HQ Player

[Miska]

Some people use optical isolation ... it may be better solution than switch.

 

Yes optical is good choice and one reason I created NAA in first place. My switch has module slots for two these kind (SFP) of plug-in optical interface modules:

https://www.hpe.com/us/en/product-catalog/options/pip.transceivers.3619488.html

https://www.verkkokauppa.com/fi/product/27020/tnqf/HPE-X121-1G-SFP-LC-SX-Transceiver-moduuli

[tboooe]

Yes optical is good choice and one reason I created NAA in first place. My switch has module slots for two these kind (SFP) of plug-in optical interface modules:

https://www.hpe.com/us/en/product-catalog/options/pip.transceivers.3619488.html

https://www.verkkokauppa.com/fi/product/27020/tnqf/HPE-X121-1G-SFP-LC-SX-Transceiver-moduuli

I have a managed switch that supports fiber (Cisco SG300) as well as a fiber NIC in my server pc running HQP and Roon. Unfortunately, my little low powered pc acting as NAA does not support adding a separate NIC card so I have to use an FMC between the switch and NAA pc. The sound with the server and NAA connected to the switch is definitely different than with the NAA and server connected directly.

[bogi]

Miska, I run the things you asked for on my Xenial NAA. As you can see the networkaudiod does not see my XMOS based DAC! The DAC was powered up and connected.

 

login as: bogi
bogi@fe80::8639:beff:fe63:e9c3%19's  password:
Welcome to Ubuntu 16.04.2 LTS (GNU/Linux 4.4.52-jl+ x86_64)

 * Documentation:  [url]https://help.ubuntu.com[/url]
* Management:      [url]https://landscape.canonical.com[/url]
*  Support:        [url]https://ubuntu.com/advantage[/url]

0 packages can be updated.
0 updates are security updates.


Last login: Tue Mar  7 20:42:34 2017
bogi@NAA:~$ systemctl status networkaudiod
●  networkaudiod.service - Network Audio Adapter daemon
  Loaded: loaded  (/lib/systemd/system/networkaudiod.service; enabled; vendor pr
  Active:  active (running) since Tue 2017-03-07 20:46:29 CET; 55s ago
Main PID: 1334  (networkaudiod)
   Tasks: 1
  Memory: 252.0K
     CPU: 48ms
   CGroup: /system.slice/networkaudiod.service
          └─1334  /usr/sbin/networkaudiod

Mar 07 20:47:15 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso
Mar 07 20:47:16 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso
Mar 07 20:47:17 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso
Mar 07 20:47:18 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso
Mar 07 20:47:19 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso
Mar 07 20:47:20 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso
Mar 07 20:47:21 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso
Mar 07 20:47:22 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso
Mar 07 20:47:23 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso
Mar 07 20:47:24 NAA networkaudiod[1334]: [/usr/sbin/networkaudiod] (1334): clSocket::SetOption(): setso


lines 1-20/20  (END)
bogi@NAA:~$ sudo bash
[sudo]  password for bogi:
root@NAA:~# ps -ef | grep  netw
root      1395     1  0 20:49 ?        00:00:00  /usr/sbin/networkaudiod
root      1411  1400  0 20:50 pts/0    00:00:00 grep  --color=auto netw
root@NAA:~# kill -9  1395
bash: kill: (1395) - No such process
root@NAA:~# ps -ef | grep netw
root      1491  1400  0 21:02 pts/0    00:00:00 grep --color=auto netw
root@NAA:/usr/sbin# systemctl stop networkaudiod
root@NAA:/usr/sbin# ps -ef | grep netw
root      1500  1400  0 21:03 pts/0    00:00:00 grep --color=auto netw
root@NAA:/usr/sbin# ./networkaudiod
[./networkaudiod] (1501): networkaudiod Copyright (C) 2011-2017 Jussi Laako / Signalyst. All rights reserved.
[./networkaudiod] (1501): asoundlib version: 1.1.0
[./networkaudiod] (1501): clSocket::SetOption(): setsockopt(): No such device
[./networkaudiod] (1501): clSocket::SetOption(): setsockopt(): No such device
[./networkaudiod] (1501): clSocket::SetOption(): setsockopt(): No such device
[./networkaudiod] (1501): clSocket::SetOption(): setsockopt(): No such device
[./networkaudiod] (1501): clSocket::SetOption(): setsockopt(): No such device
^C[./networkaudiod] (1501): ALSA backend uninitialized
root@NAA:/usr/sbin#

[bogi]

Miska, the release version crashes in the same way as beta: My IPv6 NAA is not found (it probably searched for my Win10 NAA), Setup dialog opens automatically. As I click OK in that dialog the crash occurs afterwards.

