Tír Na HiFi

nige2000 wrote:

sbgk wrote:uploaded 2496 5.31 avx2 - 24/96 seems to highlight weaknesses in the sound, maybe due to the greater data volumes ?

Versions might not be directly transferable to hi res

the only difference is the buffer size, so no of times the loop runs. 24 times for 1 24/96 buffer and 8 for 16/44

Noone hear anything interesting about the 3 ms version
Maybe not perfect but still

sbgk wrote:uploaded 5.35 which is the same as 5.31, but without the setting which may have been affecting the treble.

5.34 is a 3ms version

uploaded 5.36 which is same as 5.27, but with the enhanced treble setting of 5.31 so comparison should be easier.

that's enough versions, I like 5.31, haven't listened to 5.36 yet, but hopefully there will be a consensus

I'm amazed at the sq of these last few versions, didn't think wasapi would deliver that.

maybe I should do a kickstarter for finishing off mqn. the guy who did it for a meal raised £50 k.

I would subscribe.

Especially if it would give you time to make an installation guide which would increase the number of folks willing to give it a try.

As long as you do not overcomplicate it for commercial purposes. Still wish you would put a polarity switch within it!

For me, in my system: 5.31avx2>>5.27avx2>5.35avx2=5.38avx2
Listening to jazz (Oscar Peterson)and rock (prog), 5.31 is just much more detailed and the bass is incredible, very tuneful and rhythmic. Music feels alive. A big step forward.The other versions are very good but 5.31 is much better. I'm delighted.

Adolfo

uploaded 5.42 and 5.43 avx2

experiments in variable and label names, had tried before and lower case single letter made the sound softer,
so tried single upper case and the sound was a bit better, then tried reordering the labels so the most frequent loops were the lowest letters so instead of A,B,D it was D,A,B, can't use C. So there is an outer loop and 2 inner loops in the assembly code.

anyway which is best 5.42 or 5.43 ?

and if anyone can come up with a reason... the difference is not subtle.

5.42 was G DEF where G is the procedure name D is the outer loop and E, F were the inner loops
5.43 is G DAB
5.44 is A BDE - procedure name renamed to A
5.45 is G ABD
5.46 is A EBD

think 5.44 sounds promising, this is an added complication I never would have thought about, but there it is.

5.44 and 5.46 sound better and I think 5.44 just shades it.

Any chance us Sub Standard Ever-hopeful (SSE) followers can have an occasional bite of the cherry?

jrling wrote:Any chance us Sub Standard Ever-hopeful (SSE) followers can have an occasional bite of the cherry?

+1

John Dot wrote:

jrling wrote:Any chance us Sub Standard Ever-hopeful (SSE) followers can have an occasional bite of the cherry?

+1

+1

Gordon

I like your 5.47
Great job

Don't have time right now to test every single version, but took some apparent milestones again, all 1644 avx2:

5.44 > 5.47 > 5.40 > 5.30 > 5.37

How often during the past months have I been amazed by the fact that even more improvement could be squeezed out of already great sounding releases again and again. And it still goes on!

5.30 (and 5.31) was a real milestone, better than all previous versions. Then it went down a bit again for me, but it picked up again with 5.40, improving with 5.44, then declining again. Not sure what was the step from 5.44 to 5.47, but that didn't work for me. I'll spend some more time on the range from 5.40 up to 5.46 to see exactly which one's my favourite.

Also tested some 2496, which all sound much better on my system than the 1644, because my DACs run on native 2496 and I pre-process my sound files, which means 2496 will be more accurate than a dithered 1644. There aren't that many hires versions available, but this is my preference of the latest avx2 releases:

5.31 > 5.47 > 5.37 > 5.07

I'll also repeat that avx sounds better, more organic, fluid and musical, than avx2 on my system.