Euphony OS w/Stylus player setup and issues thread

[TheAttorney]

On 10/24/2020 at 6:39 AM, edwardsean said:

However, with the new DC umbilical I went up again to 0-1 gstp 2-7 with some nice results. So, I decided to go further than I had before and tried 0 gstp 1-7, giving the playback engine 7 cores. 

 

I don't know, if your system can handle it, the benefits to SQ are knocking me out. More dynamics, clarity, soundstage, definiteness, detail, and extension in the highs.

 

I've been watching temperatures, because 3 of the 4 physical cores have dropped in temp, but core 1 is now up and holding around 74 degrees. That's acceptable to me.

 

I could resist no longer and enabled hyper-threading on my 7i7DN (8650U) so I could try out these latest core isolations.

 

My first reaction was how different HT sounded compared to my usual HT off, even with the same speeds and equivalent core isolation. Smoother and more natural. At first I wasn't sure if this smoothness was good or bad, but going back to HT off resulted in a thinner, "digital" sound, so HT is staying on. And no difference to core temperature or power consumption.

 

So good news, and now the better news: Whilst I broadly agree with your findings, I think I've found an even better combination than [0 gstp 1-7]. To my ears, [0-1 gstp 1-7] gives even more of the above SQ benefits, plus the further bonus of a significantly lower temperature for physical core 0. The temperatures are now more balanced across the cores, which isn't intuitive because it's the same physical core 0 in both cases. And virtual core 1 is now sharing gstp with everything else. Core Isolation moves in mysterious ways.

 

Compared to above SQ differences, I found turbo boost to be marginal, so I'm sticking with 1.1Ghz for the peace of mind of much lower temperatures. So my new favourite is:   1.1 Ghz, HT on, [0-1 gstp 1-7]

 

And I'm smiling because I've gained a SQ boost that I had previously spent 4 figures sums to achieve by h/w ugrades.

 

 

 

[TheAttorney]

11 hours ago, edwardsean said:

I tried out 0-1 gstp 1-7, and the temps got displaced to the first two cores. So 1 went down and 2 went up; both are low 60s. This is welcome.

To my ears the sound suffered slightly. It's really subtle, but it sounded smaller and flatter to me. 0 gstp 1-7 still sounds a bit larger and more vivid to me.

I'm still getting used to my new reference point, where the most obvious change was enabling HT. so the difference between core isolation 0 and 0-1 (and other close variants) is more subtle and subject to imagination. On further listening, 0-1 seemed a touch brighter than 0, which can give the impression of greater detail and impact, depending on the recording. With further listening, I tend to agree that 0 is a touch better overall. 

 

The temperature issue of squeezing "everything else" onto a single virtual core only applies to high CPU speeds. Any speed up to my 7i7DN's base frequency of 1.9Ghz has minimal effect on temperature no matter what I throw at it. Another thing to bear in mind with the single virtual core 0 is that CPU load does max out at 100% when I update the database or buffer files, etc. Spreading the load across 0-1 definitely helps here. So best to temporarily add more cores when doing any major database work.

 

Anyway, I also tried adjusting CPU speeds with my new reference and these are even more subtle and subject to imagination, but I felt that 1.5Ghz was fractionally better than 1.1Ghz, so my new favourite is 1.5Ghz, HT on, [0, gstp 1-7].

And that's enough testing for me - back to just enjoying the music.

 

BTW, for the avoidance of doubt, I have nothing to do with the law 🙂

 

 

[TheAttorney]

18 hours ago, Randomrunner said:

There is a new version that just came out, where can one find the release notes?

Under Settings > Update tab, click on "Changes" near top of the screen.

Also, the changes log is displayed when clicking the version number of current version.

 

My update went smoothly. As before, if updating direct from ramroot, ramroot gets automatically disabled, so needs to be re-enabled and then rebooted.

 

First impressions: The settings tab is much neater now with 3 sub-tabs including Expert, so quicker to get to that compared to the previous scrolling.

Radio now works.