[Miska]

Miska, I run the things you asked for on my Xenial NAA. As you can see the networkaudiod does not see my XMOS based DAC!

 

It doesn't seem to get that far... It doesn't really care about DAC before HQPlayer is talking to it.

 

[./networkaudiod] (1501): clSocket::SetOption(): setsockopt(): No such device

 

This error happens when the network interface is not properly up.

 

When you run "ifconfig" the ethernet interface statistics should show following line:

"UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1"

 

It should also show address for both "inet" and "inet6".

[blue2]

... Also for networking and USB gear, improving the clocks used on those make the noise signature worse. Best is to use one with spread spectrum clocking, where jitter is added on purpose to produce less spurious tones. It is good because none of these clocks have anything to do with audio timing.

One good example why the original M2Tech hiFace is still one of the best USB-to-S/PDIF converters because it doesn't have strict packet timing interval, unlike USB Audio Class devices. Main problem with USB Audio Class devices is the static 8 kHz packet interval that tends to pop up at measurements, especially for USB bus powered devices. When that interval becomes less deterministic the level of the 8 kHz spurious tone drops.

 

You're a mine of useful information! I didn't know the M2Tech hiFace was so advanced for its time. I still have one. I believe there will soon be a new USB cleaner on the market using this principle (decorrelated noise), which I thought would be a first.

[bogi]

@Miska:

 

root@NAA:/usr/sbin# ifconfig
enp1s0    Link encap:Ethernet  HWaddr 84:39:be:63:e9:c3
         inet6 addr: fe80::8639:beff:fe63:e9c3/64 Scope:Link
         UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
         RX packets:794 errors:0 dropped:0 overruns:0 frame:0
         TX packets:2868 errors:0 dropped:0 overruns:0 carrier:0
         collisions:0 txqueuelen:1000
         RX bytes:98845 (98.8 KB)  TX bytes:348933 (348.9 KB)

lo        Link encap:Local Loopback
         inet addr:127.0.0.1  Mask:255.0.0.0
         inet6 addr: ::1/128 Scope:Host
         UP LOOPBACK RUNNING  MTU:65536  Metric:1
         RX packets:9440 errors:0 dropped:0 overruns:0 frame:0
         TX packets:9440 errors:0 dropped:0 overruns:0 carrier:0
         collisions:0 txqueuelen:1
         RX bytes:698560 (698.5 KB)  TX bytes:698560 (698.5 KB)

 

I am connected with PuTTY, can do WinSCP, so the IPv6 interface is functioning.

 

Should I enable IPv4 too? My interfaces file is:

 

root@NAA:/usr/sbin# cat /etc/network/interfaces
# This file describes the network interfaces available on your system
# and how to activate them. For more information, see interfaces(5).

source /etc/network/interfaces.d/*

# The loopback network interface
auto lo
iface lo inet loopback
auto enp1s0
iface enp1s0 inet6 auto

[Jud]

You're a mine of useful information!

 

Funny how that works.

[bogi]

You're a mine of useful information! I didn't know the M2Tech hiFace was so advanced for its time. I still have one. I believe there will soon be a new USB cleaner on the market using this principle (decorrelated noise), which I thought would be a first.

 

I was using hiFace2 about a year with Harman Kardon HD970 CD player, which had also coax input and could be used as 192k capable DAC. Then I bought Gustard U10 digital USB converter and I found it was significantly better than hiFace2. The problem with hiFace2 is weak isolation from USB noise. It's hard to expect much in this area from such a small device. Therefore those battery mods etc. could have such big effect. Gustard U10 used internal LPS.

