A toast to PGGB, a heady brew of math and magic

[Zaphod Beeblebrox]

49 minutes ago, Atriya said:

My Holo Cyan 2 came in! Looking forward to trying PGGB (both PCM and DSD) with it.

Looking forward to what you find.

 

49 minutes ago, Atriya said:

For PCM, @Zaphod Beeblebrox, while this DAC does even 32fS/32-bits, is 32fS/20-bits still preferred?

Yes start with 32fS/20, but also try to go a little lower (19, 18,17) and see where you find it best. Though you may just prefer DSD.

[Atriya]

On 5/7/2024 at 6:31 PM, Zaphod Beeblebrox said:

Looking forward to what you find.

 

Well, first observations are that, at least using when USB from a Windows laptop and the Holo Audio ASIO driver, DSD1024 seems really unstable.

 

It does work (at least it works when using HQPlayer; Foobar2000 just plays noise even with the right settings & plugins, and Roon of course doesn't support it), but the music will randomly turn into a single frequency whine.

 

I wonder how stable DSD1024 was when you tested with the Holo May? Or maybe you used I2S and not USB?

[hols]

4 hours ago, Atriya said:

 

Well, first observations are that, at least using when USB from a Windows laptop and the Holo Audio ASIO driver, DSD1024 seems really unstable.

 

It does work (at least it works when using HQPlayer; Foobar2000 just plays noise even with the right settings & plugins, and Roon of course doesn't support it), but the music will randomly turn into a single frequency whine.

 

I wonder how stable DSD1024 was when you tested with the Holo May? Or maybe you used I2S and not USB?

I can share my experience on DSD1024. I have tried it on my Holo May via USB from Windows 10 machine and it plays very stable and the SQ is at another level above DSD512.

There are one or two things we need to pay attention to. One is we must use the appropriate firmware. The 'older' firmware can play DSD1024 and the new firmware cannot. Second is we need to use a USB port from the motherboard directly and cannot use a USB port from a PCIe card. See whether this helps or not.

[seeteeyou]

5 hours ago, Atriya said:

It does work (at least it works when using HQPlayer; Foobar2000 just plays noise even with the right settings & plugins, and Roon of course doesn't support it), but the music will randomly turn into a single frequency whine.

 

First of all, ASIO means exclusive access to USB DACs and therefore having multiple software players running at the same time would only cause more harm than good. In reality the biggest "offender" turned out to be Roon and therefore we've gotta make sure that everything related to Roon is terminated everywhere.

 

If a tray icon of Roon were found in the System Tray, just take care of that. In addition, always check Task Manager and confirm that absolutely nothing related to Roon is still lurking somewhere.

 

Once the ASIO is completely freed up and no longer occupied by Roon at all, things are gonna be back on track.

 

---

 

https://www.head-fi.org/threads/holo-audio-may-dac-speculation.865892/page-54#post-17433524

Quote

DSD modes all the way up to DSD1024 work fine though.

 

https://www.kitsunehifi.com/product/holo-audio-may-dac/

Quote

*note: May DAC DAC does NOT need firmware update in some cases. L2 May and KTE May come shipped with 30.14 firmware (pre aug2021) and Gen2.0 Enhanced USB Module. (compatibility firmware), however if you need to use DSD1024 and PCM 1.536mhz than you would need to install 30.12 firmware which is less compatible with some devices. However the 30.12 firmware does support high res content with Intel Gen7,Gen8,Gen9, Gen10 type Intel boards and many AMD comparable motherboards are incompatible with this firmware. And some dedicated cards with Altera Chipsets. Please contact us about the Gen2.1 USB module if you have AMD CPU. Post Aug2021 KTE Spring3, L2 May and KTE May come shipped with Gen2.1 Enhanced USB Module.  Between Gen2.0 and 2.1 there are no differences in sound quality, only added support for AMD CPU. Many users will not need this module, however if you have AMD CPU, contact us to discuss our swap out program and how you may qualify for this program.

