i have built some beta builds based on a new kernel with pelt added. they only difference are, they come with different pre-defined settings like cpu max freq, mif values etc. some of them may be visible like max freq allowed but others are not
based on my tests on geekbench 5, they scored as high as 22xx (~10% increase) with multi cores and 70x (~15% increase) with single core, which both of them were new heights (for me) since the bm v2 released a few mths ago. and, it's not all about the scores but the smoothness, battery consumptions as well for instance, and in fact higher freq does not gurantee a better, overall performance due to the extra heat generated. i have seen huge degrades with higher cpu freq but more stable and higher ipc (12xk with cpu throttling test) with lower freq
now, i would like to invite you guys to help me to test them out thoroughly since i am unable to test them one by one or otherwise you will have to wait for a few years
. and in order not to screw up this thread, i tend to distribute them via pm and discussions will be carried out in another new thread
if you are brave enough and interested in taking the risks and share your findings together, pls drop me a pm. please also note that they are NOT supposed for normal users and experiemental, and they are NOT FREE! you will need to provide feedbacks!!
again, this is for you guys, i am more than satisifed with the current kernel
NOTE although i have tested them by myself, i cannot gurantee they are 100% safe. so you will have to bare all the risks, as usual
ps, a bit more info
i know may be some of you are frustrated, why this but not more oc and dealing with the throttling first?
well, as i have discussed with some of you guys, i wanna one for optimizing the overall performance instead of fast but inefficient kernel. in short, i am willing to scarifice for a few percentage gain in speed with relatively less power consumption per percentage gain in performance. i want to find the equalibrium in stead of speed chasing with substantial losses in battery power. i understand it is some what an ideal situation which may not be fruitful
and once this and other queuing things etc are solved, you can raise it as high as you like even if it is beyond the optimal point or with diminishing returns. it's up to you, no problem and no contradiction. on the other hand, if i go for the speed now and the new scheduler make you sad later, it will be irreversible (hard to reverse i mean), at least for me, kernel wise
wow, such a long post...
i hope you will understand my philosophy behind and the logic in the development, and of coz, my poor english
thanks a lot