[Miska]

Should I enable IPv4 too?

 

Yes, NAA is either dual-stack or IPv4-only...

[Miska]

I was using hiFace2 about a year with Harman Kardon HD970 CD player, which had also coax input and could be used as 192k capable DAC. Then I bought Gustard U10 digital USB converter and I found it was significantly better than hiFace2. The problem with hiFace2 is weak isolation from USB noise. It's hard to expect much in this area from such a small device. Therefore those battery mods etc. could have such big effect. Gustard U10 used internal LPS.

 

Important to note that hiFace2 is USB Audio Class device (XMOS-based), it is using completely different protocol on USB than the hiFace... Meaning that hiFace2 is vulnerable to this 8 kHz packet noise.

 

By the way, exaSound DACs go to the same category as the first hiFace. Not using USB Audio Class but a different protocol.

 

P.S. Apart from original hiFace, there was of course hiFace Evo, Young and Vaughan using the same interface. Plus bunch of third party DACs with the OEM interface part.

[bogi]

Yes, NAA is either dual-stack or IPv4-only...

 

 

 

OK, that I didn't know. I had enabled IPv4 with Win10 NAA but I left it set for DHCP (although none is provided) and it functioned with HQPlayer computer which was set up in the same way.

 

I enabled IPv4 and tried both the dhcp option and static IP address option. But no luck. This is for the static IP case:

 

root@NAA:/usr/sbin# cat /etc/network/interfaces
# This file describes the network interfaces available on your system
# and how to activate them. For more information, see interfaces(5).

source /etc/network/interfaces.d/*

# The loopback network interface
auto lo
iface lo inet loopback
auto enp1s0
iface enp1s0 inet static
 address 192.168.1.1
 gateway 192.168.1.1
 netmask 255.255.255.0
 network 192.168.1.0
 broadcast 192.168.1.255
iface enp1s0 inet6 auto

root@NAA:/usr/sbin# /etc/init.d/networking restart
[....] Restarting networking (via systemctl): networking.serviceJob for networking.service failed because the control process exited with error code. See "systemctl status networking.service" and "journalctl -xe" for details.
failed!
root@NAA:/usr/sbin#
root@NAA:/usr/sbin# /etc/init.d/networking restart
[....] Restarting networking (via systemctl): networking.serviceJob for networking.service failed because the control process exited with error code. See "systemctl status networking.service" and "journalctl -xe" for details.
failed!
root@NAA:/usr/sbin# ifconfig
enp1s0    Link encap:Ethernet  HWaddr 84:39:be:63:e9:c3
         inet addr:192.168.1.1  Bcast:192.168.1.255  Mask:255.255.255.0
         inet6 addr: fe80::8639:beff:fe63:e9c3/64 Scope:Link
         UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
         RX packets:1973 errors:0 dropped:0 overruns:0 frame:0
         TX packets:3763 errors:0 dropped:0 overruns:0 carrier:0
         collisions:0 txqueuelen:1000
         RX bytes:244807 (244.8 KB)  TX bytes:512003 (512.0 KB)

lo        Link encap:Local Loopback
         inet addr:127.0.0.1  Mask:255.0.0.0
         inet6 addr: ::1/128 Scope:Host
         UP LOOPBACK RUNNING  MTU:65536  Metric:1
         RX packets:16000 errors:0 dropped:0 overruns:0 frame:0
         TX packets:16000 errors:0 dropped:0 overruns:0 carrier:0
         collisions:0 txqueuelen:1
         RX bytes:1184000 (1.1 MB)  TX bytes:1184000 (1.1 MB)

