Popular Post Schafheide Posted November 3, 2022 Popular Post Share Posted November 3, 2022 FWIW. UltraRendu now has NAA 4.2.5 available. Mike Rubin, semente and MikePid 3 Link to comment
Popular Post IgorSki Posted November 3, 2022 Popular Post Share Posted November 3, 2022 In this fairly long post I will share in details my experience on how to get stable and reliable HQPlayer performance at SDM @ 1024 with ASDM7ECv2 modulator and some (not all) filters of choice. Weather anyone (and why) would need DSD @ 1024 is another question... Sometime, not so long ago, on this forum in "best CPU for hqplayer" I was very excited to share that DSD 1024 with ASDM7ECv2 was very much a reality on i9-12900K server without any graphics card (no CUDA, that is) in combination with classical HQP NAA architecture. While the results were obvious, and I was not the first one to say ASDM7ECv2 @ 1024 is just possible - my system was apparently cooking at the top edge of its capacity, being highly sensitive to micro variations of al types of settings both in HQP (like length of buffer for example, or even the logging) and BIOS (like thermal control features or such...). And dispite long hours of play this was coming unfortunately with occasional "deadly" crashes (not really frequent, but utterly annoying, since CMOS reset would normally be required) And because of that my 12900K build did never make it to the cellar, away from listening room, but... But on the other hand the fact that ASDM7ECv2 @ 1024 is doable, made me think (may be logically) that "KS" model of CPU with its extra headroom would bring the machine down to more stable state. And I was right!!! To have stable ASDM7ECv2 @ 1024 the i9 - 12900KS chip needs to be overclocked (certainly), but finding a sweet spot for stable performance is easier than for 12900K. No real need in core by core dialing (although possible) and therefore not a factor of "silicon lottery" as overclockers call it. As a result - after reasonable time of experimenting, the 129KS server made it to the cellar room and stayed there for a good month or so. This was until the latest HQP 4 Embedded release 4.33.0 and to be more precise - hqplayerd_4.33.0-145avx2_amd64 has disturbed an equilibrium. This event in turn, gave me an opportunity to recheck all my settings again in more structured manner and now "fresh from the press" I am happy to share with you the details of my base-line config for stable ASDM7ECv2 @ 1024 performance (obviously 5ECv2@1024 works even better, and PCM 1,5MHz is a light breeze without saying, and I'm not discussing SDM@512 which plays smooth as silk and leaves CPU temps in its 40s) My architecture in brief: Source - ROON ROCK (local library and Qobuz) on 10th gen NUC (classics) - (ethernet) HQP server 12900KS (ethernet, of course) NAA - 8th gen fanless NUC connected via USB to DAC Holo Audio May KTE -> Serene -> AHB2 -> ProAc D20R Software: OS: Ubuntu Server 22.04 LTS - normal installation, with all latest updates Kernel: 5.15.0-52-lowlatency - note: I have also tested with custom kernels 52 and 71 from Jussi, these were not friendly with KS chip pushed to its limits and would occasionally drop out, ordinary lowlatency kernel seemed to do the job well. At the time of the post I see kernel 76 jl+, I did not try this one. But equally, I understand that highest value for Jussi's kernels is in set up with direct USB connection to DAC, that is not the case here. libgmpris_2.2.1-10_amd64.deb - prerequisite hqplayerd_4.32.4-141avx2_amd64.deb - I would call this a Golden DSD 1024 release, as it performs totally stable in this set up for SDM 1024 purpose. The version 143avx2 is good, but would drop out time to time on SDM sources or Nx: hirez-mp filter for example. Release 145avx2 would hick up more frequently and always bring system to a frozen crash state after 30 to 90 min of play - playing with BIOS would resolve drop outs, but would leave system generally unstable. NB: very important is the setting in HQP xml config file multicore="1", the default is "auto" - it would shutter on my system. For system monitoring I use couple of known utilities: htop - visual CPU load by threads, frequencies, temps (screen 12h+ run of HQP in action) i7z - CPU load by cores, frequencies, temps, voltages (secure boot needs to be disabled in BIOS to run this one) Hardware: CPU: i9-12900KS MB: MSI Pro Z690-A (DDR5) NB: Microcode or BIOS version is important! They seem to perform differently and for my build I am using the one released in May 2022, it knows KS chip well, but is not yet touched with future architectures like 13th