Should have thought of this before, but disabling the onboard wifi in bios solved this issue. fingerprint now the same. woot.
Was doing some more tests and I think I see the difference between HQ Player OS and Ubuntu. I think it has to do with thermald. On Ubuntu the CPU never crosses 80 C, whereas I have seen the cpu temp on HQ OS get up to as high as 84 C once in a while. i.e. Ubuntu seems to have a max of 80 C and is thermal throttling a little bit.
Not sure on AIO longevity but I will warn you that shoving the 120mm AIO cooler into the DeskMeet requires extremely tight fit/bending of the water hoses.
How are getting your CPU to go above 80C with Ubuntu 22.04? I canāt ever get mine to cross 80, which seems to be limiting me a bit.
I can get 14900K do ASDM7EC-super at DSD1024 without GPU and without breaking a sweat using 280mm AIO cooling. But building HQPlayer OS will run the CPU to thermal throttling pretty quick (all cores at 100% constantly for couple of hours).
On server type of OS I just run āsensorsā. But on desktop OS I run for example GKrellM. When I see cores sticking at ācriticalā temperature level constantly I know it is being thermal throttled. I have cooling set to āquietā profile. So the CPU easily runs around max temps.
What I say is without convolution at default filters. I donāt use ext3 for Nx, that is certainly heavier. But what kind of load figures do you get with that for E-cores set to filter? My tests are done with E-cores running filters.
Are you sure for example ext2 wouldnāt be enough for Nx? I would consider steepness of ext2 way enough for 192k source.
ok, so I randomly decided to remove hqplayerd_5.6.2-19_amd64.deb and install hqplayerd_5.6.2-19intel_amd64.deb instead. The install bailed towards the end as HQPlayer user account already existed but I manually enabled in systemctl and it starts up fine. And lo and behold, itās now more performant :!@#
I can run 192k + ext3 + convolution ā ASDM7EC-super @ DSD1024 without overclocking or gpu. With e-cores=filters and blocks=16. This is better than what I can get out of HQ Player OS without having to overclock a bit.
The 2 main p-cores run a little hotter but it looks like in general more processing is being offloaded to e-cores - they are running at up to 70% whereas before they never went above 50%. The result is overall lower thermals, never going above 70C even with 192k source.
I thought everyone recommended the _amd build for latest gen Intel? Somewhat stumped but whatever! I recommend everyone running Ubuntu 22.04 on 14th gen Intel to at least give the intel_amd build a try.
Iāve not changed any specific features other than just enable Turbo Boost boot, but now is disabled and I donāt see any spikes of 100c, which are only for 2000ms. I guess Iām ok, just exaggerated a bit, not really experienced 
Thanks for the tips of the AIO, i think Iāll pass for the moment as I seems the noctua fan is not that bad.
I got over 80c when using DSD1024 with ul modulators and basic Nx filters. But it drops to 70ish and for brief I see bursts of 90C (2000ms). Now is less as I disabled the turbo boost on BIOS and leave instead to maximum.
If using DSD 512 is on 60Cs with standard settings and more demanding will jump to 70s.
I guess itās fine so far. Can keep as it is 
Agree with you! I also found the same results. Intel version is performing better. Specifically on intensive tasks
@jussi_laako I am assuming that based on prior naming conventions that the hqplayer-embedded-5.6.2-x64avx2 image has the āamdā build (formerly āavx2_amdā) and not āintel_amdā (formerly āamdā). The change in naming conventions is somewhat confusing 
Just to summarize my recent findings, and what @Moy also experienced, it seems like at least with the i9-14900, the avx2 optimized āamdā build has less overall load, presumably from those optimizations, but it can also results in dropouts. Whereas the more compatible āintel_amdā build has higher overall load but is free of dropouts (assuming you have cpu headroom of course).
fyi, if anyone is interested, I was able to get all of the core cpu temps to line up and output correctly in htop on Ubuntu 22.0.4. 3 things you need to do:
Result is each coreās cpu%, frequency, and temperature is shown in htop. I have mine configured with the hyper-threaded p-cores on left, e-cores on right, and overall cpu, frequency and package temp on top.
hqserver:~$ cat /etc/sensors.d/i9-14900
chip "coretemp-isa-0000"
label temp2 "Core 0"
label temp3 "Core 2"
label temp4 "Core 4"
label temp5 "Core 6"
label temp6 "Core 8"
label temp7 "Core 10"
label temp8 "Core 12"
label temp9 "Core 14"
label temp10 "Core 16"
label temp11 "Core 17"
label temp12 "Core 18"
label temp13 "Core 19"
label temp14 "Core 20"
label temp15 "Core 21"
label temp16 "Core 22"
label temp17 "Core 23"
label temp18 "Core 24"
label temp19 "Core 25"
label temp20 "Core 26"
label temp21 "Core 27"
label temp22 "Core 28"
label temp23 "Core 29"
label temp24 "Core 30"
label temp25 "Core 31"
Same things are offloaded regardless of the build. So if the E-cores load is higher with one build, then it means that the particular build takes more resources to do the same thing. Assuming same settings and same track.