root@NAA:/usr/sbin# ping 192.168.1.2
PING 192.168.1.2 (192.168.1.2) 56(84) bytes of data.
64 bytes from 192.168.1.2: icmp_seq=1 ttl=128 time=0.206 ms
64 bytes from 192.168.1.2: icmp_seq=2 ttl=128 time=0.338 ms
64 bytes from 192.168.1.2: icmp_seq=3 ttl=128 time=0.492 ms
^C
--- 192.168.1.2 ping statistics ---
3 packets transmitted, 3 received, 0% packet loss, time 2000ms
rtt min/avg/max/mdev = 0.206/0.345/0.492/0.117 ms
root@NAA:/usr/sbin# ./networkaudiod
[./networkaudiod] (1919): networkaudiod Copyright (C) 2011-2017 Jussi Laako / Signalyst. All rights reserved.
[./networkaudiod] (1919): asoundlib version: 1.1.0
[./networkaudiod] (1919): clSocket::SetOption(): setsockopt(): No such device
[./networkaudiod] (1919): clSocket::SetOption(): setsockopt(): No such device
[./networkaudiod] (1919): clSocket::SetOption(): setsockopt(): No such device

 

While networkaudiod running I tried connect from HQPlayer Desktop again, but NAA wasn't seen both for IPv4 and IPv6.

 

For all cases I substituted my DIY USB cable with Supra and I tried also other USB port on my miniPC, but no luck.

 

I am old UNIXman from 20 years ago but last 10 years I used UNIX/Linux only little. I'm not familiar with audio related things under Linux ... I'd like to check if USB class 2 audio (ALSA on Linux?) is working at all. Is there some easy way available how to check it?

[Margus]

Is there some easy way available how to check it?

 

margus@NAA:~$ uname -a
Linux NAA 4.4.52-jl+ #12 SMP PREEMPT Thu Mar 2 11:29:46 EET 2017 x86_64 x86_64 x86_64 GNU/Linux
margus@NAA:~$ cat /etc/network/interfaces
# This file describes the network interfaces available on your system
# and how to activate them. For more information, see interfaces(5).

source /etc/network/interfaces.d/*

# The loopback network interface
auto lo
iface lo inet loopback

# The primary network interface
auto enp0s25
iface enp0s25 inet static
       address 192.168.10.8
       netmask 255.255.255.0
       gateway 192.168.10.1
       dns-nameservers 8.8.8.8
margus@NAA:~$ps ax | grep networkaudiod
1108 ?        S<s    0:00 /usr/sbin/networkaudiod

My pure Linux based trace

 

HQP@i5/Kubuntu16.04.2 --> switch@home network -> NAA@NUC/Ubuntu server -> DIY DSC-1 -> DSD256

[tipunch]

Hello Miska,

I always have this error when I quit HQPlayer Desktop 3.15.1. It is not too grave, but if you can correct it would be better (Hum, I am on OS X Mavericks 10.9.1)

 

Best regards

 

 

error.txt

[bogi]

My pure Linux based trace

 

Thanks Margus, my network is functioning, but it seems audio need some configuration:

 

root@NAA:~# cat /proc/asound/cards
1 [X20            ]: USB-Audio - XMOS USB Audio 2.0
                     XMOS XMOS USB Audio 2.0 at usb-0000:00:14.0-3, high speed
root@NAA:~# aplay -l
**** List of PLAYBACK Hardware Devices ****
card 1: X20 [XMOS USB Audio 2.0], device 0: USB Audio [uSB Audio]
 Subdevices: 1/1
 Subdevice #0: subdevice #0
root@NAA:~# speaker-test -t pink

speaker-test 1.1.0

Playback device is default
Stream parameters are 48000Hz, S16_LE, 1 channels
Using 16 octaves of pink noise
Sample format not available for playback: Invalid argument
Setting of hwparams failed: Invalid argument
root@NAA:~# aplay test.wav
Playing WAVE 'test.wav' : Signed 24 bit Little Endian in 3bytes, Rate 96000 Hz, Stereo
aplay: set_params:1233: Sample format non available
Available formats:
- S32_LE
root@NAA:~#

[bogi]

I performed as root:

adduser root audio

adduser bogi audio

 

To set default playback device I created /etc/asound.conf and .asoundrc in /home/bogi with content

 

pcm.!default { type plug slave { pcm "hw:1,0" } } ctl.!default { type hw card 1 }

 

Then I could play a WAV file via aplay, it played really through my USB DAC:

 

root@NAA:/usr/sbin# aplay /home/bogi/test.wav
Playing WAVE '/home/bogi/test.wav' : Signed 24 bit Little Endian in 3bytes, Rate 96000 Hz, Stereo

