Page 653 of 804
Re: MQN
Posted: Thu Jan 01, 2015 7:06 pm
by jkeny
Thanks for reporting, Pierce - it's good to get further confirmation of the improvements of a RAMdisk OS
Similar improvements across the board, I take it? Jplay & MQN both improve although I hear what you are saying about MQN's low resource usage & the bigger uplift in sound with MQN & RAMdisk.
Re: MQN
Posted: Thu Jan 01, 2015 8:05 pm
by Sligolad
cvrle59 wrote:Pearse - one thing is not clear to me if read your post right. Have you managed ti install mqnload to RAMdisk, or you have to use traditional method with .bat program?
No problem installing MQNLoad to RamDisk, in fact I just copied my full MQN Folder with MQNLoad inside to the Compacted VHD File so it is all loaded to RAM.
The compacted VHD file is missing JPlay, AO and MQN files after compacting so I just install again on the compacted VHD and all goes OK.
Re: MQN
Posted: Thu Jan 01, 2015 10:09 pm
by sbgk
Hi JKeny, did you ever find out what jplay mean't by ultra low latency ram ?
the only think I can think of is large memory pages which enables contiguous ram to be allocated.
Re: MQN
Posted: Thu Jan 01, 2015 11:29 pm
by jkeny
sbgk wrote:Hi JKeny, did you ever find out what jplay mean't by ultra low latency ram ?
the only think I can think of is large memory pages which enables contiguous ram to be allocated.
That's what I also figured they meant but there could be more to it than that?
Re: MQN
Posted: Thu Jan 01, 2015 11:49 pm
by Octagon
Sligolad wrote:cvrle59 wrote:...to the Compacted VHD File so it is all loaded to RAM.
Hi Pearse,
I am not sure what you mean with "compacted VHD File"? But I understand you have the whole OS in RAM, not just MQn loaded onto a RAMDisc? Your comments show you got there :D great!
As mentioned before the whole R2 2012 will work without any compression if you build a 10GB VHD. That will leave 2GB space for other files like MQn and music files. ;) Boot this VHD with Grub4Dos into RAM which leaves 6GB RAM of a 16GB RAM availability for the RAM OS. You will be able to confirm what Pearse, myself and others are reporting for yourself.
What I see from a lot of reports trying to reduce the size, that is where the trouble starts. Mainly this is tried due to less availability of RAM. I do not to use the R2 possibility to reduce storage use as it harms again the performance as it adds additional compression/decompressing tasks. My lowest RAMOS VHD is 8GB and I have heard about versions where R2 has been stripped down to 2.9GB which allows 8GB RAM system to be used. But I am not concentrating on the lowest storage for the moment but sq. ;) In case you are interested in more details, I would be happy to share more of my learnings from trying to reduce the size of R2. But I didn't want to go to much off topic.
And yes, having tested MQN, BHE, JPlay and Foobar, MQn is still best in sq. Pearse, i can also confirm your measurements about activity in general.
I can only encourage everyone to try, it is a major improvement.
Just my 2cents
Thomas
Re: MQN
Posted: Fri Jan 02, 2015 12:00 am
by Sligolad
Hi Thomas, going to open a separate thread for this so as not to dilute the MQN thread, I will post more there. Cheers, Pearse.
Re: MQN
Posted: Fri Jan 02, 2015 9:22 am
by TioFrancotirdor
sbgk wrote:anyone tried the 3.94/9.07 versions ?
uploaded a 9.07 1 loop version which I think is an improvement, think it only works on cuinas devices though.
the advantage of 1 loop is less code and the function calls having optimal alignment.
wonder if it's an indication that if the driver was part of the player then the sq would be better.
think the interaction between user and kernel is part of the problem and if it was all in kernel that would better.
Hi Gordon,
I ended up eventually with 3.94/9.00 tf. Previously I used 3.87 avx2. I really did not expected mqn control giving much improvement untill I tried 3.94. 3.94 added more clarity in my setup. The difference was easily noticable. Seems like the way files are loaded into ram matters as well.
I would love to try 9.07 1 loop if you could make tf version.
Cheers,
Re: MQN
Posted: Fri Jan 02, 2015 4:54 pm
by sbgk
TioFrancotirdor wrote:sbgk wrote:anyone tried the 3.94/9.07 versions ?
uploaded a 9.07 1 loop version which I think is an improvement, think it only works on cuinas devices though.
the advantage of 1 loop is less code and the function calls having optimal alignment.
wonder if it's an indication that if the driver was part of the player then the sq would be better.
think the interaction between user and kernel is part of the problem and if it was all in kernel that would better.
Hi Gordon,
I ended up eventually with 3.94/9.00 tf. Previously I used 3.87 avx2. I really did not expected mqn control giving much improvement untill I tried 3.94. 3.94 added more clarity in my setup. The difference was easily noticable. Seems like the way files are loaded into ram matters as well.
I would love to try 9.07 1 loop if you could make tf version.
Cheers,
uploaded aa/tf/cuinas 9.08 versions, all 1 loop, works with cuinas, don't know about others. 3.5 kb, don't know if that's significant ie < 4 kb which is the page size, but sounds quite intimate.
Re: MQN
Posted: Fri Jan 02, 2015 5:02 pm
by sbgk
jkeny wrote:sbgk wrote:Hi JKeny, did you ever find out what jplay mean't by ultra low latency ram ?
the only think I can think of is large memory pages which enables contiguous ram to be allocated.
That's what I also figured they meant but there could be more to it than that?
yep, trying to get it working, may not be worth it.
Re: MQN
Posted: Fri Jan 02, 2015 5:05 pm
by sbgk
Sligolad wrote:Have not tried latest versions yet but below some feedback on recent versions in RamDisk OS Setup.
I got Server 2012 R2 Essentials setup and reduced to Minimal Server with AO1.26b and then everything Optimised in AO except Wasapi/MMCS and USB optimisations.
I had been listening mostly to Control 3.87avx2 and Play 1644 24 bit 9.01 avx2 gk launched from MQNLoad before I converted to RamDisk and the combination sounded excellent wit same OS setup listed above.
Moving this setup to RamDisk has produced an extraordinary improvement in the Music I have listened to so far with Control 3.87avx2 and Play 1644 24 bit 9.01 avx2 gk.
The added detail dynamics and musicality of playback is astounding, its almost like I have added a very expensive set of Ceramic Speakers and a whole new DAC and amplification system.
Today I looked at the resource monitor to see if there are any clues.
When MQNPlay launches a 1644 track I am seeing the following:
CPU usage at on average 0.10% most of which is used by the resource manager with only about 0.01% used by MQNPlay on Core 0 (that is one hundred of one percent!!!)
No Disk Activity once track is loaded.
No Network Activity.
I then went to JPlay and ran the same test:
CPU usage at on average 5.5% most of which is used by JPlay at about 3.8% used across changing cores 3, 4, 5.
Lots of Disk Activity during playback of first track which eventually settles down.
Continuous Network Activity even though I have Network disabled in Bios and no network connections whatsoever.
Bottom line is MQN sounds much better and all I can say is that you appear to be on the right track with recent developments.
Hope to try more recent versions over the coming days, it is a more convoluted process now to update the VHD and then load to RamDisk so hopefully it works OK.
At the moment I just cannot stop listening to music!!
Cheers, Pearse.
try the 3.94/9.08 versions, you'll have an excuse to listen to your music all over again.