gen. COOLING: Arctic Liquid Freezer II 360 - absolute winner! MEM: Kingston DDR5-RAM FURY Beast 6000 MHz 2x8 GB SSD: Samsung SSD M2 250G POWER: EVGA SuperNOVA 850 G3 80 Plus Gold 850W CASE: Fractal Define R5 The cost of this server from all new components in Switzerland (dont forget the secondary market exists) at the time of post is just nothing but CHF 1450,- BIOS/Overclocking. Is done in two very simple steps and nothing more is required to keep it going ! NB: for previous 12900K build I was recomending to adjust hardware fan/pump curve - but for this build please leave it at default, otherwise HQP drops out. BIOS bug or feature - I don't know, and don't care as temps are reasonable. Step 1. Flash BIOS, Set XMP Profile - 1 (fastest), turn off the following: hd audio controller, ftpm 2.0, ez led control, disable secure boot. Reboot Step 2. Light overclocking summary is on the picture: Comments: ring ratio should not be higher than CPU frequency, as such having it aligned with e-cores at 44 keeps system stable. CPU core voltage mode - adaptive mode - would preserve from overheating. Cpu core voltage 1,45v - is normal for 12th gen. Voltages will be fluctuating from 1,2 up to 1,49... The following will cause hick ups if not disabled: Intel adaptive thermal monitor Intel speed shift technology EIST TVB - all tvb settings needs to be disabled HQPlayer Set-up. The picture is self explanatory, but few comments. The upsampling goes to SDM1024 sharp - no x48 ! Unfortunately 7ECv2@1024x48 drops out. But ordinary EC like 7EC 512+ can do 1024x48, if you wish. Logging is disabled - last time I tried it was causing drop outs, apparently I/O pressure was too much. Buffer is 250, the maximum, it never hicks up with this setting - you may experiment of course. Testing criteria Overclocking (OC) for HQPlayer blues. This comes in several iterations (its f.. pain!) 1st Circle: trying to establish stable playback for a composition, normally i take 192/24 source at poly-sinc-short-mp filter, important to have more than 4 min playback in testing, once HQP passes 4 min line of playback, there's high chance you found some type of OC base-line to build upon. This is most tedious circle requires a lot of (we talking dozens) system restarts... 2nd Circle: exploring the settings with different filters and sources such as always gauss-xla, ext3 or hirez-mp and DSD64, often here is a disappointing moment when we find that base line is not fit for all. Back to circle 1. 3rd Circle: "Album" runs - when 2nd circle is a success, test the album length runs... unstable OC system would crash in 30/40/60/90 mins eventually, some times you'll start getting drop outs at 30 min, like thermals are not good, etc... 4th Circle: once we can see that album length is stable, run a marathon tests of 12h+ mixed playbacks. Importantly after 12h+ playback play with filters and sources to check system stability. Also important are mixed playbacks - unstable OC system hangs on source changes. 5th Circle: Once we are happy with a long runs (that is may be on day 3 or 4 or 5...) set system at its permanent place in cellar away from screens and listening area, and after a couple of days of regular "usual" use, if it still works - we may say - it's stable... Results: Voila! We have a system that runs hours and hours of smooth and uninterrupted DSD@1024 playback through ASDM7ECv2 modulator for wide range of sources like PCM 44,1/16 to 192/24 and DSD64/128... This comes at reasonable power cost as well - only 240-250W at full power (DSD@512 is cheaper in this regard). Gradually stepping in turmoil of European winter energy crisis this consideration may become extremely important. The system can do most of the filters but not all. Also - i did not try it with convolution but I highly suspect it will not manage. The simplest "bit perfect" speaker position adjustment I have to do in ROON for 7ECv2@1024 playback as at these settings HQP hicks up. Here are the filters I've tested: 1x: poly-sinc-short-mp, gauss-long, gauss-xla, ext3, pretty much all "-2s" ones Nx: hires-lp/ip/mp, poly-sinc-short-mp It can not do Nx: ext3 or gauss-xla It can not do 1x/Nx sinc-l or sinc-Ll (or orher sinc-xx) at this rate. And these filters are my favorite combos in this set up: 1x: gauss-xla / Nx: hirez-lp 1x: poly-sinc-short-mp / Nx: poly-sinc-short-mp -> listening to prog rock on this is totally insane, you would never ever want to stop... "-2s" filters also sounds very decent. Annoying is initialization times for some filters, such as poly-sinc-short-mp for 44,1 takes up to 4-5min to initialise (but only once), long-mp would probably take hours. Although for 96 or 192/24 short-mp initialization is almost immediate. DSD sources go via FIR2 / XFi - "creme of a creme" settings. 