implemented as two independent 64-bit
> controllers rather than a single 128-bit interface. This
> allows the memory of Phenom platforms to work in two
> modes: ganged and unganged. The first one is analogous to
> the usual 128-bit dual-channel mode. The second one implies
> that the CPU can deal with two independent 64-bit memory controllers
> thus processing two memory requests simultaneously, which is good
> for multi-threaded environments."
>
> Linkage, refers to the relationship between core and memory clocks
>
> "Phenom memory controller is also interesting because it runs at
> a different frequency than the CPU clock: it uses its own clock
> frequency multiplier and runs at 2GHz on all currently available
> CPU models. They did it this way to ensure that the actual memory
> frequency always matches the desired one, which was not always the
> case with previous generation AMD processors. They frequency was
> obtained by applying integer dividers to the CPU clock speed value.
> So, the Phenom memory controller may set the DDR2 SDRAM frequency
> at exactly 400, 533, 667, 800 or 1066MHz at any clock speeds."
>
> HTH,
> Paul

Thanks for the info. This is a heck of a lot farther than I got.
This is pretty freakin dizzying.
I'm not going to complain about AMD not having a web site where they
could be the information in concise format. Where there customers
could find it and apply to there rigs. I mean there are so many
operating systems that are running on AMD dual core and AMD Phenom
processors that it would be impossible for them to find one that might
cover 90%+ of the user base.

I forgot in my original post that one of the users was running the
same program on an AMD 3800+ chip in 1/5th the time mine was running,
so this is becoming a little frustrating.

I don't know if it is possible, but could it be throttling down the
CPU some how. Because when I run the program in VMware I see the CPU
usage go up to 85-95%. When I run it on the other system, I only see
10% max usage. Running though the same problem. On one system it
takes 12 minutes and on the other it takes 1.5 minutes.

wrote:
> On Sep 29, 1:41 pm, Paul <nos...@needed.com> wrote:
>> Paul wrote:
>>> phwashing...@tx.rr.com wrote:
>>>> I have been trying to test cpu performance using Nbench and have found
>>>> that I get better results running XP on VMWare system using a quad
>>>> core than I do on an XP system running a dual core AM2 5200+ or
>>>> 5000+ . I clocked the same system with the two different processors
>>>> and was only able to obtain results about half of what I was able to
>>>> achieve on an XP system running in VMWare on top of a quad core 9850
>>>> linus system. For measurement I was most interested in Floating point
>>>> and Integer MOPs/sec.
>>>> I originally saw the problem when running a database program and found
>>>> out it was five times slower on the system running on system A running
>>>> directly on dual core vs system B running in VMware on a quad core
>>>> 9850. The VMware system has only one CPU allocated to it.
>>>> When I realized I had a problem, I suspected first the hard drive,
>>>> then the MB, then the CPU. All have been changed out, along with the
>>>> memory.
>>>> Any ideas on why I am seeing this difference? I just feel like there
>>>> is something I am not understanding here. I don't really want to
>>>> install a quad core into a system when it is only utilizing one CPU
