September 12, 2022 at 19:53 #40101
In February I bought a new laptop with an AMD Ryzen 5 PRO 4650U processor and started having a problem I never had before with Intel processors.
Using free version nxviewer version 7.10.2_8 on Windows 10 connection to nxserver version 7.9.2_1 on Linux Debian 11 (with a physical display), the response of the remote seems very sluggish. Not at all what I’m used to. I’ve tried upgrading and downgrading the NX versions with no effect.
It almost feels like a machine loading issue on the remote end, but the load average is typically < 1.0 , and it’s not always sluggish by the same amount. Sometimes it can take several seconds to react to a character typed into a console window or a button pressed in a graphical application. Other times it’s less than a second, but still enough to be noticeable.
For a while I thought it was related to multiple connect/disconnect cycles, perhaps due to some processes running on the Windows client being left running after disconnecting. It would get worse with each connect/disconnect until I rebooted the Windows machine, then it would get better for a while. Recently, however, that doesn’t seem to help much.
While researching this a few months ago, I think I saw a reference to issues with AMD graphics drivers, but I haven’t been able to locate that reference again.
I’d sure love to know what is causing this, because it’s making it almost painful to use NX any more. Should I be sorry I bought a laptop with an AMD processor?September 13, 2022 at 13:00 #40110
Your description indeed lets us think that it is a problem with the drivers, but we can’t be sure. Can you tell us if the AMD processor is on the Windows machine (the client) or the Debian one (the server)?
We suggest to try to disable hardware decoding on the client, and please also show us the CPU usage on client and server.
Looking forward to your answer.September 13, 2022 at 17:04 #40117
Thanks for your reply.
I’ll answer some of the questions, with a commitment to provide an update with CPU usage information soon (but maybe not today, a lot of things going on around here right now, and after a few changes described below and reboots, things are doing ok for the moment.)
Windows 10 Pro, 21H2 release with all updates applied
AMD Ryzen 5 Pro 4650U with Radeon graphics 2.10 GHz
16 GB RAM
Nomachine version 7.10.2
NX Server #1:
Debian 11 Bullseye fresh install (yesterday) fully updated OS and packages. XFCE4 desktop version 4.16.
Intel i3 2120 CPU @ 3.3 GHz
AMD Cedar [Radeon HD 5000/6000/7350/8350 Series] graphics @ 33 MHz clock
32 GB RAM
Updated to Nomachine version 7.10.1_1_amd64 yesterday
NX Server #2:
Debian 11 Bullseye fully updated OS and packages. XFCE4 desktop version 4.16.
Intel Core i7-3630QM @ 2.40 GHz
Intel 3rd Gen Core processor Graphics Controller
8 GB RAM
Updated to Nomachine version 7.10.1_1_amd64 yesterday
Both servers are connected via a Gigabit switch to a Netgear R6400 v2 router. The Windows client is connected to the servers via the router’s WiFi link (5Ghz). The wireless link is pretty lightly loaded. At the moment the Windows connection to NX server #1 is the only active link. I sometimes have a wireless link to an Amazon Fire Stick 4k and/or a Samsung Galaxy Tab 3 active on the link, but those typically don’t make much difference.
When I see the issue, it affects both of the Debian servers, which, along with the fact that the problem didn’t show up until I replaced my Laptop in February with an AMD processor based one, leads me to suspect something on the Windows end.
This morning I disabled hardware encoding on the Windows client, and rebooted. For the moment, the performance seems ok, but it sometimes takes a day or more after rebooting Windows before I start seeing the issue. Since I’ve made some changes, I’ll watch it for a while and provide CPU usage numbers if it shows up again. The Linux load averages should be easy enough to collect, but what form would you like to see the Windows numbers in? (I’m not much into Windows)
Thank you for your time.September 14, 2022 at 12:11 #40128
You can check the following link to find out more about collecting usage on Windows:
Is that what you are looking for, or did I misunderstand you ?
We look forward to your update on the CPU usage 🙂
Best regards,September 15, 2022 at 19:12 #40159
Yes, the link you provided gives me an idea of what needs to be done.
It’s been a couple days now since I disabled hardware decoding on the client and I haven’t noticed the sluggishness I had earlier. I’ve attached CPU usage for the client and server #1, which is the one I use most.
I think the next step will be to re-enable hardware decoding and see if the sluggishness comes back. If so, I’ll update with new CPU usage info.
Attachments:September 16, 2022 at 09:57 #40163
One of your files was too big and couldn’t be uploaded (“CPU-Usage-hw-decode-disabled.png”).
Can you please send it by email to forum[at]nomachine[dot]com, making sure to reference the topic in subject ?
Thank you!September 16, 2022 at 11:10 #40169sil04Keymaster
may you please tell us also which is the driver version of graphic card on the client?September 16, 2022 at 13:26 #40170
I emailed the file as requested. As I mentioned in the email, I don’t know for sure if this is what you’re looking for. It’s a screen shot of the CPU utilization during a time that the nomachine connection was active. If you were looking for individual processes that were running at that time, I’ll have to send something else.
Here is the information of the client graphics driver:
Driver Provider: Advanced Micro Devices, Inc.
Driver Date: 12/29/2021
Driver Version: 30.0.13040.5001
Driver Signer: Microsoft Windows Hardware Compatability PublisherSeptember 23, 2022 at 14:15 #40311BritgirlKeymaster
I think the next step will be to re-enable hardware decoding and see if the sluggishness comes back. If so, I’ll update with new CPU usage info
Ideally, we would need a screen of the Details tab of the Task Manager. If you reproduce it again, submit that.
You must be logged in to reply to this topic.