1024? What's in it? I follow the DSD 1024 discussions here with great interest. To myself I have a lucky chance to actually be able ro reproduce and hear how it sounds at one of best filter and modulator combinations. Without going to deep into unmeasurable subjectivism I mention only two elements that definitely stand out - the size of sound stage is inflated, separation or imaging inside this stage is amazingly sharp. Very good question - would I capture this in blind difference with DSD 512? I want to know as well. But overall experience is euphonic, that's certain. The source quality matters a lot, good mastering blows away, but frankly it would be true to other systems. But spatial or dimensional effect @1024 is there! I'd say I have a quality concert hall sound with just enough reverb sitting beside the stage with my eyes closed... but it's just me... You are welcome to try! pavi, marcosax, k6davis and 3 others 2 2 2 Link to comment
Miska Posted November 3, 2022 Share Posted November 3, 2022 22 minutes ago, IgorSki said: Release 145avx2 would hick up more frequently and always bring system to a frozen crash state after 30 to 90 min of play This is related to something else than HQPlayer, and would indicate a likely hardware problem somewhere. 22 minutes ago, IgorSki said: Buffer is 250, the maximum, it never hicks up with this setting - you may experiment of course. I would recommend to stick to 100 ms, it is likely more stable. Signalyst - Developer of HQPlayer Pulse & Fidelity - Software Defined Amplifiers Link to comment
IgorSki Posted November 3, 2022 Share Posted November 3, 2022 4 minutes ago, Miska said: 30 minutes ago, IgorSki said: Release 145avx2 would hick up more frequently and always bring system to a frozen crash state after 30 to 90 min of play This is related to something else than HQPlayer, and would indicate a likely hardware problem somewhere. @Miska thank you for reply. I have compared 141avx, 143avx and 145avx under same conditions. That is not changing anything at BIOS or HQPlayer settings, only changing the releases. Base-line were settings where 141avx performs just fine. 143avx would play forewer as well, but would occasionally drop. But 145avx would start hicking up after like 30-40 min, especially on SDM sources and than it all ends in crash. Setting “nblocks” in “engine” element to “8” to force past behavior has no effect. But I speak only 7ECv2@1024 which is apparently on the edje, Jussi, SDM 512 works just fine... of course. I will check out buffer of 100... now with the system performing at its best really afraid to tweak :) hehe. I wish you can hear how it sounds! Link to comment
Miska Posted November 3, 2022 Share Posted November 3, 2022 3 minutes ago, IgorSki said: But 145avx would start hicking up after like 30-40 min, especially on SDM sources and than it all ends in crash. Sounds like you are hitting either thermal limits, or you get RAM errors. Have you left memtest86 running for 24 hours with the BIOS settings you are using? 3 minutes ago, IgorSki said: Setting “nblocks” in “engine” element to “8” to force past behavior has no effect. OK, I guess it remains a mystery then where the difference is coming from. Signalyst - Developer of HQPlayer Pulse & Fidelity - Software Defined Amplifiers Link to comment
ted_b Posted November 3, 2022 Share Posted November 3, 2022 @IgorSkiand @Miska Do you think the above DSD1024 server would work in Windows, as well? My NAA continues to be a Fitlet2 in my media room, but server would remain upstairs in my office. I ask cuz today (after working on some unrelated electrical issues and needing to shut service down while installing a new GFI plug in the garage) my server (WIn10, I7-6700k) is failing to power up. I haven't done a deep dive, but this thing is getting old anyway. "We're all bozos on this bus"....F.T. My JRIver tutorial videos Actual JRIver tutorial MP4 video links My eleven yr old SACD Ripping Guide for PS3 (needs updating but still works) US Technical Advisor, NativeDSD.com Link to comment
