kernel configuration for maximum gaming / desktop perfomance
hi,
im curious how to config the config file in order to achieve maximum performance for a gaming / desktop setting. things i already changed was setting cpu-scheduler to bfq, optimize kernel for size and change cpu type to core2 instead of generic - so 3 changes to the default config of liquorix amd64. but there are so many options i have no clue of. maybe someone is able to suggest some changes that would boost performance? my hardware is: :: Code ::
System: Host: biatchXFCE Kernel: 3.8-8.dmz.1-liquorix-amd64 x86_64 (64 bit, gcc: 4.7.2) Desktop: Xfce 4.10.0 (Gtk 2.24.13) Distro: Ubuntu 12.10 quantal Machine: Mobo: MICRO-STAR model: P35 Platinum(MS-7345) version: 1.0 Bios: American Megatrends version: V1.12 date: 12/30/2009 CPU: Dual core Intel Core2 Duo CPU E6750 (-MCP-) clocked at 3336.003 MHz Graphics: Card: Advanced Micro Devices [AMD] nee ATI Juniper [Radeon HD 5700 Series] bus-ID: 01:00.0 X.Org: 1.13.0 driver: fglrx Resolution: 1440x900@75.0hz GLX Renderer: ATI Radeon HD 5700 Series GLX Version: 4.2.12173 - CPC 12.10.17 Direct Rendering: Yes Network: Card: Realtek RTL8111/8168B PCI Express Gigabit Ethernet controller driver: r8169 ver: 2.3LK-NAPI port: e800 bus-ID: 04:00.0 Drives: HDD Total Size: 2000.4GB (1.0% used) Info: Processes: 152 Uptime: 4:52 Memory: 1441.8/3952.1MB Runlevel: 2 Gcc sys: 4.7.2 Client: Shell inxi: 1.8.4 thx! Back to top |
|||||
For gaming, what matters more than throughput is behavior, so all your changes except tuning the compiler for the core2 architecture would actually reduce performance.
Optimizing for size actually reduces throughput. This is not application throughput, but kernel throughput. This means the average time it takes for the process scheduler to do its job has increased. More time is spent calculating the next process to run, which takes time away from the application itself. For a game that needs a frame rendered in 16ms, each millisecond lost is a big deal. Now, on the other hand, if you're running on a system with an old cpu with a limited size of L2 cache, optimize for size is more optimal. The kernel itself can thrash the L2 cache enough to reduce application performance. This isn't a problem on modern x86-64 cpus. And last, I'm not clear about your cpu scheduler change. BFQ is a disk IO scheduler (aka elevator), and has nothing to do with process scheduling. Maybe you meant BFS? Back to top |
|||||
yes, thats right. i switched cpu scheduler to BFS. is CFS or BFS better for cpu scheduling when it comes to gaming mainly? i also noticed that the kernel optimized for size makes the system slower. so the only real thing i could do, is to set processor type to core 2 instead of generic x86_64? how much % performance would i gain by this?
thanks for your answer, master damentz. i bow before you. ^^ Back to top |
|||||
All times are GMT - 8 Hours
|