 

Firmware explanations.txt

https://www.kitsunehifi.com/wp-content/uploads/2021/03/HoloAudio-5.00-DriverFirmware-multiDFU-2021.zip

Quote

take note version 30.14 are for those with either JCAT Femto PCI-E card (renasas chipset) or the Matrix Element or Element H card (texas instruments chipset)
these are LIMITED TO DSD512 and PCM768 at this time.

MOST COMMON upgrade firmware - version 30.12 is for motherboard USB chipset which is commonly Intel base USB controller and should support DSD1024 and PCM1.536Mhz as stated above in overview.

 

Therefore we'll ask @hols to take a look and find out if version 30.12 were meant for stable operations of DSD1024 playback or not, many thanks.

 

---

 

Other than that, Jeff Zhu uploaded the latest version 5.62.0 of ASIO driver here

 

https://pan.baidu.com/share/init?surl=b3ZNu62H1aSzH2zuKUpN2A&pwd=8888

 

FYI - an exact copy of HoloAudio_v5.62.0_2023-09-29_setup.exe was mirrored as follows

 

https://drive.google.com/file/d/1cp-FI7Qjnb735ueJMPj3D75DzGultEXb/view

 

---

 

TuneBrowser with DSD1024 could be found below

 

https://docs.google.com/spreadsheets/d/1rHcXP7PExksxZfkif3Rv2CT1Y_ESpNSv2b0nkKlBySk/view#gid=1964212797

 

https://tunebrowser.tikisoft.net/download/

 

https://tunebrowser.tikisoft.net/summary/

[hols]

Yes. firmware 30.12 is the one I installed that can make DSD1024 work.

[seeteeyou]

Thanks @hols for the confirmation.

 

---

  

10 hours ago, Atriya said:

It does work (at least it works when using HQPlayer; Foobar2000 just plays noise even with the right settings & plugins, and Roon of course doesn't support it), but the music will randomly turn into a single frequency whine.

 

I just realized that you're having Holo Cyan 2 while both @hols and @rayon got DSD1024 working fine with Holo May, sorry about that.

 

Both Spring 3 and May should share the same USB module, we could actually see that XMOS chip was present on either version

 

 

OTOH, the USB module of Cyan 2 was changed from XMOS to Altera as shown below

 

https://eliseaudio.com/products/holo-audio-cyan-2

Quote

• Proprietary firmware for USB interface ensures ultra-low latency and highly reliable data transmission.
• USB firmware contributes to 2-4 times higher performance than official firmware.

 

No wonder they didn't post any firmware downloads on both pages linked below

 

https://www.kitsunehifi.com/product/holo-audio-cyan-2-dac/

https://wildism-audio-hk.myshopify.com/products/holo-audio-r2r-cyan-2-r2r

 

However, ASIO driver version 5.62.0 was indeed found on these pages of Cyan 2

 

https://www.ebay.com/itm/156150774703

https://www.taobao.com/list/item/760569850699.htm

https://www.aliexpress.com/item/1005006455531491.html

 

Therefore it might work better than version 5.57.0 or below then?

[Atriya]

@hols and @seeteeyou: thanks for your inputs!

 

It is indeed a Cyan 2 I have and not a May/Spring.

 

The USB port is indeed directly on the motherboard and not from a PCI-e card. I am also not running Roon (checked in Task Manager).

 

I upgraded the ASIO driver to 5.62.0 (thanks for that link). However, I still get audio for anywhere between a few seconds to a few mins, then just a monotonic whine. This noise doesn't go away when I stop the audio file, but it goes away when I pull out the USB cable.

 

I will try using TuneBrowser to see if it works any better.

 

EDIT: TuneBrowser is working like a dream! No issues at all so far.

[Atriya]

Some very preliminary notes on comparing PGGB on

1. My Weiss DAC501-MkII, which I have had for a while, and
2. The Holo Audio Cyan 2, which arrived just a few days ago, and is my first and only R2R DAC.

 

Signal chain to DAC:

Music server -> USB -> Cyan 2

Music server -> direct ethernet connection -> Weiss

 

Downstream chain used to compare: Mass Kobo 433 headphone amp, Focal Utopia, nice cables.

 

PGGB settings:

I understand that the following two settings are very dissimilar (after all, one is PCM and the other DSD), but they represent the best possible PGGB settings for their respective DACs, making this post as much a comparison of the two DACs in question, as it is a comparison of PGGB settings.