All my playlists give "no items found" message, but I haven't used playlists in months - and they generally were cached in optane drive which I also don't use for caching any more since file buffering became available.

[TheAttorney]

9 hours ago, flkin said:

Restarting the Euphony app twice solves this [playlists]. 

Yes indeed, although in my case playlists still didn't work after 2 reboots if I include the original reboot to re-enable ramroot.

On a whim, I then did a "save root fs to disc", then a 3rd reboot and this time playlists came back. Not sure which of those last 2 actions mattered.

 

Going back several releases, I had occasional glitches when cached files were added to a playlist. Some files would later not be found. This may have been because the cached files had since been swapped out when the cache filled up, or maybe another reason. Anyway, now the playlists are more robust because Stylus automatically picks up the original file location if the cached file no longer exists. Not sure if it was this particular release that fixed that one, but it runs smoother now.

[TheAttorney]

A quicker way to resolve the playlist issue is to "Restart App" from the power menu drop-down. Much faster than rebooting and can quickly solve other issues when something goes out of sync. I had used this in the past and then forgot about it.

[TheAttorney]

4 hours ago, Chopin75 said:

Well, I just tried after the latest update, the entire Euphony is offline, and it is supposed to switch back on after a set amount of time ? The default is 206, so it means 206 min ?

 

You get more info on this new feature if you hover over the ? symbol. It's the number of minutes that Euphony has calculated to play whatever is in the buffered queue. If you subsequently change your mind and want to gain control earlier, then the only action is to physically reboot the server, which is not recommended because such an immediate hard stop could cause problems - so only to be used as last resort.

 

I'm intrigued that this function seems to work for wifi as well as ethernet. I use my NUC's stock wifi card for all networking - primarily for convenience and for a complete reduction of router-based spaghetti. The wifi may well have an impact on SQ, but after a year of use, I don't think it's a major impact. I'm aiming to find out later today - fingers crossed that this new function does work and that the wifi comes back after the designated number of minutes.

[TheAttorney]

4 hours ago, TheAttorney said:

I'm intrigued that this function seems to work for wifi as well as ethernet.

Željko has informed me that wifi won't automatically reconnect when the play queue completes, so I won't be trying this for now. Maybe will be implemented in a future release.

[TheAttorney]

On 1/9/2021 at 5:13 PM, TheAttorney said:

Željko has informed me that wifi won't automatically reconnect when the play queue completes, so I won't be trying this for now. Maybe will be implemented in a future release.

Seems that Željko never rests. Yesterday he RC'd in to add the final step to enable wifi reconnection.

The good news is that it works very well. near the end of the last song in the queue, wifi kicks in again and everything back to normal. I expect wifi support will be formally added to next release for the minority who use this feature.

 

The better news, for me, is that I couldn't reliably tell the difference between wifi enabled of disabled. Which could be due to:

a. My system or ears aren't resolving enough

b. Wishful thinking because I find wifi networking to be so convenient

c. On my particular NUC configuration, Euphony Stylus does a great job of minimising the impact of built-in wifi - to the point I can stop worrying about it and worry about more productive things instead.

 

I've chosen to believe "c" because of "b". In which case, maybe others with similar set-ups may find they can stop fussing over super switches, super clocks and super ethernet cables, and just enable their NUC's built-in wifi? 🙂

 

On 1/9/2021 at 11:38 PM, austinpop said:

You need a safe word. 

Maybe Željko can recognize an event — like one press on power button — to abort and re-enable the network interface(s).

This awakened my latent s/w designer tendency 🙂.

If this new network-disable feature is primarily for occasional testing to get a baseline reference point, then the current interface works fine. But if users will continue to use this long term to listen to music, then something like the following would be better:

 

When the network is disabled, instead of "Euphony is offline" being displayed, you see "Euphony is playing offline - press [special key] to interrupt". The special key sets a flag on the browser. Meanwhile, the server will, every 5 minutes, briefly reconnect the network just to check flag status and then fully reconnects if flag requests it. Which means that if you change your mind after loading 20 albums, you have to wait no longer than 5 minutes, or configurable time, to regain control.

