quad core tweaks when compiling linux kernel?

[graysky]

Can anyone point me to quad-core (or multicore) specific options I can compile into my kernel?

Thus far, I found the following:

Processor type and features ---> Processor family ---> (X) Core 2/newer Xeon
Processor type and features ---> Timer Frequency ---> (X) 1000 HZ
Processor type and features ---> Preemption Model ---> (X) Preemptible Kernel (Low-Latency Desktop)

Enabling these gave me a little boost when doing x264 video encoding via handbrake, but I still don't get approx 80 % CPU usage when encoding (80 % ave on each core).

Thanks all!

[davidletterboyz]

Hi.
I assume you are using an Intel Core 2 Quad series CPU. I think the config u showed is OK, as long as u enable SMP and enter a value at "Maximum number of CPUs" to be no lesser than 4)

I guess the reason handbrake does not use up above 80% CPU utilization is because it is not a fully CPU bound program. There are times when process are waiting for I/O events (like waiting to read the hard disk).

[graysky]

I guess the reason handbrake does not use up above 80% CPU utilization is because it is not a fully CPU bound program. There are times when process are waiting for I/O events (like waiting to read the hard disk).

You'd think that, but doing the same encode to the same source using the Windows version (Windows XP 64-bit Ed) encodes faster than doing it under Debian/Lenny-amd64.

Hardware details: X3360 @ 3.40 GHz (8.5x400), 2x2 GB @ 5-5-5-15 running at 1,000 MHz (4:5), SATA2 HDDs, DFI LT-P35-T2R motherboard.

handbrake-win64
Run1: 233.5 fps (pass 1) and 41.8 fps (pass 1)
Run2: 234.8 fps (pass 1) and 41.8 fps (pass 2)

Time to encode both passes: 03:29

handbrake-linux64
Run1: 192.0 fps (pass 1) and 37.8 fps (pass 2)
Run2: 193.8 fps (pass 1) and 37.7 fps (pass 2)

Time to encode both passes: 03:54

That's about 12 % slower under LINUX. I can't believe that WIndows XP is more efficient at using a quad-core processor than Linux!