And speaker-test also plays to my DAC:

 

root@NAA:/usr/sbin# speaker-test -t pink

speaker-test 1.1.0

Playback device is default
Stream parameters are 48000Hz, S16_LE, 1 channels
Using 16 octaves of pink noise
Rate set to 48000Hz (requested 48000Hz)
Buffer size range from 16 to 131072
Period size range from 8 to 65536
Using max buffer size 131072
Periods = 4
was set period_size = 32768
was set buffer_size = 131072
0 - Front Left
Time per period = 0.014993
0 - Front Left
Time per period = 2.732622
0 - Front Left
Time per period = 2.730580
0 - Front Left

 

 

But networkaudiod is still not visible for HQPlayer Desktop ...

[Miska]

I always have this error when I quit HQPlayer Desktop 3.15.1. It is not too grave, but if you can correct it would be better (Hum, I am on OS X Mavericks 10.9.1)

 

At the moment HQPlayer only works on macOS 10.11 and newer. The overall rule is latest and one before. This is also what I can test on my side. I know for sure that Nvidia's CUDA 8 stack explodes on anything older than 10.10.

[Miska]

OK, that I didn't know. I had enabled IPv4 with Win10 NAA but I left it set for DHCP (although none is provided) and it functioned with HQPlayer computer which was set up in the same way.

 

On Windows protocols are enabled for the interface as long as the check box is ticked on the protocol list in network adapter properties.

 

[./networkaudiod] (1919): clSocket::SetOption(): setsockopt(): No such device

 

While networkaudiod running I tried connect from HQPlayer Desktop again, but NAA wasn't seen both for IPv4 and IPv6.

 

As long as you are getting this error it won't be seen because it has not been able to subscribe to the multicast packets... It is just running in loop for a while waiting for network to become available.

 

I suspect most OS by default assign the interface where default route is going.

[Miska]

But networkaudiod is still not visible for HQPlayer Desktop ...

 

You need to get it running cleanly first, audio devices won't matter until HQPlayer can talk to the networkaudiod in first place...

 

Just in case, did you install the Xenial networkaudiod package? Mixing Stretch packages to Xenial may result in funny behavior.

 

Try starting networkaudiod in IPv4 only mode to see if the problem is related to IPv6 by setting NETWORKAUDIOD_IPV6=0 environment.

 

For example you can just start networkaudiod with "NETWORKAUDIOD_IPV6=0 /usr/sbin/networkaudiod".

[bogi]

Just in case, did you install the Xenial networkaudiod package? Mixing Stretch packages to Xenial may result in funny behavior.

Xenial for sure, according to OS.

 

For example you can just start networkaudiod with "NETWORKAUDIOD_IPV6=0 /usr/sbin/networkaudiod".

I tried that, but no change for IPv4 only. HQPlayer Desktop sees no device when NetworkAudioAdapter backend is selected.

 

Did you ever try direct cable config?

[Superdad]

Important to note that hiFace2 is USB Audio Class device (XMOS-based), it is using completely different protocol on USB than the hiFace... Meaning that hiFace2 is vulnerable to this 8 kHz packet noise.

 

By the way, exaSound DACs go to the same category as the first hiFace. Not using USB Audio Class but a different protocol.

 

P.S. Apart from original hiFace, there was of course hiFace Evo, Young and Vaughan using the same interface. Plus bunch of third party DACs with the OEM interface part.

 

Ah yes, the benefit of the nightmare that is USB bulk mode. Turns DAC makers into perpetual software developers and turns their users against them.

We all should remember why M2Tech moved away from bulk mode....

 

And did exaSound ever release a Linux driver?

 

There are ways to deal with the 8kHz packet data rate issue. It's not that big a problem.

 

But I ageee that we would benefit from competition to XMOS. Both their base code and their development tools leave a lot to be desired. Of course Cmedia and Tenor (are they still around?) are not particularly compelling alternatives. Would be nice if there were a selection of open source UAC2-compliant USB protocol cores for Altera or other nice FPGA families.

