MQN

[sbgk]

uploaded control haswell v26 which has the c:\mqnplayer removed

[elaprince]

Thank you Gordon
Now I can play no problem

[sbgk]

tonight's supercharged offering - play haswell v44, based on v39, but uses a few tricks I learnt from v40-v43

might be better than v39

now at 9 uops, v39 had 11 and before that was 13.

forgot about loop alignment optimisation in v44 - uploaded v45

the type of instruction immediately before the loop entry affects things as it is read when looping, so even though in v44 and v45 the loop entry was aligned on 16 bytes because v44 had a long instruction it affected the sq. v45 just has a 1 byte instruction immediately before the loop entry, which is optimal.

[sbgk]

I did not test all the versions but I did test v39 and I think it is very good but I like it more with control 19 (compared against 25 only). 39 with control 19 has more energy then control 25, also more clarity, 25control is slightly mushy and too polite.

LDN15 with control 7 and loader 7 is still the best in my book but you are narrowing the margin. LDN15 is another step-up in clarity, both width and depth of sound stage and bass is more extended and realistic.

The difference are actually very easy to be heard on any high quality acoustic recording like Chris Jones - Long After You' re Gone (https://www.youtube.com/watch?v=D-h6MoF7HLA) or high quality electronica as well.

Cheers

the music I'm testing it against is a bit more challenging in terms of speed and contrast between bass and treble, I still think I'm hearing better/more detail with v45 etc

lack of noise is sometimes judged to be too polite, I judge it by how many layers I can uncover in quiet passages, which again, I think v25 is better at.

[sebna]

This track is not my usual test track. I just added it to the mix as I was growing tired of my usual test tracks ;)

Anyway when I test I very rarely listen the whole track. Usually I start comparisons by comparing just previously cut and trimmed snippets of tracks with fragments which are good for testing of different aspects of sound reproduction. Usually those fragments last anything between 5 to 20 seconds and can be taken from middle of the original track if needs be. Switching between version has to be fast and there is no point to wait 2minutes for the fragment of the track you are really interested for comparison as there is high chance you will forget some of information by then from your first listen with other version of MQN one is comparing against.

Once I am finished with my snippets I usually listen first minute or two of few of my test tracks to confirm what I learned from the snippets comparison. If I really like what I hear I will leave the version for the evening but even then I will do occasional random switch to my ref version which currently still is LDN15.

ATM I compare each version against LDN15 (conrtol, loader: 7) but if I like one of the new version more then others I would also pit them against each other ;)

Anyway it is time consuming process so I do not do it every day. Usually once or twice a week. But I would do quick test of "recommended" versions in between bigger testing sessions.

That is how I do it.

Cheers

[janh]

Listened to play v41, 43 and 45, all with control/loader v25
I heard: v45>v43>v41
More clarity to 45 and finer music.

Cheers
Jan H.

[Aleg]

...
More clarity to 45 and finer music.

Cheers
Jan H.

+1

[sbgk]

there was minor alignment error with v45, so v46 is another attempt at getting the alignment and instructions in the best order, maybe slightly more solid bass.

[sbgk]

This track is not my usual test track. I just added it to the mix as I was growing tired of my usual test tracks ;)

Anyway when I test I very rarely listen the whole track. Usually I start comparisons by comparing just previously cut and trimmed snippets of tracks with fragments which are good for testing of different aspects of sound reproduction. Usually those fragments last anything between 5 to 20 seconds and can be taken from middle of the original track if needs be. Switching between version has to be fast and there is no point to wait 2minutes for the fragment of the track you are really interested for comparison as there is high chance you will forget some of information by then from your first listen with other version of MQN one is comparing against.

Once I am finished with my snippets I usually listen first minute or two of few of my test tracks to confirm what I learned from the snippets comparison. If I really like what I hear I will leave the version for the evening but even then I will do occasional random switch to my ref version which currently still is LDN15.

ATM I compare each version against LDN15 (conrtol, loader: 7) but if I like one of the new version more then others I would also pit them against each other ;)

Anyway it is time consuming process so I do not do it every day. Usually once or twice a week. But I would do quick test of "recommended" versions in between bigger testing sessions.

That is how I do it.

Cheers

agree that it's getting very difficult to pick up on improvements. The feedback has been very useful, though.

I tried v45 as a hard coded device name and didn't think it was an improvement.

[sbgk]

there was minor alignment error with v45, so v46 is another attempt at getting the alignment and instructions in the best order, maybe slightly more solid bass.

not that good, back to v45 and a smaller change, uploaded v47, sounds ok - call that the final one.

amazing how just moving the instruction order affects the sound, guess it's putting pressure for resources on the cpu in different ways.

v47 had a small bug, v48 sounds better.

Anyone using win 10 yet ? Other forums seem to be praising it's sq.