Miska Posted November 4, 2022 Share Posted November 4, 2022 1 minute ago, ted_b said: @IgorSkiand @Miska Do you think the above DSD1024 server would work in Windows, as well? My NAA continues to be a Fitlet2 in my media room, but server would remain upstairs in my office. I ask cuz today (after working on some unrelated electrical issues and needing to shut service down while installing a new GFI plug in the garage) my server (WIn10, I7-6700k) is failing to power up. I haven't done a deep dive, but this thing is getting old anyway. I think eventually it will become pretty straightforward. Maybe with i9-13900K + RTX4090 or Ryzen 9 7950X + RTX4090 already. I have not seen any results with these machines and latest generation Nvidia GPUs yet. Signalyst - Developer of HQPlayer Pulse & Fidelity - Software Defined Amplifiers Link to comment
IgorSki Posted November 4, 2022 Share Posted November 4, 2022 6 minutes ago, Miska said: 12 minutes ago, IgorSki said: But 145avx would start hicking up after like 30-40 min, especially on SDM sources and than it all ends in crash. Sounds like you are hitting either thermal limits, or you get RAM errors. Have you left memtest86 running for 24 hours with the BIOS settings you are using? Good point on RAM, I need to try the memtest86. I can imagine it's not obvious since 141avx is just fine... like totally fine. As for the thermals, visually nothing critical. 145avx seems to do the same stuff, but with adaptive voltage the range of temps is like from 30 to 80 so really difficult to say where it may misbehave. Link to comment
Popular Post Fredc Posted November 4, 2022 Popular Post Share Posted November 4, 2022 54 minutes ago, IgorSki said: In this fairly long post I will share in details my experience on how to get stable and reliable HQPlayer performance at SDM @ 1024 with ASDM7ECv2 modulator and some (not all) filters of choice. Weather anyone (and why) would need DSD @ 1024 is another question... Sometime, not so long ago, on this forum in "best CPU for hqplayer" I was very excited to share that DSD 1024 with ASDM7ECv2 was very much a reality on i9-12900K server without any graphics card (no CUDA, that is) in combination with classical HQP NAA architecture. While the results were obvious, and I was not the first one to say ASDM7ECv2 @ 1024 is just possible - my system was apparently cooking at the top edge of its capacity, being highly sensitive to micro variations of al types of settings both in HQP (like length of buffer for example, or even the logging) and BIOS (like thermal control features or such...). And dispite long hours of play this was coming unfortunately with occasional "deadly" crashes (not really frequent, but utterly annoying, since CMOS reset would normally be required) And because of that my 12900K build did never make it to the cellar, away from listening room, but... But on the other hand the fact that ASDM7ECv2 @ 1024 is doable, made me think (may be logically) that "KS" model of CPU with its extra headroom would bring the machine down to more stable state. And I was right!!! To have stable ASDM7ECv2 @ 1024 the i9 - 12900KS chip needs to be overclocked (certainly), but finding a sweet spot for stable performance is easier than for 12900K. No real need in core by core dialing (although possible) and therefore not a factor of "silicon lottery" as overclockers call it. As a result - after reasonable time of experimenting, the 129KS server made it to the cellar room and stayed there for a good month or so. This was until the latest HQP 4 Embedded release 4.33.0 and to be more precise - hqplayerd_4.33.0-145avx2_amd64 has disturbed an equilibrium. This event in turn, gave me an opportunity to recheck all my settings again in more structured manner and now "fresh from the press" I am happy to share with you the details of my base-line config for stable ASDM7ECv2 @ 1024 performance (obviously 5ECv2@1024 works even better, and PCM 1,5MHz is a light breeze without saying, and I'm not discussing SDM@512 which plays smooth as silk and leaves CPU temps in its 40s) My architecture in brief: Source - ROON ROCK (local library and Qobuz) on 10th gen NUC (classics) - (ethernet) HQP server 12900KS (ethernet, of course) NAA - 8th gen fanless NUC connected via USB to DAC Holo Audio May KTE -> Serene -> AHB2 -> ProAc D20R Software: OS: Ubuntu Server 22.04 LTS - normal installation, with all latest updates Kernel: 5.15.0-52-lowlatency - note: I have also tested with custom kernels 52 and 71 from Jussi, these were not friendly with KS chip pushed to its limits and would occasionally drop out, ordinary lowlatency kernel seemed to do the job well. At the time of the post I see kernel 76 jl+, I did not try this one. But equally, I understand that highest value for Jussi's kernels is in set up with direct USB connection to DAC, that is not the case here. libgmpris_2.2.1-10_amd64.deb - prerequisite hqplayerd_4.32.4-141avx2_amd64.deb - I would call this a Golden DSD 1024 release, as it performs totally