 

Yet I just don't see bulk mode making a comeback...

 

Sorry for the off topic. You know I am a big HQP fan.

[Miska]

Ah yes, the benefit of the nightmare that is USB bulk mode.

 

There's no nightmare, it is great! Just compare amount of source code in the hiFace Linux driver and UAC driver. The hiFace driver is like 1/10th of the UAC driver.

 

And the Linux driver still works very well, I just recently fixed one bug from it. One of the great things with Linux, I could do it!

 

Turns DAC makers into perpetual software developers and turns their users against them.

 

They should be anyway, because they should be providing ASIO drivers for Windows too. To properly deal with firmware and drivers you need to employ software engineers. I think this will be one of the major difference makers between manufacturers in future. The ones who employ software engineers vs the ones who don't...

 

There is no way to avoid dealing with software in modern world.

 

We all should remember why M2Tech moved away from bulk mode....

 

Because of Apple. Apple loves to break other people's software. HQPlayer as suffered couple of times from this too.

 

If people insist getting UAC devices and it is less work for manufacturer that way, it may be easier to give them what they ask?

 

And did exaSound ever release a Linux driver?

 

No, but that has nothing to do with bulk mode.

 

There are ways to deal with the 8kHz packet data rate issue. It's not that big a problem.

 

Yes, best is to forget stupid USB and use Ethernet instead. Like I said, bake NAA straight into the DAC and use bare I2S/DSD data wires from there. No USB.

 

But I ageee that we would benefit from competition to XMOS.

 

There are many non-XMOS implementations. Amanero, TEAC, Marantz/Denon, Fostex, etc.

 

Would be nice if there were a selection of open source UAC2-compliant USB protocol cores for Altera or other nice FPGA families.

 

Why use FPGA to host MCU when you can have MCU doing the job? It is just inefficient.

 

Yet I just don't see bulk mode making a comeback...

 

It comes back, but is just called Ethernet...

[louisxiawei]

Hi Miska,

 

Can you please make any comment for recently released AMD Ryzen 1700 8-core CPU? What's your opinion of it against intel 7700k (processing speed, power consumption, etc)?

 

However, from the userbenchmark website, 7700k's single-core core and quad-core are much faster than AMD 1700's.

 

I plan to run HQplayer for stereo system DSD-512 upsampling, which one will be the best choice? Will HQplayer take advantage of AMD 1700's 8-core processing speed for a stereo system?

[Miska]

Did you ever try direct cable config?

 

No...

 

I've tried with IPv6 auto config in my network and it works, but it could be because I have full IPv6 internet access from the network too.

 

It is intended to work in a regular home network. Same network that is commonly used to connect other home devices like TV's, iPad's, IoT stuff, NAS, etc.

[Miska]

Can you please make any comment for recently released AMD Ryzen 1700 8-core CPU? What's your opinion of it against intel 7700k (processing speed, power consumption, etc)?

 

Only once I've got a chance to try it out myself...

 

I'm planning to get 1800X.

 

However, from the userbenchmark website, 7700k's single-core core and quad-core are much faster than AMD 1700's.

 

There's quite a bit of clock speed difference, 7700K base clock is 4.2 GHz.

 

I plan to run HQplayer for stereo system DSD-512 upsampling, which one will be the best choice? Will HQplayer take advantage of AMD 1700's 8-core processing speed for a stereo system?

 

There is a limit how much extra cores can be utilized, depending on use case. So if you don't run convolution engine 7700K is probably safer choice for your use case. Because base clock is usually related to number of cores, rule of thumb CPU selection for upsampling without convolution or such is to get CPU with high as possible base clock frequency and with 2xChannels number of cores (so for stereo a quad core). If you run convolution or Roon you can throw in two extra cores so that the processing doesn't starve due to other activities. So for stereo 4 to 6 cores is usually optimal point. The 10-core 6950X I have is for running 8-channels with convolution at DSD256 to exaSound e28 DAC.

 

I have added some new things to make next HQPlayer version utilize a bit more cores so that it would be better on those 8 cores in basic cases too. So that it would behave a bit more like it now does with Nvidia CUDA offload.