The change aligns better. The build branded āintelā is only for Intel CPUs, primarily for older ones that lack AVX2 support. While the generic build is non-brand specific AVX2-only, just like for example the Fedora build, or the Windows build.
Yep thanks that is clear now. Unfortunately, there seems to be some issue with the amd avx2 build that causes dropouts on my i9-14900k, all else being equal with same settings and same track(s).
One other thing I did notice is that it even happens with 1x and Nx 48/96/192k source ā ASDM7EC-super @ DSD1024x48, but itās ok with 1x and Nx 44.1/176k ā ASDM7EC-super @DSD1024. i.e. across the board at 49.152M output with -super modulator. I can mitigate this with overclocking. None of that is necessary with intel build. If you need any more info happy to oblige.
with I7-4930K which is made in 2013, I was able to do DSD 256 for 44.1 and multiples (including 176), with all filters I was using for PCM except sinc-long. for 48K it downgrades to DSD 128. @jussi_laako HQPlayer is not that bad in consuming resources:-)
Then it is better to just use the Intel build. Nothing wrong with that. These CPUs are complex beasts and the results are not always predictable. I have a hunch why that happens, but not much I can do about. In your case it is once again a trade-off between two things. AMD architecture is not as susceptible to this particular behavior as Intel. The āul/light/super 512xfsā flavor seems to be even more into this category.
Wow, That is cool! Thank you for sharing!
As these temps were always āmessed upā in htop display I was always using i7z for monitoring. There it is per core not per threadā¦
Many thanks!
Also Iāve upgraded my Deskmeet ITX-B660 to a Sugo 16 with a Noctua NH-L12S, Itās a total game changer! no more thermals!
I know the case can fit a GPU! but Iām quite satisfied for the long run!
The empty Deskmeet case! More like CPU oven!
Sure here is my ~/.config/htop/htoprc.
# Beware! This file is rewritten by htop when settings are changed in the interface.
# The parser is also very primitive, and not human-friendly.
fields=0 48 17 18 38 39 40 2 46 47 49 1
sort_key=46
sort_direction=-1
tree_sort_key=46
tree_sort_direction=-1
hide_kernel_threads=1
hide_userland_threads=1
shadow_other_users=0
show_thread_names=1
show_program_path=0
highlight_base_name=0
highlight_megabytes=1
highlight_threads=0
highlight_changes=0
highlight_changes_delay_secs=3
find_comm_in_cmdline=1
strip_exe_from_cmdline=1
show_merged_command=0
tree_view=0
tree_view_always_by_pid=0
header_margin=1
detailed_cpu_time=0
cpu_count_from_one=0
show_cpu_usage=1
show_cpu_frequency=1
show_cpu_temperature=1
degree_fahrenheit=0
update_process_names=0
account_guest_in_cpu_meter=0
color_scheme=0
enable_mouse=1
delay=15
left_meters=CPU Blank LeftCPUs Blank Load Systemd NetworkIO
left_meter_modes=1 2 1 2 2 2 2
right_meters=Memory Blank RightCPUs Blank Hostname DateTime Uptime
right_meter_modes=1 2 1 2 2 2 2
hide_function_bar=0
Can sensors be added to the embedded image version? Iām getting jealous of all this data.
This was one reason I moved from the image. I wanted to add more monitoring tools. But there are lots of other things that also need to be redone with every update. Config save/reload, recreating ssh user, htop config, etc, etc.
Now that I have Ubuntu 22.04 running instead, I canāt believe I stuck with the image updates for so long! Installing updates takes 2 commands. Big bonus is that I was able to switch to the intel build of hqplayerd and get better overall performance as a result.
I would strongly suggest running embedded on a full OS install!
I have a script that I built to do all these things. I boot from USB stick so I just pop in a second stick, run my script, reboot, pull old stick and I have new version.