stable in this set up for SDM 1024 purpose. The version 143avx2 is good, but would drop out time to time on SDM sources or Nx: hirez-mp filter for example. Release 145avx2 would hick up more frequently and always bring system to a frozen crash state after 30 to 90 min of play - playing with BIOS would resolve drop outs, but would leave system generally unstable. NB: very important is the setting in HQP xml config file multicore="1", the default is "auto" - it would shutter on my system. For system monitoring I use couple of known utilities: htop - visual CPU load by threads, frequencies, temps (screen 12h+ run of HQP in action) i7z - CPU load by cores, frequencies, temps, voltages (secure boot needs to be disabled in BIOS to run this one) Hardware: CPU: i9-12900KS MB: MSI Pro Z690-A (DDR5) NB: Microcode or BIOS version is important! They seem to perform differently and for my build I am using the one released in May 2022, it knows KS chip well, but is not yet touched with future architectures like 13th gen. COOLING: Arctic Liquid Freezer II 360 - absolute winner! MEM: Kingston DDR5-RAM FURY Beast 6000 MHz 2x8 GB SSD: Samsung SSD M2 250G POWER: EVGA SuperNOVA 850 G3 80 Plus Gold 850W CASE: Fractal Define R5 The cost of this server from all new components in Switzerland (dont forget the secondary market exists) at the time of post is just nothing but CHF 1450,- BIOS/Overclocking. Is done in two very simple steps and nothing more is required to keep it going ! NB: for previous 12900K build I was recomending to adjust hardware fan/pump curve - but for this build please leave it at default, otherwise HQP drops out. BIOS bug or feature - I don't know, and don't care as temps are reasonable. Step 1. Flash BIOS, Set XMP Profile - 1 (fastest), turn off the following: hd audio controller, ftpm 2.0, ez led control, disable secure boot. Reboot Step 2. Light overclocking summary is on the picture: Comments: ring ratio should not be higher than CPU frequency, as such having it aligned with e-cores at 44 keeps system stable. CPU core voltage mode - adaptive mode - would preserve from overheating. Cpu core voltage 1,45v - is normal for 12th gen. Voltages will be fluctuating from 1,2 up to 1,49... The following will cause hick ups if not disabled: Intel adaptive thermal monitor Intel speed shift technology EIST TVB - all tvb settings needs to be disabled HQPlayer Set-up. The picture is self explanatory, but few comments. The upsampling goes to SDM1024 sharp - no x48 ! Unfortunately 7ECv2@1024x48 drops out. But ordinary EC like 7EC 512+ can do 1024x48, if you wish. Logging is disabled - last time I tried it was causing drop outs, apparently I/O pressure was too much. Buffer is 250, the maximum, it never hicks up with this setting - you may experiment of course. Testing criteria Overclocking (OC) for HQPlayer blues. This comes in several iterations (its f.. pain!) 1st Circle: trying to establish stable playback for a composition, normally i take 192/24 source at poly-sinc-short-mp filter, important to have more than 4 min playback in testing, once HQP passes 4 min line of playback, there's high chance you found some type of OC base-line to build upon. This is most tedious circle requires a lot of (we talking dozens) system restarts... 2nd Circle: exploring the settings with different filters and sources such as always gauss-xla, ext3 or hirez-mp and DSD64, often here is a disappointing moment when we find that base line is not fit for all. Back to circle 1. 3rd Circle: "Album" runs - when 2nd circle is a success, test the album length runs... unstable OC system would crash in 30/40/60/90 mins eventually, some times you'll start getting drop outs at 30 min, like thermals are not good, etc... 4th Circle: once we can see that album length is stable, run a marathon tests of 12h+ mixed playbacks. Importantly after 12h+ playback play with filters and sources to check system stability. Also important are mixed playbacks - unstable OC system hangs on source changes. 