Quote:
Most of them came from the classic mb actually. So this seems to confirm that they had very little effect on performance - it's still good that they are gone because it simplifies things.
You could try out the new MB again - I optimized it quite a lot.
Perhaps it confirms. I don't know. There's still the issue of how you judge what is optimized, and whether or not altering your process to test for normal usage scenarios vs. filters in isolation would reveal better performance for cache-limited (< 2MB per core) systems. It may be that 1MB cache is just insufficient no matter what, but as time has passed, there's less and less incentive to focus on 1MB systems. I still think you should change the testing methodology though.
I can attempt to try the new MB. It will probably be fine, so long as I don't attempt to use loudness. That was somewhat the point of Pusher - a fairly clean, but relatively loud setting that doesn't use loudness / advanced clipper.
Edit: Turning off declipper and oversampling in Bojcha's newest preset has the CPU load at about 70%, which is approximately where the classic MB + Advanced Clipper goes. The preset only uses 7 bands though, but perhaps that is enough. I think I had tried with 9 bands + Advanced Clipper, and that didn't go so w..w..w..w...we...we...we..well...l...l...l.
