MQN

[sbgk]

I like the sse4 intrinsic versions, the sse2 just sounds unnatural and a bit thin, might be able to improve on that.

Perhaps I change me monniker to Jack Sprat :)

the intrinsics are a whole new can of worms, will take a while to work out the best configuration, but initial trials seem to have produced some interesting results.

[sbgk]

Think its called avx2 on the intel spec sheet under instruction set

the 256 bit may or may not sound better, but haswell and windows 8.1/r2 also has hardware offloading, so if a device mfr enables that in the future you will be able to offload the sound processing to a hardware device, not sure if this would improve the sound, but the idea is that the cpu can be at a low power state and the music still streams.

thx.

haswell vs 3970X
one has AVX2
one has much more L1 L2 L3 cache.

3970X can have low power consumption. just downclock.

which CPU is more likely better for SQ?

i have used media player classic optimized for AVX. compared to VLC and others, motions are much much better on my low end graphic card. AVX can be good.

the only avx that MQn would use would be the 256 bit load and writes. Worth reading about haswell as it is more geared towards multimedia streaming, 512 bit coming in 2014/2015

[goon-heaven]

the intrinsics are a whole new can of worms, will take a while to work out the best configuration, but initial trials seem to have produced some interesting results.

The Intrinsics are Worms? ... Doh, I tought dey woz dem super aliens

[tony]

Take the morning off from komputers and when I return War Pigs is now the main test track and myriad new MQn releases.

Thanks to all re foobar/dbpoweramp got sorted and dbpoweramp works fine.

Just to add on Nigels comments on I3-7 and haswell. When doing the first build if you bother to go through the 1000 posts on Buildanaudiopc
there was agonising for ages over what processor to get. Most went for i7 simply just in case as the thinking wasn't clear at the time and the zuma build on computeradiophile used the i7. Sanity/experience suggests i5 just in case is now more than enough and if budget is tight i3 is fine. Has to be haswell again unless budget dictates otherewise. Haswell simply because it is newest technology and the advances outlined above.

Last time in my place no difference could be detected between zuma i7 sandybridge(I think) and Nige'ls haswell.Go figure.

Imho some of the stuff throughout the thread is splitting hairs. music in root,r2/core server/haswell/i3/i7. Maybe it is my ears but after a certain point of implementation you are doing stuff because it is recommended best practice.Often it costs virtually nothing to implement so why not. I fine it hard to distinguish some of the more subtle suggestions. It doesn't mean there not worthwhile and don't work.

On Davef's suggestion re blind tests the easiest way to try that is in a group situation where one person controls what version is used. The others can guess. If people are up to doing it that can be done relatively easy. Personally would just use that for a sanity check.The notion of spending valuable hobby/relaxation time conducting 'perfect' tests to convince a population of people who are too lazy to do it themselves and who often don't care about the result(i.e not looking for the info to help make up their minds) is futile. I know Dave is not in that category and his interest is genuine.

[DaveF]

On Davef's suggestion re blind tests the easiest way to try that is in a group situation where one person controls what version is used. The others can guess. If people are up to doing it that can be done relatively easy. Personally would just use that for a sanity check.The notion of spending valuable hobby/relaxation time conducting 'perfect' tests to convince a population of people who are too lazy to do it themselves and who often don't care about the result(i.e not looking for the info to help make up their minds) is futile. I know Dave is not in that category and his interest is genuine.

Exactly. Not looking for definitive proof or such but its a very worthwhile experiment. Time and time again it has been shown that the ears are easily fooled. If in this case there is enough evidence from the above tests that differences are correctly identified a high percentage of the time then it should harden our resolve to dig further into the technical side to find out a why. Once the mechanism is understood clearly, it can be further improved then. Right now it all seems rather hit and miss without a clear understanding of whats going on underneath.

[nige2000]

Take the morning off from komputers and when I return War Pigs is now the main test track and myriad new MQn releases.

Thanks to all re foobar/dbpoweramp got sorted and dbpoweramp works fine.

Just to add on Nigels comments on I3-7 and haswell. When doing the first build if you bother to go through the 1000 posts on Buildanaudiopc
there was agonising for ages over what processor to get. Most went for i7 simply just in case as the thinking wasn't clear at the time and the zuma build on computeradiophile used the i7. Sanity/experience suggests i5 just in case is now more than enough and if budget is tight i3 is fine. Has to be haswell again unless budget dictates otherewise. Haswell simply because it is newest technology and the advances outlined above.

Last time in my place no difference could be detected between zuma i7 sandybridge(I think) and Nige'ls haswell.Go figure.

Imho some of the stuff throughout the thread is splitting hairs. music in root,r2/core server/haswell/i3/i7. Maybe it is my ears but after a certain point of implementation you are doing stuff because it is recommended best practice.Often it costs virtually nothing to implement so why not. I fine it hard to distinguish some of the more subtle suggestions. It doesn't mean there not worthwhile and don't work.

On Davef's suggestion re blind tests the easiest way to try that is in a group situation where one person controls what version is used. The others can guess. If people are up to doing it that can be done relatively easy. Personally would just use that for a sanity check.The notion of spending valuable hobby/relaxation time conducting 'perfect' tests to convince a population of people who are too lazy to do it themselves and who often don't care about the result(i.e not looking for the info to help make up their minds) is futile. I know Dave is not in that category and his interest is genuine.

just to over complicate things
as usual
Last time in tonys there was some other differences that will have had an effect too
I was running r2 without any script
my ram at that time was unable to work at 800mhz like tonys and i was at a determental 1066mhz
and had not yet figured out that running the haswell cpu at 800mhz was optimal

im ready for my rematch

[sbgk]

Page 67 of 67 [ 666 posts ] Go to page Previous 1 ... 63, 64, 65, 66, 67


Nige just got the 666th post, is that a sign. Maybe Iron Maiden - Number of the beast should be the test track.

[jesuscheung]

just to over complicate things
as usual
Last time in tonys there was some other differences that will have had an effect too
I was running r2 without any script
my ram at that time was unable to work at 800mhz like tonys and i was at a determental 1066mhz
and had not yet figured out that running the haswell cpu at 800mhz was optimal

im ready for my rematch

dont forget CPU voltage affects SQ. you need more voltage than minimum for best SQ.

[jkeny]

Exactly. Not looking for definitive proof or such but its a very worthwhile experiment. Time and time again it has been shown that the ears are easily fooled. If in this case there is enough evidence from the above tests that differences are correctly identified a high percentage of the time then it should harden our resolve to dig further into the technical side to find out a why. Once the mechanism is understood clearly, it can be further improved then. Right now it all seems rather hit and miss without a clear understanding of whats going on underneath.

DaveF, I've no problem with doing a blind test but you do realise MQN was already tested blind, twice, by Clive - once inadvertently, by himself & another time in an arranged blind test where he had a number of listeners who were mainly disinterested in the outcome & all seemed to identify differences & agree on which MQN version was best.

Maybe Clive can say more about these blind tests & fill in some of the details as they seem to have been glossed over or missed?

Edit: I'm not sure if further discussion of this belongs in the MQN measurements thread

[DaveF]

also, turning your pc sideways will give you a much wider soundstage