5th Circle: Once we are happy with a long runs (that is may be on day 3 or 4 or 5...) set system at its permanent place in cellar away from screens and listening area, and after a couple of days of regular "usual" use, if it still works - we may say - it's stable... Results: Voila! We have a system that runs hours and hours of smooth and uninterrupted DSD@1024 playback through ASDM7ECv2 modulator for wide range of sources like PCM 44,1/16 to 192/24 and DSD64/128... This comes at reasonable power cost as well - only 240-250W at full power (DSD@512 is cheaper in this regard). Gradually stepping in turmoil of European winter energy crisis this consideration may become extremely important. The system can do most of the filters but not all. Also - i did not try it with convolution but I highly suspect it will not manage. The simplest "bit perfect" speaker position adjustment I have to do in ROON for 7ECv2@1024 playback as at these settings HQP hicks up. Here are the filters I've tested: 1x: poly-sinc-short-mp, gauss-long, gauss-xla, ext3, pretty much all "-2s" ones Nx: hires-lp/ip/mp, poly-sinc-short-mp It can not do Nx: ext3 or gauss-xla It can not do 1x/Nx sinc-l or sinc-Ll (or orher sinc-xx) at this rate. And these filters are my favorite combos in this set up: 1x: gauss-xla / Nx: hirez-lp 1x: poly-sinc-short-mp / Nx: poly-sinc-short-mp -> listening to prog rock on this is totally insane, you would never ever want to stop... "-2s" filters also sounds very decent. Annoying is initialization times for some filters, such as poly-sinc-short-mp for 44,1 takes up to 4-5min to initialise (but only once), long-mp would probably take hours. Although for 96 or 192/24 short-mp initialization is almost immediate. DSD sources go via FIR2 / XFi - "creme of a creme" settings. 1024? What's in it? I follow the DSD 1024 discussions here with great interest. To myself I have a lucky chance to actually be able ro reproduce and hear how it sounds at one of best filter and modulator combinations. Without going to deep into unmeasurable subjectivism I mention only two elements that definitely stand out - the size of sound stage is inflated, separation or imaging inside this stage is amazingly sharp. Very good question - would I capture this in blind difference with DSD 512? I want to know as well. But overall experience is euphonic, that's certain. The source quality matters a lot, good mastering blows away, but frankly it would be true to other systems. But spatial or dimensional effect @1024 is there! I'd say I have a quality concert hall sound with just enough reverb sitting beside the stage with my eyes closed... but it's just me... You are welcome to try! You mentioned "Very good question - would I capture this in blind difference with DSD 512?". I guess that's the $64K question. May be not $64K, but I would have to upgrade to a 1024DSD Dac and beef up my PC to your specification. Fun to try I guess but unless there is a clear upgrade in sound from DSD512, it's difficult to justify for me. With my current setup, I can play DSD512 very smoothly with 7ECV2 and just as about all filters while having a silent PC setup. k6davis and chipvn 2 Link to comment
IgorSki Posted November 4, 2022 Share Posted November 4, 2022 9 minutes ago, ted_b said: @IgorSkiand @Miska Do you think the above DSD1024 server would work in Windows, as well? My NAA continues to be a Fitlet2 in my media room, but server would remain upstairs in my office. I ask cuz today (after working on some unrelated electrical issues and needing to shut service down while installing a new GFI plug in the garage) my server (WIn10, I7-6700k) is failing to power up. I haven't done a deep dive, but this thing is getting old anyway. @ted_b that would need testing if objective is 7ECv2@1024. May be now 139K would be a better choice in this regard. My hands are itching to get 13K but I know it needs healthy couple of month for hardware and software updates to properly recognise it. If I look at 12900KS dynamics it took couple of month for my MSI to update the microcode for my MB for its full potential... Link to comment
IgorSki Posted November 4, 2022 Share Posted November 4, 2022 1 hour ago, Miska said: 1 hour ago, IgorSki said: Setting “nblocks” in “engine” element to “8” to force past behavior has no effect. OK, I guess it remains a mystery then where the difference is coming from @Miska, Its weird, agree. I do now another "clean cut" test for 145avx... No changes to BIOS, no changes to OS environment (no vodka, hands not shaking... :) - 145avx installed - xml config updated with nblocks="8", saved - reboot nothing more, I leave it running for a night with mixed playlist, starting with 192/24 sources Link to comment
Miska Posted November 4, 2022 Share Posted November 4, 2022 10 minutes ago, IgorSki said: - xml config updated with nblocks="8", saved You can safely leave it there. Previous versions don't understand it and just ignore it. Signalyst - Developer of HQPlayer Pulse & Fidelity - Software Defined Amplifiers Link to comment