Weiss: 4fS, dither only (after a lot of trial & speculation, this setting seemed most reasonable for the Weiss).

Cyan 2: DSD1024, 9th-order noise shaping.

256-bit precision for both.

 

Initial findings:

The Weiss sounds "cleaner" than the Cyan 2, but also "leaner". The imaging is also "sharper" than the Cyan 2, and there is a greater sense of tangibility to vocals, which seem to "pop out" at the listener as if the singer is "alive" in the recording. The soundstage is both wider and deeper (set back from the listener) with the Weiss, but it is also much flatter. The Cyan 2 wraps instruments around my head (I believe "holographic" is the word for this), whereas the Weiss lays them all in front of me, perfectly positioned on a wall of sound that is expansive in both height and breath. The Weiss is probably more detailed (not 100% sure yet, need to listen closely for microdetails). I like the bass on the Cyan 2 much better: it is warm and palpable and inviting and floats in the 3D space the Cyan 2 creates, whereas the Weiss' bass is just clinical in comparison, though maybe a tad punchier.

 

Overall, from these initial impressions, I still think the Weiss is the better DAC technically, but I can absolutely see why R2R DACs (if the Cyan 2 is at all a good ambassador for R2R DACs in general) are so popular. The Cyan 2 would probably be further improved by using a Holo Red (-> I2S -> Cyan 2), instead of USB.

 

[Zaphod Beeblebrox]

16 minutes ago, Atriya said:

Some very preliminary notes on comparing PGGB on

1. My Weiss DAC501-MkII, which I have had for a while, and
2. The Holo Audio Cyan 2, which arrived just a few days ago, and is my first and only R2R DAC.

 

Signal chain to DAC:

Music server -> USB -> Cyan 2

Music server -> direct ethernet connection -> Weiss

 

Downstream chain used to compare: Mass Kobo 433 headphone amp, Focal Utopia, nice cables.

 

PGGB settings:

I understand that the following two settings are very dissimilar (after all, one is PCM and the other DSD), but they represent the best possible PGGB settings for their respective DACs, making this post as much a comparison of the two DACs in question, as it is a comparison of PGGB settings.

Weiss: 4fS, dither only (after a lot of trial & speculation, this setting seemed most reasonable for the Weiss).

Cyan 2: DSD1024, 9th-order noise shaping.

 

 

Initial findings:

The Weiss sounds "cleaner" than the Cyan 2, but also "leaner". The imaging is also "sharper" than the Cyan 2, and there is a greater sense of tangibility to vocals, which seem to "pop out" at the listener as if the singer is "alive" in the recording. The soundstage is both wider and deeper (set back from the listener) with the Weiss, but it is also much flatter. The Cyan 2 wraps instruments around my head (I believe "holographic" is the word for this), whereas the Weiss lays them all in front of me, perfectly positioned on a wall of sound that is expansive in both height and breath. The Weiss is probably more detailed (not 100% sure yet, need to listen closely for microdetails). I like the bass on the Cyan 2 much better: it is warm and palpable and inviting, whereas the Weiss' bass is just clinical in comparison.

 

Overall, from these initial impressions, I still think the Weiss is the better DAC technically, but I can absolutely see why R2R DACs (if the Cyan 2 is at all a good ambassador for R2R DACs in general) are so popular.

 

Thanks for this, when comparing two different DAC's it will certainly come down to preferences and to a large extent you are also comparing the analog output stage of the two, as you have 'maxed out' the digital filters/processing.

 

With Cyan 2, I would suggest couple of things if you have not already tried it.

1. Try DSD512 with both 7th and 9th order modulator, as the highest rate may not always be the best for all DACs
2. Try DSD 1024 with 7th order modulator, to see if the decreased out of band noise has any audible effect.

[Zaphod Beeblebrox]

6 minutes ago, mpaulson540 said:

For me, this "bite-size" method has worked well, it just took a selection process that at first seemed difficult but became fun as a I rediscovered my music collection. Listening to a 256x1 version of Dena DeRose "A Walk in the Park" with Martin Wind on Bass and Matt Wilson on Drums that sounds amazingly lifelike (even with my simple audio system). 