 

There may be browser limitations that mean it can't be done exactly as above, but I'm sure something equivalent could be achieved by clever people. I'm just the ideas man...

[TheAttorney]

1 hour ago, flkin said:

Or even return online after current track is finished rather than a set time. 
 

I considered that, but then thought about the classical works that have lots of short tracks, causing network to keep switching in/out, and the prog rock tracks that last half an hour. I would expect the duration parameter to be configurable. So that purists (suckers for punishment) could choose a longer duration and realists choose shorter.

 

As an aside, I've been trialing Qobuz for a couple of months and have been delighted with the SQ - broadly as good as local files.

And also pleasantly surprised with Qobuz/Stylus integration. It's not as fully integrated as with Roon as you might expect, but it's more than usable in its own way, particularly if you make full use of Qobuz "favourites" function.

[TheAttorney]

50 minutes ago, biosailor said:

Euphony's library management is a disgrace!

 

I basically have to jump from album to album and correct each one for proper tagging. In such a way, it works, but it is tedious, to say the least!

Further to @dminches reply, whenever I've had such tagging issues where, say, one of the tracks appears elsewhere under a different artist or different album, I've always found this to be due to incorrect tags in the first place.

 

Basically, Euphony Stylus unconditionally believes what you give it and arranges the library to match. You don't see this in Roon because Roon assumes it knows better that you for any given album and recreates the library from its own online database, which works well most of the time, but isn't perfect either. E.g. for Classical albums, Roon often imposes the wrong cover art and I have to manually choose the original version.

 

Stylus' library management isn't it's strong point, that's undeniable. And it may well have some bugs outstanding, but most other apps have their own foibles. I love Roon's UI, but I struggled for ages to try to understand its library management system, much of which is deliberately hidden from users, for good and for bad.

 

I get around all of the above by having a master filestore for music files that's managed by JRiver (the first app I happened to buy years ago). JRiver's library and file management far exceeds Roon's, particularly with it's batch commands which are very powerful (if rather unintuative). Then import to Roon or Stylus or any other app. Which means I retain full control of my music files, which means it's easier to change to different music players whenever I want to.

[TheAttorney]

12 hours ago, biosailor said:

Ok, so you do the metadata handling in JRiver, if I understand you correctly? I need to look into this as I happen to have a JRiver license.

Yes that's correct.

Having a "master" music filestore  is in some ways more effort because you have to go back to the master to make any corrections, and then copy any changes back to your music app.

 

But it starts making sense if you have more than one music app (say, Roon, Stylus and HQP) on different systems, each with their own way of library management. And even more sense when you change apps with the knowledge that you have full control of the original folders and files - as opposed to the previous app's unique interpretation of your files. Some people like to have full control of their folders/files and others are happy to let the music app sort it for them. I differentiate "files and folders", which contain the actual music, from "library" which is unique and internal to each music app and is out of my control.

 

A further benefit is that my master system also has a CD drive, so using JR's ripping function on a master filestore is the most reliable way to keep all my other music systems in sync.

 

The main issue with JRiver's library/tag/file management is that it is so powerful there are usually several ways of making any correction, and it's not obvious which is the best way - Google is your friend here.

 

For example, say you have a number of albums where the artist is spelt McVoy, and 1 album where it's spelt MacVoy. You can correct this directly in the UI Album view, or the Artist view. Maybe one row at a time, or maybe highlighting a number of rows. Or you can create a batch command to do it in batch mode. Or you can go to the actual folder and manually change the file or folder name and get JR to update it's internal library from the filename. Definitely a learning curve required to get the best out of it, but Google helps.

[TheAttorney]

18 hours ago, Joerg D said:

Does it matter whether I select in the setting that the music is loaded 100% into the memory or whether I select the function 
"Buffer queue to Ram " ?

 

Best to always select "Buffer before play=100%" to load the current track into RAM before play starts (and also loads next track for gapless switchover).

