fra81

[fra81]

Unfortunately we’re not able to reproduce the issue with hardware encoding in our labs. Would you test one more thing? That would be to restore the EnableHardwareEncoding key back to the original value and change the encoder’s rate control mode instead (EncoderMode key), i.e.:

EnableHardwareEncoding 1

EncoderMode bitrate

[fra81]

This is strange indeed. Is it possible that the problem is with screen refresh and not with mouse and keyboard? Do you have any other means for accessing the server’s physical screen and verify?

You may try to check the ‘Use acceleration for display processing’ option in Server preferences -> Performance tab. You can do this before assigning the problematic card, then restart the server and try to connect.

Please also gather logs from the server as explained in https://www.nomachine.com/AR10K00697. You can send them to forum[at]nomachine[dot]com.

[fra81]

Hi,

I’m not sure I understand, do you have this problem only when the Mac is the client? Or also when the client is Linux or Windows?

I assume you are using an external keyboard on your Mac. Could you show a picture of the keyboard?

Aren’t you able to switch the NumLock status of the Linux/Windows server via the Clear key on your Mac keyboard?

[fra81]

Note that server and client side can either use the GPU or the CPU independently: server side can use the hardware encoder or a software fallback; client side can use the hardware decoder or a software fallback. And both sides can use the hardware (GPU) at the same time, or one of them can, or none. It will depend on hardware capabilities and on settings you’re using. To answer your question, when the GPU is not used (on either side), the CPU does the work.

That said, it seems there is a problem with hardware encoding. Please provide more info so we can investigate further:

– server OS version (as I understand Fedora 29)

– Nvidia card model

– video drivers type and version

Also server side logs would be useful. You can gather them as explained in https://www.nomachine.com/AR10K00697 and send them to forum[at]nomachine[dot]com, if you prefer.

[fra81]

Hi,

can I assume you already tried to use VP8 encoder and that solved the problem?

Please check if any of the following options solves the problem with H.264 (try them in this order):

1) disable hardware decoding by checking the ‘Disable client side hardware decoding’ option in the session menu, Display settings

2) disable hardware encoding on server side by adding this line to the /usr/NX/etc/node.cfg  file (then restart the server):

EnableHardwareEncoding 0

[fra81]

Hi Dario,

generally reducing image quality doesn’t decrease CPU usage, but reducing frame rate does. It is possible to set any value from 1  to 1000 by editing manually the node.cfg file on server side. For example, to have 10 frames per second, you can uncomment and set the following keys:

DisplayServerVideoFrameRate 10

DisplayServerUseVideoFrameRate 1

Another thing you can do is disabling the “lossless refinement” algorithm:

DisplayServerExtraOptions “-refinement 0”

[fra81]

Hi Dario,

at the moment NoMachine needs Nvidia Video Codec (NVENC) and that is not supported on Jetson.

There is work in progress to leverage OMX encoders and decoders in the future, mainly for mobile and embedded platforms, but I wouldn’t be able to say a date yet.

[fra81]

Please send us logs from the host machine as explained in https://www.nomachine.com/AR10K00697.

And client side logs as explained in https://www.nomachine.com/DT10O00163#2.3.

You can also send them to forum[at]nomachine[dot]com.

[fra81]

At the moment, the only way not to use all monitors is unplugging them physically from the server. Artifacts could be due to a bug in the H.264 software encoder with such a high horizontal resolution. Please try to use a different codec: open Server preferences -> Performance tab, check ‘Use a specific display encoding’ and make sure VP8 is selected, then restart the server and connect again.

[fra81]

Hi Juerg,

this problem is unknown. As a first try, please open Server preferences -> Performance tab on your host machine, and unselect ‘Use acceleration for display processing’.

[fra81]

You seem to have two 4K monitors, that give the total screen size reported in the logs:

    Info: Using screen size 7280×1080.

Unfortunately NVENC (Nvidia’s encoder) supports up to 4K, as you can see here:

https://developer.nvidia.com/nvidia-video-codec-sdk#NVENCFeatures

However this Feature Request, when implemented, will address your case:

https://www.nomachine.com/FR05N03113

For the records, you can see that the hardware encoder is used when there is a single monitor:

    Info: Using screen size 3840×1080.
    Info: Using Nvidia H.264 hardware encoder.

 

[fra81]

Hi,

can you please send all the logs. Whole ‘/usr/NX/var/log’ directory if possible. You can send it to forum[at]nomachine[dot]com.

[fra81]

15% of CPU usage doesn’t seem too much if the screen is updating. Are you using latest NoMachine versions on client and server? Recent versions always ship the H.264 encoder which minimizes CPU usage. Anyway you can check the used codec in the Display settings panel.

[fra81]

Hi,

for the moment NoMachine doesn’t have an option to select a specific GPU and NoMachine just iterates by default through available cards. However we’ve created a new feature request that will cover this case:

https://www.nomachine.com/FR07Q03854

[fra81]

Does setting max quality in https://www.nomachine.com/DT07M00087#5.7 make any difference? You can also try to check ‘Disable network-adaptive display quality’ and ‘Disable multi-pass display encoding’ in the same panel.

You may also try to disabled UDP. To do so, enter the Edit panel of the connection, click Advanced and uncheck ‘Use UDP communication for multimedia data’.

If nothing works, please send a video recording showing the issue.

[Author]

Posts