Kalpesh Posted November 4, 2022 Share Posted November 4, 2022 On 11/2/2022 at 1:35 PM, Kalpesh said: I have been experiencing with web radios kind of stutter (not of the overload type) with mp3 flux and worked quite well with 24/96 HD radios, so it definitely does not look like cpu bandwidth ram buffer what have you limitation does not work at all with aac while some radios of major interest (I dare not to add afaic) such as France Musique which offers a broad choice of direct and recorded live classical and jazz performances use that tech. I understand HQP does not support aac beyond wishful thinking it would, is there a way to direct aac to HQP desktop Mac even if a third party app is required?? Looked into HQPD control with my iPhone but no luck up/short : is there a way to listen to aac web radios ? Link to comment
Argon Posted November 4, 2022 Share Posted November 4, 2022 I am trying to track down a random repeating problem streaming from a Roon/HQPlayer Embedded (SonicTransport I9) to a Sonore Signature SE NAA endpoint. Sometimes the music just stops, no rhyme or reason, could be after 40 min or 4 hours... Roon still shows it's "playing" but the progress bar doesn't move. Selecting a different track, same result. Playback will just not start. There is nothing I can do short of power cycling the NAA endpoint, Roon/HQPlayer computer and PS Audio DSD to solve the problem. What do these "Empty Transport" or "Push to FIFO Failed" indicate? 2022/11/04 05:24:36 Stop... & 2022/11/04 05:24:37 Playlist clear # 2022/11/04 05:24:37 clControlThread::ParseMsg(): Invalid authentication tag: ChaCha20Poly1305 tag check failed & 2022/11/04 05:24:37 Play # 2022/11/04 05:24:37 clControlThread::ParseMsg(): clHQPlayerEngine::Play(): Empty transport 2022/11/04 06:07:50 Initialization complete, starting audio engine 2022/11/04 06:07:50 Teams: 1 2022/11/04 06:07:50 Places: 1 2022/11/04 06:07:50 Parallel threads: 8 2022/11/04 06:07:50 Nested parallelism: 4 2022/11/04 06:07:50 Parallel pipelines: 4 ! 2022/11/04 06:07:50 NAA output clNetEngine::PushSDM(): not connected to adapter ! 2022/11/04 06:07:50 clHQPlayerEngine::Execute(): push to FIFO failed 2022/11/04 06:07:50 Stop request (reset) & 2022/11/04 06:07:50 Stop... - 2022/11/04 06:07:50 Playback engine stopped & 2022/11/04 06:07:50 ...stopped Link to comment
frieddr Posted November 4, 2022 Share Posted November 4, 2022 Help. I'm trying really hard to find settings on HQPlayer that will run upscaling fairly stable and that does not cause popping and scratchy distortion. I'm sure my settings must be wrong. I am streaming from SMS200-Ultra to Chord Qutest. If anyone can look at my settings and tell me what's wrong, I would be very grateful. Link to comment
musicjunkie917 Posted November 4, 2022 Share Posted November 4, 2022 If this is with DSD, uncheck 48K DSD as the Chord DACs don't support that if I recall correctly. Link to comment
IgorSki Posted November 4, 2022 Share Posted November 4, 2022 16 hours ago, IgorSki said: 18 hours ago, Miska said: 18 hours ago, IgorSki said: Setting “nblocks” in “engine” element to “8” to force past behavior has no effect. OK, I guess it remains a mystery then where the difference is coming from Expand @Miska, Its weird, agree. I do now another "clean cut" test for 145avx... No changes to BIOS, no changes to OS environment (no vodka, hands not shaking... :) - 145avx installed - xml config updated with nblocks="8", saved - reboot nothing more, I leave it running for a night with mixed playlist, starting with 192/24 sources @Miska now it works correctly ! must have been my spelling error or smth... Release - hqplayerd_4.33.0-145avx2_amd64.deb is stable as a brick and runs now non stop for 12+ hours with mixed playback on 7ECv2 @ SDM 1024 / 1x: ext3 / Nx: hires-mp and FIR2/Xfi for SDM with exact BIOS OC settings mentioned in my post above. No issues. I say - we have HQPe 7ECv2 @ 1024 golden standard from Jussi ! Important manual configuration in hqplayerd.xml configuration file for <engine> : - multicore="1" - (manually add) nblocks="8" 19 hours ago, Miska said: 19 hours ago, IgorSki said: Buffer is 250, the maximum, it never hicks up with this setting - you may experiment of course. I would recommend to stick to 100 ms, it is likely more stable. The buffer was set to 100 - runs smoothly. PS: It is Friday night and it feels like Friday night! I cant be more happy with my recent discoveries and insanely addictive SDM@1024 with ECv2 modulators, euphonic become euphoric. @Miska I raise a glass for you tonight, thank you for fantastic work you do !!! Miska 1 Link to comment