The "Buffer albums added to queue" is an extension of the above and buffers the whole album(s) into RAM (up until the RAM becomes nearly full). Some report that this further improves SQ, I haven't noticed that so much. Both these options apply to local music files (streamed files from Qobuz etc only buffer current and next tracks).

 

As has been mentioned, I and some others think that the biggest single SQ boost you can do is to enable ramroot, which loads the entire Euphony operating system and Stylus player into memory, which means that disc I/O is greatly reduced. This is in Expert Settings and I don't think it's described in the user guide. Boot time is obviously increased with ramroot, but it's worth it.

 

There are other SQ-boosting options in System -> Expert Settings, which requires further explanation, but first get ramroot going.

 

 

[TheAttorney]

On 1/26/2021 at 7:50 PM, RickyV said:

On Monday 18 January I ordered this memory from Mouser, https://nl.mouser.com/ProductDetail/Apacer/D2323240S004/?qs=GBLSl2AkiruWco00G6Jb1Q%3D%3D and collected it on Friday 22 January in Holland. This 8 GB ram stick replaced this type of Apacer ram 4GB 75.B93GJ.G010B which was a good up-tick. This mouser ram is I believe the same ram that Nenon provided a while back. And again this ram is a really good up-tick to previous ram, depth, dynamics and smoothness are all better.

This sounds interesting.  I also have  the 4GB 75.B93GJ.G010B on my NUC7i7DN, which I found to be a very worthwhile improvement over the stock 8GB RAM.

 

So you're saying that the new D23.23240S.004 8 GB Apacer was a further step above your old 4GB Apacer?

Approximately how big an improvement was that compared to the original improvement over stock?

 

Although I'm still finding that 4GB is enough for ramroot and a handful of buffered redbook albums, a bit more RAM capacity would help future-proofing.

 

[TheAttorney]

11 hours ago, RickyV said:

So I started with the old version after a week of D23 listening, depth: seems flatter, it has depth but the D23 is more homogeneous that goes also for the sound stage more homogeneous, seems bigger and more separated/ defined and fuller. 
It’s well worth the 100 euros.

Pitty mouser didn’t had 4GB Simms otherwise I would have bought two of those. 

Thank you for doing that extra comparison.

I've put an order in for with mouser uk - on back-order, so I have to wait over a month. But that's still better than the situation when Nenon first mentioned the new models - at that point they were unobtainable from the usual suppliers and so I lost interest.

 

I'm very happy with my current setup, but the SQ improvements you listed were exactly in the direction I'm aiming for.

[TheAttorney]

3 hours ago, Joerg D said:

That is the error message. Željko then activated it for me. I don't know how he did it. The setting is gone after every update. I use Tidal, Qobuz and my music from the NAS. I didn't set the number of songs in the playlist.

Jörg

 

Ramroot with 4GB RAM only uses 39% of the RAM, so plenty of space for a handful of redbook FLAC albums to also be buffered to play queue. I don't know what you mean by "set the number of songs in the playlist", but there is no need to enter anything into the "Buffering RAM (GB)" field. Let Stylus sort it out automatically, so it will allow you to keep adding tracks to the play queue, but it will stop buffering them (if you've chosen the buffer option) as RAM nears it's limit.

 

I did get that error message a long time ago, which caused Željko to adjust the RAM-limit check to cater for 4GB, but I've not had that error for several Euphony/Stylus updates now.

 

I think your issue may be that you are trying to Enable ramroot when there are items in the playlist, so Stylus has less free RAM available to it at the point of checking the limit. Try clearing the playlist, then enable ramroot. Once ramroot is enabled, you can then add as many tracks as you like to the playlist, as above.

 

Regarding 4GB vs 8GB SQ, the more important question is whether your are using a better sounding RAM. And the better sounding Apacer RAM is currently not readily available in 4GB, as per the recent posts in the last page or two of this thread.