 

If I remember correctly, your DAC is the Denafrips Ares ||, are you preferring DSD 256 x 1 now to your previous method?

[mpaulson540]

40 minutes ago, Zaphod Beeblebrox said:

If I remember correctly, your DAC is the Denafrips Ares ||, are you preferring DSD 256 x 1 now to your previous method?

Yes, 256 x 1 is superb. Even with my HiFiman Sundaras, the improvement is more than a little noticeable. I have always preferred DSD on the Denafrips, I just prefer it more now :) 

[Atriya]

On 5/9/2024 at 4:29 PM, Atriya said:

Some very preliminary notes on comparing PGGB on

1. My Weiss DAC501-MkII, which I have had for a while, and
2. The Holo Audio Cyan 2, which arrived just a few days ago, and is my first and only R2R DAC.

 

Signal chain to DAC:

Music server -> USB -> Cyan 2

Music server -> direct ethernet connection -> Weiss

 

Downstream chain used to compare: Mass Kobo 433 headphone amp, Focal Utopia, nice cables.

 

PGGB settings:

I understand that the following two settings are very dissimilar (after all, one is PCM and the other DSD), but they represent the best possible PGGB settings for their respective DACs, making this post as much a comparison of the two DACs in question, as it is a comparison of PGGB settings.

Weiss: 4fS, dither only (after a lot of trial & speculation, this setting seemed most reasonable for the Weiss).

Cyan 2: DSD1024, 9th-order noise shaping.

256-bit precision for both.

 

Initial findings:

The Weiss sounds "cleaner" than the Cyan 2, but also "leaner". The imaging is also "sharper" than the Cyan 2, and there is a greater sense of tangibility to vocals, which seem to "pop out" at the listener as if the singer is "alive" in the recording. The soundstage is both wider and deeper (set back from the listener) with the Weiss, but it is also much flatter. The Cyan 2 wraps instruments around my head (I believe "holographic" is the word for this), whereas the Weiss lays them all in front of me, perfectly positioned on a wall of sound that is expansive in both height and breath. The Weiss is probably more detailed (not 100% sure yet, need to listen closely for microdetails). I like the bass on the Cyan 2 much better: it is warm and palpable and inviting and floats in the 3D space the Cyan 2 creates, whereas the Weiss' bass is just clinical in comparison, though maybe a tad punchier.

 

Overall, from these initial impressions, I still think the Weiss is the better DAC technically, but I can absolutely see why R2R DACs (if the Cyan 2 is at all a good ambassador for R2R DACs in general) are so popular. The Cyan 2 would probably be further improved by using a Holo Red (-> I2S -> Cyan 2), instead of USB.

 

 

Further listening reinforces my initial impressions: definitely more detail and accuracy on the Weiss, but the Cyan 2 has a much more relaxed and 3D presentation that I'm growing to love. While the Weiss can be fatiguing, the Cyan 2 has not been at all: instead, the holographic presentation draws the listener into the music.

 

If I didn't care for the relative portability of the Weiss (I do), I would be tempted to sell it and get a Holo May KTE.

[Zaphod Beeblebrox]

10 minutes ago, Atriya said:

If I didn't care for the relative portability of the Weiss (I do), I would be tempted to sell it and get a Holo May KTE.

 

And the headphone amp will come handy if you want a 'transportable' setup.

 

Did you get a chance to compare DSD512 with DSD1024?

[Atriya]

13 minutes ago, Zaphod Beeblebrox said:

 

And the headphone amp will come handy if you want a 'transportable' setup.

 

Did you get a chance to compare DSD512 with DSD1024?

 

Yes, though I realized that the Weiss' built-in headphone amp doesn't sound quite as good as (for example) my Mass Kobo 433 even with my Utopias. The difference would probably be even starker with planars.

 

Not yet, but soon...

[Zaphod Beeblebrox]

Just now, Atriya said:

Yes, though I realized that the Weiss' built-in headphone amp doesn't sound quite as good as (for example) my Mass Kobo 433 even with my Utopias. The difference would probably be even starker with planars.