>>> The Phenom has a different cache structure. It has a shared L3. The
>>> dual does not. The memory subsystem is also different. I believe the
>>> Phenom has the ability to run the two channels independently (and is
>>> probably configured that way by default), whereas the dual core would
>>> have them linked together.
>>> If the clock rates were identical, it would suggest your benchmarks
>>> or activities have a high dependence on cache and memory.
>>> Try disabling Cool N' Quiet, forcing both processor to run at full
>>> clock speed all the time.
>>> This thread was a discussion of the things needed to properly install
>>> WinXP on AMD. With SP3 out, I expect "Update 4" is already taken care
>>> of. Leaving the "CPU driver" ? Start reading the thread from back to
>>> front, to get the latest scoop.
>>> http://forum.notebookreview.com/showthread.php?t=60416
>>> (Last post)
>>> http://forum.notebookreview.com/show...=60416&page=67
>>> Processor drivers are available here, for the various processor types.
>>> http://www.amd.com/us-en/Processors/...30_182,00.html
>>> Paul
>> I should get my terminology right.
>>
>> http://www.xbitlabs.com/articles/cpu...-phenom_3.html
>>
>> Ganged versus unganged.
>>
>> "The second peculiarity of the integrated memory controller
>> is the fact that it is implemented as two independent 64-bit
>> controllers rather than a single 128-bit interface. This
>> allows the memory of Phenom platforms to work in two
>> modes: ganged and unganged. The first one is analogous to
>> the usual 128-bit dual-channel mode. The second one implies
>> that the CPU can deal with two independent 64-bit memory controllers
>> thus processing two memory requests simultaneously, which is good
>> for multi-threaded environments."
>>
>> Linkage, refers to the relationship between core and memory clocks
>>
>> "Phenom memory controller is also interesting because it runs at
>> a different frequency than the CPU clock: it uses its own clock
>> frequency multiplier and runs at 2GHz on all currently available
>> CPU models. They did it this way to ensure that the actual memory
>> frequency always matches the desired one, which was not always the
>> case with previous generation AMD processors. They frequency was
>> obtained by applying integer dividers to the CPU clock speed value.
>> So, the Phenom memory controller may set the DDR2 SDRAM frequency
>> at exactly 400, 533, 667, 800 or 1066MHz at any clock speeds."
>>
>> HTH,
>> Paul
>
> Thanks for the info. This is a heck of a lot farther than I got.
> This is pretty freakin dizzying.
> I'm not going to complain about AMD not having a web site where they
> could be the information in concise format. Where there customers
> could find it and apply to there rigs. I mean there are so many
> operating systems that are running on AMD dual core and AMD Phenom
> processors that it would be impossible for them to find one that might
> cover 90%+ of the user base.
>
> I forgot in my original post that one of the users was running the
> same program on an AMD 3800+ chip in 1/5th the time mine was running,
> so this is becoming a little frustrating.
>
> I don't know if it is possible, but could it be throttling down the
> CPU some how. Because when I run the program in VMware I see the CPU
> usage go up to 85-95%. When I run it on the other system, I only see
> 10% max usage. Running though the same problem. On one system it
> takes 12 minutes and on the other it takes 1.5 minutes.

