search

Guichaguri / BetterFps

Performance Improvements
GNU Lesser General Public License v2.1
99 stars 22 forks source link

No difference between algorithms? #127

Open EnragedN3wb opened 5 years ago

EnragedN3wb commented 5 years ago

So over the past few weeks I've been trying to figure out which algorithm to use, but test results seem to be completely random...

I started out testing in batches of 20 to see which one works best for me. I'd run the test 20 times & record results to see which algorithm was fastest the most times. Problem was, with every batch of 20 results, a different algorithm was best... So I tried working with batches of 50... Same result. Then batches of 100... Same result.

Literally the ONLY thing I've learned from the 500+ tests I've done is that Java's algorithm is trash, & every other one, including vanilla, is exactly the same...

CPU: i5 3570k @4.6GHz RAM: 16GB DDR3 Dual Channel @1600MHz GPU: R9 290x

Is there any point to me using BetterFPS or should I just remove it & run 1 less mod? Anyone else have similar results or a similar system with different results?(Maybe something is wrong with mine?)

alexmercerind commented 5 years ago

Well Java Math is fastest one for me.

marharols commented 5 years ago

Run it and click benchmark, 90% of the time anything but java is good, Rivens Full was best for me.

Bearturnedhuman commented 4 years ago

Even though this was a year ago, maybe somebody will find this and it'll help them.

If you run the test a good 10 times and document the results in something like google spreadsheets, you can average all of the algorithms and find the best one for you.

Example: https://docs.google.com/spreadsheets/d/1rMzflEC8SlZ0k5jqnA2bWSnyUN04dbtNcPFsWqHDQu4/edit?usp=sharing

You can use that as a template, for me, the fastest algorithm was Riven's "Half" Algorithm. This is the most effective way I know to find which is the best. While there are different results every time, and there is almost always a different "better algorithm" it recommends. If you find the average ms and pick the one with the fastest response time you can pretty accurately set the best algorithm for your computer.