Re: Asf fastforward performance

"Wael_Bakr" <WaelBakr@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:670D41C8-4875-40B0-9919-9FC26464E51F@xxxxxxxxxxxxxxxx
> [...]
> 3-what i realy don't understand is how can i get very smooth 2X
> with single processor and only 2.15X with dual processors , i
> think from my description above that i should ge getting very
> smooth 4X instead , does anyone know why i don't ?

In addition to the very useful comments provided by Alessandro, I'll point
out that as far as I can tell, you haven't established the CPU(s) as your
bottleneck. There are a number of components in a PC that can slow you
down, including but not limited to memory, PCI bus, hard disk, and AGP bus.
Adding more CPUs only helps if the only thing holding you back is the CPU.

If you think this is a problem that is worth solving, and you're willing to
put the hours into it that will be required, then what you really need to do
is to start actually profiling your code. That is, measuring the time
various elements of the code take. Done correctly, this will show you what
limit you're running into.

Given your system specifications (3.0Ghz), the fact that you are seeing a
CPU utilization of "only" 90% does suggest to me that the CPU isn't the
bottleneck. You can confirm this by looking at what's taking the other 10%.
If the System Idle "process" is consuming the bulk of the rest of the CPU
time, that tells you that you are spending time just waiting on something
(ie i/o), rather than computing something.

Furthermore, if you mean to say that with one CPU you get 97% utilization,
but with two CPUs you only get 90% utilization, that is an additional
indication that the CPU is the bottleneck with one CPU, but not with two.
However, it's also an indication that even if you removed the two-CPU
bottleneck, you're not going to go much faster before the CPUs become the
bottleneck again.

If you have something else consuming the 10%, the CPU might be the
bottleneck, and you should try to get that something else to stop running,
because it's distracting "your" CPU from the work you want it to do. :)
Fixing that, however, would still only result in a marginal increase in
performance.

Pete


.

Relevant Pages

  • Re: We quadrupled hardware power and reduced performance
    ... the bottleneck from Application/Query/Database Design -> Operating ... it seems that the high CPU ... evaluation system perfomance is to first analyze memory, ... HOW TO: Troubleshoot Application Performance Issues ...
    (microsoft.public.sqlserver.server)
  • Re: Upgrade to gigabyte NIC?
    ... bottleneck occurs in the communication, ... to focus on the CPU as the main hardware reason for it, ... well known Interrupt Livelock situation, ...
    (microsoft.public.backoffice.smallbiz)
  • Re: We quadrupled hardware power and reduced performance
    ... Seems to be CPU where the bottleneck is, ... > CPU, Memory, Disk i/o, Index information, long running queries etc. ... >> We did several tests with new box serving as server to other boxes. ...
    (microsoft.public.sqlserver.server)
  • Re: Bottleneck I/O or CPU
    ... Looks like a CPU bottleneck to me. ... VMSTAT shows most of the CPU time is being spent in OS for the ... but the sar output is showing slightly the opposite. ...
    (comp.unix.solaris)
  • RE: ATI vsync CPU usage
    ... the workaround technique and see if you get different results. ... workings of the DirectX runtime and how it uses the display driver. ... This results in a spin loop that consumes 100% CPU. ... windowed = 10-20% utilization ...
    (microsoft.public.win32.programmer.directx.graphics)