If you disable Cool N' Quiet, the machine should operate at the rated
clock speed. (Check the BIOS.)

You can also try the power monitor application, and see what it reports.
Another program would be CPUZ, but I don't know how it handles the
dynamic nature of Cool N' Quiet.

http://www.amd.com/us-en/assets/cont...onitor_123.zip

Paul

On Sep 29, 6:56 pm, Paul <nos...@needed.com> wrote:
> phwashing...@tx.rr.com wrote:
> > On Sep 29, 1:41 pm, Paul <nos...@needed.com> wrote:
> >> Paul wrote:
> >>> phwashing...@tx.rr.com wrote:
> >>>> I have been trying to test cpu performance using Nbench and have found
> >>>> that I get better results running XP on VMWare system using a quad
> >>>> core than I do on an XP system running a dual core AM2 5200+ or
> >>>> 5000+ . I clocked the same system with the two different processors
> >>>> and was only able to obtain results about half of what I was able to
> >>>> achieve on an XP system running in VMWare on top of a quad core 9850
> >>>> linus system. For measurement I was most interested in Floating point
> >>>> and Integer MOPs/sec.
> >>>> I originally saw the problem when running a database program and found
> >>>> out it was five times slower on the system running on system A running
> >>>> directly on dual core vs system B running in VMware on a quad core
> >>>> 9850. The VMware system has only one CPU allocated to it.
> >>>> When I realized I had a problem, I suspected first the hard drive,
> >>>> then the MB, then the CPU. All have been changed out, along with the
> >>>> memory.
> >>>> Any ideas on why I am seeing this difference? I just feel like there
> >>>> is something I am not understanding here. I don't really want to
> >>>> install a quad core into a system when it is only utilizing one CPU
> >>> The Phenom has a different cache structure. It has a shared L3. The
> >>> dual does not. The memory subsystem is also different. I believe the
> >>> Phenom has the ability to run the two channels independently (and is
> >>> probably configured that way by default), whereas the dual core would
> >>> have them linked together.
> >>> If the clock rates were identical, it would suggest your benchmarks
> >>> or activities have a high dependence on cache and memory.
> >>> Try disabling Cool N' Quiet, forcing both processor to run at full
> >>> clock speed all the time.
> >>> This thread was a discussion of the things needed to properly install
> >>> WinXP on AMD. With SP3 out, I expect "Update 4" is already taken care
> >>> of. Leaving the "CPU driver" ? Start reading the thread from back to
> >>> front, to get the latest scoop.
> >>>http://forum.notebookreview.com/showthread.php?t=60416
> >>> (Last post)
> >>>http://forum.notebookreview.com/show...=60416&page=67
> >>> Processor drivers are available here, for the various processor types.
> >>>http://www.amd.com/us-en/Processors/...30_182,00.html
> >>> Paul
> >> I should get my terminology right.
>
> >>http://www.xbitlabs.com/articles/cpu...-phenom_3.html
>
> >> Ganged versus unganged.
>
> >> "The second peculiarity of the integrated memory controller
> >> is the fact that it is implemented as two independent 64-bit
> >> controllers rather than a single 128-bit interface. This
> >> allows the memory of Phenom platforms to work in two
> >> modes: ganged and unganged. The first one is analogous to
> >> the usual 128-bit dual-channel mode. The second one implies
> >> that the CPU can deal with two independent 64-bit memory controllers
> >> thus processing two memory requests simultaneously, which is good
> >> for multi-threaded environments."
>
> >> Linkage, refers to the relationship between core and memory clocks
>
> >> "Phenom memory controller is also interesting because it runs at
> >> a different frequency than the CPU clock: it uses its own clock
> >> frequency multiplier and runs at 2GHz on all currently available
> >> CPU models. They did it this way to ensure that the actual memory
> >> frequency always matches the desired one, which was not always the
> >> case with previous generation AMD processors. They frequency was
> >> obtained by applying integer dividers to the CPU clock speed value.
> >> So, the Phenom memory controller may set the DDR2 SDRAM frequency
> >> at exactly 400, 533, 667, 800 or 1066MHz at any clock speeds."
>
> >> HTH,
> >> Paul
>
> > Thanks for the info. This is a heck of a lot farther than I got.
> > This is pretty freakin dizzying.
> > I'm not going to complain about AMD not having a web site where they
> > could be the information in concise format. Where there customers
> > could find it and apply to there rigs. I mean there are so many
> > operating systems that are running on AMD dual core and AMD Phenom
> > processors that it would be impossible for them to find one that might
> > cover 90%+ of the user base.
>
> > I forgot in my original post that one of the users was running the
> > same program on an AMD 3800+ chip in 1/5th the time mine was running,
> > so this is becoming a little frustrating.
>
> > I don't know if it is possible, but could it be throttling down the
> > CPU some how. Because when I run the program in VMware I see the CPU
> > usage go up to 85-95%. When I run it on the other system, I only see
> > 10% max usage. Running though the same problem. On one system it
> > takes 12 minutes and on the other it takes 1.5 minutes.
>
> If you disable Cool N' Quiet, the machine should operate at the rated
> clock speed. (Check the BIOS.)
>
> You can also try the power monitor application, and see what it reports.
> Another program would be CPUZ, but I don't know how it handles the
> dynamic nature of Cool N' Quiet.
>
> http://www.amd.com/us-en/assets/cont...AMD_Power_Moni...
>
> Paul

Thanks for your help.
I finally got my integer and floating point Mops/sec up, but the
program I was trying to run still operated slowly.

So for anyone that may see this yes you need the lates AMD Drivers to
get the best performance even with SP3.

Everything else worked fine, I was really impressed with the
responsiveness of the system. But whenever I tested the integer and
floating point operations I got really low numbers, until I installed
the AMD Drivers.

But for this system I am going to give up for now and get a Quad Core
and run XP in VMWare. The performance matches the values my cohort
was getting on an
AMD
3800+ and I'm really spent to much time on this.

Thanks again Paul.