El Guapo Posted November 4, 2022 Share Posted November 4, 2022 9 minutes ago, IgorSki said: PS: It is Friday night and it feels like Friday night! I cant be more happy with my recent discoveries and insanely addictive SDM@1024 with ECv2 modulators, euphonic become euphoric. @Miska I raise a glass for you tonight, thank you for fantastic work you do !!! Congratulations for both of you guys! IgorSki 1 Link to comment
frieddr Posted November 4, 2022 Share Posted November 4, 2022 10 hours ago, Argon said: I am trying to track down a random repeating problem streaming from a Roon/HQPlayer Embedded (SonicTransport I9) to a Sonore Signature SE NAA endpoint. Sometimes the music just stops, no rhyme or reason, could be after 40 min or 4 hours... Roon still shows it's "playing" but the progress bar doesn't move. Selecting a different track, same result. Playback will just not start. There is nothing I can do short of power cycling the NAA endpoint, Roon/HQPlayer computer and PS Audio DSD to solve the problem. What do these "Empty Transport" or "Push to FIFO Failed" indicate? 2022/11/04 05:24:36 Stop... & 2022/11/04 05:24:37 Playlist clear # 2022/11/04 05:24:37 clControlThread::ParseMsg(): Invalid authentication tag: ChaCha20Poly1305 tag check failed & 2022/11/04 05:24:37 Play # 2022/11/04 05:24:37 clControlThread::ParseMsg(): clHQPlayerEngine::Play(): Empty transport 2022/11/04 06:07:50 Initialization complete, starting audio engine 2022/11/04 06:07:50 Teams: 1 2022/11/04 06:07:50 Places: 1 2022/11/04 06:07:50 Parallel threads: 8 2022/11/04 06:07:50 Nested parallelism: 4 2022/11/04 06:07:50 Parallel pipelines: 4 ! 2022/11/04 06:07:50 NAA output clNetEngine::PushSDM(): not connected to adapter ! 2022/11/04 06:07:50 clHQPlayerEngine::Execute(): push to FIFO failed 2022/11/04 06:07:50 Stop request (reset) & 2022/11/04 06:07:50 Stop... - 2022/11/04 06:07:50 Playback engine stopped & 2022/11/04 06:07:50 ...stopped Same thing was happening to me. Also running imbedded Hqplayer but on a sonicorbiter5. I have significantly lowered my upsampling to 4x. Obviously my server is more limited in computational power. I think I get most of the sonic benefits I'm looking for at 4x. Link to comment
sworksone Posted November 6, 2022 Share Posted November 6, 2022 Is there any technical reason not to use AMSDM7EC 512+fs at DSD256 (Holo May DAC) or is it truly suitable for DSD512 and above only? Link to comment
Miska Posted November 6, 2022 Share Posted November 6, 2022 23 minutes ago, sworksone said: Is there any technical reason not to use AMSDM7EC 512+fs at DSD256 (Holo May DAC) or is it truly suitable for DSD512 and above only? Recommendation is to use it only at DSD512 and higher speeds. But I have not prohibited it's use at DSD256. It may be OK'ish at DSD256 on Holo and T+A. But OTOH, I don't see a reason to use it instead of ASDM7ECv2. Signalyst - Developer of HQPlayer Pulse & Fidelity - Software Defined Amplifiers Link to comment
sworksone Posted November 6, 2022 Share Posted November 6, 2022 Thank you. In very limited listening this afternoon vocals seemed more holographic vs with ASDM7ECV2 but I’d have to do additional back and forth to confirm. I was curious if I was just hearing more distortion or similar. Link to comment
Miska Posted November 6, 2022 Share Posted November 6, 2022 25 minutes ago, sworksone said: Thank you. In very limited listening this afternoon vocals seemed more holographic vs with ASDM7ECV2 but I’d have to do additional back and forth to confirm. I was curious if I was just hearing more distortion or similar. If you like it more, it should be pretty OK with your DAC. But especially if you like it, I'd recommend to try to run it at DSD512 or DSD1024 instead. Of course it demands 2x or 4x more processing power than at DSD256 too. Signalyst - Developer of HQPlayer Pulse & Fidelity - Software Defined Amplifiers Link to comment
dericchan1 Posted November 7, 2022 Share Posted November 7, 2022 @Miska Hi Miska, in my new listening room, I have created the following PEQ room correction filter Filter 4: ON PK Fc 550.00 Hz Gain +3.00 dB Q 3.0 Which is a gain boost at 550 Hz by 3.00 db. Can you please advise if there is anything I need to watch for in my settings to prevent clipping? Thanks Deric Link to comment
Miska Posted November 7, 2022 Share Posted November 7, 2022 6 hours ago, dericchan1 said: Filter 4: ON PK Fc 550.00 Hz Gain +3.00 dB Q 3.0 Which is a gain boost at 550 Hz by 3.00 db. Can you please advise if there is anything I need to watch for in my settings to prevent clipping? Use -3 dB mix gain on the matrix pipelines and you are fine. dericchan1 1 Signalyst - Developer of HQPlayer Pulse & Fidelity - Software Defined Amplifiers Link to comment
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now