If you want to know what kind of bump you will get from a 500 MHz overclock, then first do a 500MHz underclock and see if you notice anything. It's relatively safe and very easy to do. It just takes a two line batch file:
Code:
powercfg -setacvalueindex SCHEME_BALANCED SUB_PROCESSOR PROCTHROTTLEMAX 85
powercfg.exe -setactive SCHEME_CURRENT
The 85 is a percent of whatever your stock clock is. For a 2600K, this would put you at about 2.9GHz, not counting turbo. Use CPU-Z to verify that it is clocking where it should.
Frequency scaling is not linear, you cannot assume that if you downclock 500Mhz, the performance lost will be same as performance gained when you overclock 500Mhz.
Let's think of CPU stock(not turbo)clock rate a baseline, that means that the more you offset this value, the change becomes less apparent. It means that if 500Mhz overclock will give you 5% increase in performance, overclocking additional 500Mhz will give let's say another 3% gain, not 5. It is also true that downclocking will cause higher performance decrease than overclocking it would give higher performance increase.
In case of downclocking by 500MHz you would likely lose 15% of your CPU's stock while going down another 500 could drop your performance by 40% or more.
Downclocking would be good idea in terms of finding best performance per watt, because stock clock is usually highest safe clock CPU is capable of withstanding. You can easily downclock your 95W CPU by few bins down and its TDP will drop to 35-50W area, which is over 50% decrease, while it will also lose some performance, it will still perform very well.
That's why hardcore overclockers who run extreme cooling and so on, have their CPUs perform only by few % faster than someone running mediocre OC on air, although they increased frequency by another few thousands. Both up and down, at some point specific for each CPU, frequency will stop to affect performance.
Facebook
Twitter