[TheAttorney]

On 2/17/2021 at 7:28 PM, c-w said:

To be precise, it gets disabled after 'image' update but not after 'feature' update. There have been several feature updates lately before this one and they did not disable ramroot.

I fell into this trap by not bothering to check ramroot after the first of the February image updates (because previous Feature updates didn't change ramroot). Then I spent 2 days wondering why I wasn't engaging with the music any more. It's not that it was bad, just that the magic was missing. And sure enough, ramroot was disabled.

 

On re-enabling ramroot (as well as latest Feb update just for completeness), the magic returned. Bliss...

Having temporarily upset my reference point, it's hard now to reliably compare SQ to the previous release, but yes the new one is possibly a bit clearer and more focused, maybe at a slight expense of fullness. I rather like it, and anyway I can easily dial in clarity vs fullness by a small change to my CPU speed, or headphones height adjustment.

 

But for those that don't like the new presentation, rather than consider ways of reverting back, it would be much better if Željko finds what caused the difference and then produces a version that gives us the best of both worlds 🙂. I think he does listen in on this thread sometimes, so maybe he's already working on it? Maybe a new expert setting if we can't have it all at the same time?

 

 

[TheAttorney]

15 hours ago, Joerg D said:

I have increased the min. max. frequency of the CPU. This made the sound nervous and cool. Have taken the old frequency again and the warmth and calm come back.

It's worth experimenting with the EXACT lower speed (I only ever change the "Max" value to change this):

 

My 7i7DN has a base speed of 1.9GHz, with 4.2GHz in Turbo mode.

For the past year, my preferred range has been between 1.0-1.5 GHz, and going below or above this range does not improve SQ. More recently, I've honed in on 1.2Ghz as being the clarity/warmth sweet spot. Going lower than this loses warmth, and above this it gets a bit smeary. Jumping straight to Turbo also sounds good, but I avoid this because any potential subtle benefits do not justify the much higher core temperatures.

 

The above differences are subtle, but worthwhile. A bigger difference in changing the character of the sound is when I enabled Hyperthreading to give a warmer sound. There are so many variables involved here as to what affects what. I use:

 

    CPU 1.2GHz, Hyperthreading on, [0, 1-7 gstp] core isolation, Ramroot on 4GB Apacer 2400 RAM, wifi networking...

 

... and I'm not getting the brightness issues with the '116/117 releases that others are getting. Maybe my system config is just too warm and unresolving to notice the difference?

 

(BTW, the consensus around here is that, the higher the CPU speed, the more important is the quality of the power supply to avoid brightness. My top-quality PH SR7 p/s removes this particular variable from the equation).

[TheAttorney]

26 minutes ago, biosailor said:

Would you mind elaborating a bit on hyperthreading, gstp, and Core isolation? Sorry, newbie here!

 

Typical CPUs have a number of physical cores (mine has 4) which spread the workload. Hyperthreading enables a greater number of virtual cores (mine has 8) which further spread the workload across the physical cores, with the CPU balancing the workload across all these virtual and physical cores. In recent published screenshots, you can see the load readings on 8 virtual cores, but there are only 4 temperature readings because only the physical cores have relevance for temperature. 

 

With core isolation, you can influence the spread of workload by allocating specific processes to specific cores, thereby isolating the music processes from other workloads, thereby subtly improving SQ. Core numbering starts from 0. Key process names are "stylus" the UI, "gstp" the lower-level process that actually plays the music. Other processes for roon etc.

 

A typical example of [0-1, 2-3 stylus, 4-8 gstp] means that gstp is allocated cores 4-8, stylus 2-3 and "everything else" is allocated cores 0-1.

In my own case [0, 1-7 gstp], the allocation is heavily stacked in favour of gstp, with everything else (including stylus) all crammed into virtual core 0 (which is half of physical core 0) because that's what sounds best. This does put a big burden on core 0, but not a problem because of the low CPU speeds means inherently low temperatures. At turno speeds, this could be a problem.

 

The explanation has evolved over many threads. Best to use the search function to read a few of them and get the hang of it.