Understood. I had the 502 before the Mk2 update, it sounded significantly better out of the balanced output from the back.

[seeteeyou]

In order to ensure that DSD files could be played from specific drives with low latency, I've been looking for relatively large ones that wouldn't cost an arm and a leg

 

https://news.ycombinator.com/item?id=38527437

Quote

Optane SSDs are really special and still have a place. SLC is 30 microseconds at best, FIO puts my 905p optane drive at 5-11 microseconds. That means for a single threaded latency bound task, a 905p may finish it 6 times faster than an SLC ssd. 20 times faster than MLC!

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/542/#comment-941760  

On 3/25/2019 at 4:50 AM, romaz said:

SSD vs Optane

 

Just for kicks, I decided to compare a 58GB Optane card against a Samsung 500GB 960 EVO NVMe SSD in the M.2 slot of the Asrock board.  This was a very brief comparison because it didn't take long to realize how much more harsh the SSD sounded even with all of my isolation schemes in place.

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/543/#comment-942228  

On 3/26/2019 at 3:11 PM, romaz said:

An Optane card can function just like an SSD if you want it to and that is how I'm using it.  Optane cards don't generate anywhere close to the same amount of noise as SSDs.  The have a noise signature closer to RAM.

 

Right now it's back to $399.99 but most likely we're gonna see even better prices at some point

 

https://www.ebay.com/itm/386184520715

https://www.newegg.com/intel-optane-905p-1-5tb/p/N82E16820167505

https://www.reddit.com/r/buildapcsales/comments/1cbnizs/ssd_intel_optane_15tb_905p_25_x15mm_u2_pcie_30x4/

And then it's also good to know that even SLC NAND chips from Kioxia should be catching up

 

https://www.servethehome.com/kioxia-fl6-xl-flash-and-dapustor-ssds-at-fms-2022/

Quote

Using the Kioxia XL-FLASH, the company was able to hit very low latency for 4K random reads. Effectively, DapuStor is using the SLC NAND, and tuning controller firmware for low latency.

 

The latency looked quite impressive when they're comparing Kioxia XL-FLASH inside X2900P to 3D XPoint inside Optane P5800X

 

https://www.servethehome.com/dapustor-xlenstor2-x2900p-800gb-review-the-100-dwpd-next-gen-slc-optane-alternative-kioxia-intel-optane/

 

As long as we aren't killing the Optane when it's being used for caching purposes, we could simply copy our DSD files from our backup to the Optane for the sake of further improving SQ

 

https://forum.level1techs.com/t/would-optane-be-good-for-running-virtual-machines-from/199508/19

Quote

905P: 10 DWPD

• 380 GB: 7 PB lifetime
• 480 GB: 9 PB lifetime
• 960 GB: 18 PB lifetime
• 1.5 TB: 27 PB lifetime

 

https://www.tomshardware.com/reviews/intel-optane-ssd-dc-p5800x-review/2

 

 

I guess that 1.5TB Optane 905P should be going for about 300 bucks during Black Friday then?

[Atriya]

23 hours ago, Atriya said:

 

Yes, though I realized that the Weiss' built-in headphone amp doesn't sound quite as good as (for example) my Mass Kobo 433 even with my Utopias. The difference would probably be even starker with planars.

 

Not yet, but soon...

 

Comparing between DSD512 and DSD1024 on the Cyan 2 (9th-order on both), I find them to be very close.

 

I think I do prefer DSD1024 as sounding just a tad more transparent & etched overall.

 

I didn't do a blind test or anything, so it could just be "bigger is better" bias, and in any case, the differences seem small.

 

What I did discover though, is that I get an error when doing 1-stage DSD1024 for some tracks (could be length-dependent), so I had to resort to 2-stage DSD1024 for the comparison. The DSD512 was 1-stage. A screenshot of the error is attached.

 

[Zaphod Beeblebrox]

4 minutes ago, Atriya said:

Comparing between DSD512 and DSD1024 on the Cyan 2 (9th-order on both), I find them to be very close.

 

I think I do prefer DSD1024 as sounding just a tad more transparent & etched overall.

 

I didn't do a blind test or anything, so it could just be "bigger is better" bias, and in any case, the differences seem small.

