its not the OS that gave me an extra 1900 points, however building my kernel to optimize and patch it for my system, still did have an impact on my score. but if you look at the different base scores i posted for the different kernels in detail, you'll see that the sngle threaded tests are all pretty much the same. where the difference lies is the multithreaded tests, which shows that the vanlla kernel wasn't quite taking advantage of the multiple cores quite right. honestly, one of the patches i added when i built the new kernel is what added the phenom achitecture support to the base kernel. for example, on my eee pc, when i built the kernel, i added smp support, which gives 2 more options, hyperthreading and multicore. i left them both in, which ended up causing me shutdown problems, because multicore was taking precedence over hyperthreading. so at shutdown, when it was trying to shut off a virtual core, it was treating it like a physical core, thus hanging the boot process. when i recompiled with HT only, that went away. so maybe something like that was afftecting my desktop as well in lowering my multithreaded performance
actually, if you look at my other
thread from when i showed the first benchmarks with different kernels before even overclocking. i did explain a few things. one of the things i mentioned was the task scheduler, BFS (brain f###), and i've noticed a difference with it in the new optimized kernel i'm running. its alot simpler than the CFS (completely fair) scheduler in that it gives more priority to active processes than trying to share the load equally, the results change depending on system load and overall time it takes to run a task will change depending on system load as well. but since running the new kernel i've noticed my video encoding times have improved like 50~%, give or take depending on the number of passes and filters. but before for just a basic 2 pass with the old kernel, it would encode at around 22fps, just under real time. now i see 40~fps. and to me it makes sense that how an OS manages things can change the outcome of a benchmark to a degree because benchmarks run a test, and calculate scores by data passed within a timeframe, the less your tasks that are being juggled by your hardware, the more ressources are allowed for the benchmark to use within that time frame, in simple terms at least. all i'm saying, is that perhaps the way tasks are managed can have an impact. but i think the main thing making a difference for me without overclocking is the actual phenom architecture support, not to mention better SMP support like i mentioned above.
im not saying geekbench is accurrate, i've dfinitely seen a few tests that seem off, but im still curious. and one thing mentioned on their site somewhere with the last version was that they improved the scoring chart somewhat, they did say themselves that the old version wasn't. that score you posted definitely looks off though. which does leave doubt as far as comparing systems to each other. but its still a good reference point for seeing how different things scale on a single system in terms of performance. its still a benchmark though, and differs from real world performance. but like a said, not a bad baseline for tweaking a single system, definitely not for comparing. looking over my phenom scores though, multithreaded performance seems to scale well against the clock speeds though. i'm still curious what you scored though
started pushing the phenom
Linear Mode