One useful shortcut: If you hit the apply key without entering anything in the field, you'll see what the current core allocations are (without changing anything).

 

EDIT: To enable/disable hyperthreading, you have go to BIOS settings, Euphony doesn't do this itself.

[TheAttorney]

9 hours ago, flkin said:

Z is able to fine tune the sound in such a subtle way it’s amazing. He has many graduations to offer. It like he’s a chef adding or taking away some ingredient affecting the overall taste. 

Good things are coming 😄

I do hope he will provide this as a configurable Expert option - so that we can then spend endless hours trying out and debating which setting sounds best. Stops us from spending money on new hardware 🙂

[TheAttorney]

On 3/14/2021 at 2:00 PM, NanoSword said:

But I am trying to explain that we need enhance the method of disabling network with euphony .

I think for better sq we need to disable network first .

second buffering the queue .

this should apply in sequence .

I've had difficulty trying to understand what problem you're trying to resolve.

 

I think what you're effectively saying is that, when buffering files into RAM, noise from the network somehow enters and stays inside the RAM alongside the 1's and 0's - thereby permanently corrupting the file in some way. And you can hear the resulting SQ degradation when you later play the buffered file - compared to when you had buffered the files into RAM when the network was disconnected.

 

How did you run this test to compare the SQ of the two different ways of buffering the file?

I don't see how you can achieve this test without Z creating the new feature you're asking for.

[TheAttorney]

On 1/26/2021 at 7:50 PM, RickyV said:

On Monday 18 January I ordered this memory from Mouser, https://nl.mouser.com/ProductDetail/Apacer/D2323240S004/?qs=GBLSl2AkiruWco00G6Jb1Q%3D%3D and collected it on Friday 22 January in Holland. This 8 GB ram stick replaced this type of Apacer ram 4GB 75.B93GJ.G010B which was a good up-tick. This mouser ram is I believe the same ram that Nenon provided a while back. And again this ram is a really good up-tick to previous ram, depth, dynamics and smoothness are all better.

I've now received the exactly the same new "2666 speed" Apacer to replace my Apacer 2400, after several weeks on back-order.

A brief history of my NUC7i7DN RAM journey:

 

1. 2x8GB stock RAM reduced to 1x8GB stock. Incremental but immediately obvious SQ improvement.

2. 1x8GB stock to 1x4GB Apacer 2400. A significantly bigger across-the-board SQ improvement.

3. 1x4GB Apacer 2400 to 1x8GB Apacer 2666. Short answer: Wow! The longer answer.....

 

With step 3, my start point happened to be a cold system and, as a precaution, ramroot disabled. From the first minute, SQ was poor. So I tried again that evening about 10 hours later and with ramroot enabled. What a difference!. The sound was remarkably focused, clear and dynamic. In one word: "Vivid". Also maybe a touch brighter than I'd like? The sound seemed to change over the next couple of days, so that I was starting to forget my original reference.

 

So after 5 days, I swapped back the Apacer 2400 - and the sound immediately became flatter, softer, rather muffled. Difficult to imagine this was a sound I was very happy with a week ago. Now it was dull and boring.

 

Back to the Apacer 2666 and Wow! Again, the one word Vivid comes to mind. Comparing some of my brighter recordings, the 2666 actually handles these better by, for example, better separating the body of a female vocal away from the sibilance and reverb. The 2400 tends to smear all these together. I'm still getting used to the intensity of the new presentation, but remember this is with the latest Euphony release that several others have had a brightness problem with (that overall hadn't bothered me).

 

I can't get over how big a change this has been for me - especially considering it cost less than £100. If anyone is still using stock RAM in their NUC, I consider this to be a mandatory upgrade. As always, YMMV. I have no idea why this was such a big change for me - yet I can't tell the difference whether or not the network is disabled when playing buffered music files.

 

PS. I see that mouser.uk has again got a long back-order lead time, despite having recently received  new shipment.

 

[TheAttorney]

4 hours ago, flkin said:

If your motherboard allows, try changing speed and voltage of the ram I had good results doing this. 

