Kernel Impact on 3D Games

July 31st, 2007 by Crusader

There’s a followup article at KernelTrap relating a recent Linux kernel mailing list thread where benchmarking thenew Completely Fair Scheduler with various games was discussed:

From: Ingo Molnar [email blocked]
Subject: Re: [ck] Re: SD still better than CFS for 3d ?(was Re:2.6.23-rc1)

I regularly test native Linux games on CFS, and they all behavewell.


people are regularly testing 3D smoothness, and they find CFS good enough:

and that matches my experience as well (as limited as it may be). In general my impression is that CFS and SD are roughly on par when it comes to 3D smoothness.


Thanks Kenneth for the testing! I’ve created a graph out of yournumbers:

(it also includes the SD numbers you got with the turn-yield-into-NOPhack applied.)

One Response to “Kernel Impact on 3D Games”

  1. Anonymous Says:

    SD seems to work well at low to mid system load. But SD just can’t handle very heavy System load.
    CFS appears to be more consistent.

    I wonder if the numbers would be more consistent if the OS was compiled and optimised from scratch, (for that system)?

    Also, GCC-4.2.1 has MP support. Could that be a good test too with SD and CFS?

Leave a Reply

You must be logged in to post a comment.