Re: FreeBSD is very slow when Memory chip sizes are imbalanced in slots

On Mon, Mar 18, 2013 at 10:30 AM, Mehmet Erol Sanliturk
<m.e.sanliturk_at_gmail.com> wrote:
>
>
> On Mon, Mar 18, 2013 at 3:01 AM, Tom Evans <tevans.uk_at_googlemail.com> wrote:
>>
>> On Sun, Mar 17, 2013 at 8:26 PM, Mehmet Erol Sanliturk
>> <m.e.sanliturk_at_gmail.com> wrote:
>> > Dear All ,
>> >
>> > Previously , in the following message , I have mentioned effect of
>> > memory
>> > chip placement on execution speed :
>> >
>> >
>> > http://lists.freebsd.org/pipermail/freebsd-current/2012-February/031836.html
>> > Effect of Processor and Memory on KDE4 execution
>> >
>> > speed<http://lists.freebsd.org/pipermail/freebsd-current/2012-February/031836.html>
>>
>> These seems to be more than different memory slot allocation between
>> those two boxes.
>>
>> Can you reproduce this on the one labelled 'FAST' by assigning memory
>> in the same manner as it is assigned in the one labelled 'SLOW'?
>>
>> >
>> >
>> > The above thread did not produce any usable result .
>> >
>> > The problem is persisting over 9.1 and 10.0 current .
>> >
>> > My opinion is that , it is NOT related to KDE only .
>> >
>> > After X is started , any desktop is behaving very slowly .
>> > This is also visible in PC-BSD and GhostBSD .
>>
>> This is very nebulous. What is 'very slowly'? Is there a test you can
>> run that is independent of X, KDE, etc that demonstrates this?
>>
>> One thing that KDE does require (iirc - from about 5 years ago,
>> probably wrong now) is that since KDE is C++, it spends a lot of time
>> loading executables/libraries into memory and prelinking them. If you
>> have dramatically lowered your RAM bandwidth, then this stage could
>> take a lot longer.
>>
>> One thing that could cause memory bandwidth to lower is by installing
>> mismatched modules. The BIOS will set all RAM up at the same speed,
>> the lowest that all of the installed RAM supports. If you fill the RAM
>> slots with mismatched modules of different sizes, it may also not
>> enable dual channel memory, further reducing the RAM bandwidth.
>>
>> Because of this, I think it is a jump to go from "My computer runs
>> slow when I put these bits of RAM in" to "FreeBSD always runs slow
>> when there is mismatched RAM".
>>
>> If you find out what is slow on FreeBSD - eg RAM bandwidth -  you can
>> then test the same thing in Linux. If Linux shows the same slowdown
>> from fast to slow, then I'm sorry, that's a hardware defect. If, on
>> the other hand, Linux is just as fast in both configurations, then I'm
>> sure a lot of people would be interested as to why.
>>
>> Cheers
>>
>> Tom
>
>
> I think , all of the answers to your questions may be found in the above
> referenced thread messages :

Nope, you tested 'running KDE' and on different computers found out
that one runs at a different speed than another. You've not done
anything else to determine why or because of what.

You're the one seeing this problem. If you want it fixed, you will
need to do some work to determine what is causing it. All you are
saying now is "When I build this complex combination of hardware and
software, it's slow. Fix my special case".

>
>  Every possible combination has been tried , and identified that the problem
> is different memory chip sizes for FreeBSD ( v9.0 , v9.1 , v10.0 )  (
> GhostBSD , PC-BSD , v9.0 , v9.1 ) :
>
> Channel A : Slot 1 : 2 GB
>                   Slot 2 : 1 GB
>
>
> Channel B : Slot 1 : 2 GB
>                   Slot 2 : 1 GB
>
>
> All of the memory chips : Kingston HyperX , same  clock frequency .
> Memory placement kind is correct .

You say this, have you actually measured/checked. sysutils/dmidecode
will interrogate your BIOS and tell us what it thinks is installed in
each RAM socket. It is not uncommon for RAM to say one thing on the
outside, and report something completely different to the BIOS.

>
> There is NO any hardware defect .
>
> Linux is insensitive to such different memory chip sizes ( I am using Fedora
> , CentOS , Mandriva , Mageia , OpenSUSE , Arch Linux , Puppy Linux , and
> some others ... )

And you've run the tests which clearly demonstrate this? Or you've
installed KDE, and it's not been 'too slow'?

I'm trying to get you to approach a more scientific approach to this.
Hopefully, this would lead to some actual conclusions and things that
can be improved, rather than "FreeBSD is slow".

You've got a problem when you have mismatched memory, your computer runs slower.
Is the memory running slower?
Test memory bandwidth in both 'fast' and 'slow' configurations. Is
there a difference?
Apparently linux is unaffected.
Test memory bandwidth in both fast and slow configurations on linux.
Is there a difference?

Doing those 4 steps should tell us more about your actual problem, and
might spur an idea in someone. You can use ramspeed to test ram speed
in both linux and freebsd:

http://alasir.com/software/ramspeed/

Problems with your memory modules should produce testable scenarios
that don't involve X and KDE.

Cheers

Tom