Thanks, do compare 7th order modulator too, so you can settle on what sounds best for your DAC.

 

5 minutes ago, Atriya said:

What I did discover though, is that I get an error when doing 1-stage DSD1024 for some tracks (could be length-dependent), so I had to resort to 2-stage for the comparison. A screenshot of the error is attached.

Most likely reason is running out of memory (please DM the logs). 

[Atriya]

1 minute ago, Zaphod Beeblebrox said:

Most likely reason is running out of memory (please DM the logs). 

 

Yes, I have "only" 64 GB right now, though I plan to buy another 64 GB soon.

[Zaphod Beeblebrox]

8 minutes ago, Atriya said:

 

Yes, I have "only" 64 GB right now, though I plan to buy another 64 GB soon.

You could compensate it with large enough page file and also run it with 'Experimental ' option on.

 

I have a calculator to help you figure out how much VM you need to allocate: VM calculator - Zaphod's Guide (remastero.com). You could consider adding a second NVME drive too.

 

[Zaphod Beeblebrox]

Version 6.1.50: The DSD Edition (PGGB Plus)

 

I have released update, you can grab v6.1.50 here. The 'Experimental' feature that I added to improve paging performance is now called 'Reduce contention' and is on by default (if you install PGGB for the first time). 

 

Logging under the hidden menu now has three modes Sparse/Simple/Verbose. Sparse and Simple mode have the potential to slightly improve processing speed.

 

On Macs, I have provided the option to specify a second location for page files, to be of any benifit, this second location needs to be a different drive. an external Thunderbolt drive can be an option.

 

 

 

Release Notes:

Version 6.1.50: The DSD Edition (PGGB Plus)
    SQ Changes:    None, same as version 6.0.42
    Features:

• Changes logging, now logging provides sparse/simple/verbose mode.
• Sparse or simple logging mode can result in up to 10% improvement in speed (not guaranteed, but potential exists for some short tracks or some long tracks).
• The 'Experimental' option is now out of beta and is called 'Reduce contention', i.e., reduces NVME contention during paging. It is on by default.
• Adds indicator to show if '1ch' or 'RC' (Reduce contention) setting is on.
• Allows addition of one more paging drive
• Mac only: Adds retry logic if processing terminated in 2ch mode.
• Mac only: Cleans up orphaned paging files.

[Atriya]

It looks like the new Laiv Harmony DAC is creating quite a splash. 768Khz, DSD256 is the max. I hope someone here buys it and tries PGGB with it. Not going to be though! 🤣

[kennyb123]

5 hours ago, Zaphod Beeblebrox said:

The 'Experimental' feature that I added to improve paging performance is now called 'Reduce contention' and is on by default (if you install PGGB for the first time)

Mac users should make sure that this is enabled.  Nice reduction in processing time.

[Zaphod Beeblebrox]

1 hour ago, Atriya said:

It looks like the new Laiv Harmony DAC is creating quite a splash. 768Khz, DSD256 is the max. I hope someone here buys it and tries PGGB with it. Not going to be though! 🤣

It is R2R DAC, but has no pure DSD path, it will convert DSD to PCM. It is best to input 16fS noise shaped PCM, I have not seen any measurements, so not sure how good the linearity measurements are.

 

[LowOrbit]

Two quick updates:

 

I've been running the latest PGGB release for the last 24 hours and can confirm that using the new speed-up options, I am seeing good improvements in processing. Maybe 15% or more for similar track lengths.

 

My low-intensity evaluation of DSD512 vs DSD1024 on the marvellous Dac200 continues. Currently I would say that with some well-recorded acoustic recordings DSD1024 offers a little extra realism, more believability in the "room" in which the music was played and subtly more robust instrument character. But DSD512 is pretty close and with most recordings seems to offer a little more musical energy.

 

I appreciate that is not an objective assessment, but these are my impressions only. I am happy with either and am generally disinclined to picking out individual albums for this or that treatment,  in the same way I use one set of transducers for all my music listening. The music is the thing and with PGGB it's all rather splendid.

 

With the improvements in processing speed I can generate more comparison materials but DSD512 gives me plenty to enjoy.