I've attached a screenshot of my RAM BIOS settings. The values on the left are display-only. I can't seem to change them.

One immediate thought: The display shows speed of 2400 despite this model's spec being 2666.

 

What do you suggest the speed and voltage be changed to?

I'm looking for one-shot change - I can't easily keep trying things out in BIOS as I don't have a monitor or keyboard.

[TheAttorney]

On 3/22/2021 at 8:13 PM, Miska said:

 

You don't need to touch any of those other settings. Just select XMP Profile 1 from the "Memory Profiles" list and it should go with the 2666 spec and correct values.

I've tried really hard going through every BIOS screen 5 times and reading lots of NUC guides, but my NUC7i7DN just doesn't support XMP profiles, doesn't support voltage updates, and only supports speed changes that are less than the default 2400 (the highest multiplier in the drop-down is 16). Therefore I can't get any better than 2400 at 1.2V. So I'm done with trying.

 

Shame because I was half hoping that Z could add some of these parameters in Expert settings as they have the potential to be more important for SQ than most of the existing settings. But I guess that there are just too many variables and limitations across the different motherboards for this to be practical.

 

Anyway, even with the defaults, this new Apacer 2666 is miles better than my previous 2400 - despite running at the same speed and voltage. Maybe it's because I run the CPU at a low speed 1.2Ghz that the speed of the RAM is not so important - there must be other quality factors that do matter.

[TheAttorney]

3 hours ago, KATARRAZZI said:

Can you tell me how to order apacer 2666 8gb ram that is suitable for my main Msi 310m pro?  thank you

I'm not at expert, but a brief look at the msi spec shows support for DDR4 RAM up to 2666 Mhz, which suggests that the Apacer I bought below will work fine. But it may depend on the exact CPU and and board configuration. So I'm not guaranteeing anything.

 

https://www.mouser.co.uk/ProductDetail/Apacer/D2323240S004/?qs=GBLSl2AkiruWco00G6Jb1Q==

 

11 hours ago, Speedy381 said:

I saw in earlier posts that you had been running between 1 and 1.5. Was this in bios as well? Just altering in expert settings does not change the CPU reported frequency for me

Yes, I make those changes in Expert settings. They are immediately effective as shown on the Temp display screen.

However, different motherboards may have different quirks. My 7i7DN allows changes up to it's base frequency of 1.9GHz, but anything above that will go straight to Turbo boost of 4.2GHz. So I guess other systems may not work at all.

[TheAttorney]

On 4/6/2021 at 1:35 PM, ASRMichael said:

Ref core isolation - Is there a maximum amount of cores/threads gstp use? 
My core isolation is 0-3 stylus 4-7 gstp 8-19

I suggest you try giving all of stylus over to gstp, i.e.   0-3 gstp 4-19   and then tinker around with variations of CPU speed vs possibly giving even more cores to gstp (but system responsiveness will go down, and temperature will go up, if you go too far)

 

On my 4 core/8 thread NUC, I initially found it hard to tell any difference with any core isolation, but there's a sweet spot combination of CPU speed and core isolation where it all seems to happen - and the differences are now much clearer to me.

 

My previous best combination of core isolaton and CPU speed was   

    0 gstp 1-7   and 1.2GHz.

... because it lost a touch of brighness that came with 0-1 gstp 1-7, even though the latter gave a clearer, more dynamic, sound.

 

As my recent Apacer 2666 RAM has smoothed out slightly with further burn-in, I feel I can now push the dynamic parameters a bit harder, with any remaining brightness controlled by adjusting CPU speed, with the following giving me the best of all worlds:

    0-1 gstp 1-7  and 1.1GHz

 

The differences are probably objectively tiny, but somehow matter more than they should in terms of musical engagement - in fact the balance of clarity vs smoothness is the best I've ever experienced on any previous system. And this is with the latest 217 version that others have had a brightness issue with, so I'm hoping the next Euphony version will be better still 🙂