Performance throttle on Ryzen 5-3550H+AMD RX560X laptop after some use

Technical Issues and Assistance Hardware
, ,
#1

I recently bought an Asus TUF FX705DY laptop with a Ryzen 5 3550H CPU and AMD RX560X GPU, I wanted to try an all AMD laptop as my previous Intel+Nvidia laptop had given me nothing but trouble on Linux.

Installation was mostly trouble-free (I had to use a Manjaro Testing live image as this cpu apparently needs kernel 5.0 or higher) but I'm noticing a strange phenomenon: The laptop runs fine after booting, all applications run smoothly, but if I run a graphics heavy app (e.g. Blender, Steam games) the whole laptop seems to grind to a halt.

Specific example: I do some video editing in Lightworks, no stuttering, cpu load during playback ~15%. Then, play a game on Steam - starts off at a satisfactory framerate of 40-45 fps, then after a few seconds slows down to 10-15 fps. Close game, exit Steam, run Lightworks again. Playback is now stuttering and cpu load is ~80-85% for the same task as before. Other applications seem to be struggling as well. Only fix is a reboot.

I'm by no means a Linux expert and don't know where to start troubleshooting. Can anyone offer some advice?

My inxi -Fxzc0 output is as follows:

Thanks for any tips

[Testing Update] 2020-02-25 - Kernels, Pamac 9.4-rc1, Firefox-Dev, Cinnamon, Calamares
#2

i dont have ryzen/amd hardware but there are threads here on the forum dedicated to ryzen problems/fixes . from what i can tell, from reading anyway is that most if not all require some special kernel parameters, either for zenstates issues, lockups, etc... but apparently they've been getting better and better with each kernel release so maybe try starting there with 5.2, (5.1 i guess is having some issues with fstrim trimming more blocks than it should, only briefly read into it)

#3

https://wiki.archlinux.org/index.php/Improving_performance

generally
set cpu governor to performance.
uninstall tlp
set swappiness to 5
vm cache pressure to 50
enable haveged
note: no battery power saving you get full
performance.

#4

Any tips on how to adjust all those settings? As mentioned, I'm not an expert and Google left me none the wiser (obviously I can figure out how to uninstall tlp).

I also tried disabling C6 state as mentioned in this thread, but nothing improved.

#5

set cpu governor to performance by editing /etc/default/cpupower
and uncommenting and editing
it should be like this
#governor=ondemand
change it to
governor=performance
then save it.
and enabling cpupower service

systemctl enable cpupower

to chamge swappiness

echo vm.swappiness=5 | sudo tee -a /etc/sysctl.d/99-sysctl.conf 
echo vm.vfs_cache_pressure=50 | sudo tee -a /etc/sysctl.d/99-sysctl.conf 
sudo sysctl -p /etc/sysctl.d/99-sysctl.conf

also uninstall tlp or it will reduce the cpu frequency

#6

This is incorrect. TLP does not reduce CPU frequency.

You've gone back to providing blanket suggestions rather than asking questions to understand the problem.

Watch this:

Can you check the output of sensors to see whether this might be a thermal issue, and the output of free (or free-m) and top to check resource usage.

You can do this before and after the issue occurs.

#7

Thanks to both of you. I will post the outputs later when I get home.

#8

OK, so I ran my test. Outputs after a fresh boot:

sensors
amdgpu-pci-0500
Adapter: PCI adapter
vddgfx:           N/A  
vddnb:            N/A  
temp1:        +39.0°C  (crit = +80.0°C, hyst =  +0.0°C)

acpitz-acpi-0
Adapter: ACPI interface
temp1:        +39.0°C  (crit = +120.0°C)

amdgpu-pci-0100
Adapter: PCI adapter
vddgfx:       +0.71 V  
temp1:        +31.0°C  (crit = +94.0°C, hyst = -273.1°C)
power1:        6.00 W  (cap =  48.00 W)

k10temp-pci-00c3
Adapter: PCI adapter
Tdie:         +39.2°C  (high = +70.0°C)
Tctl:         +39.2°C

free 
              total        used        free      shared  buff/cache   available
Mem:           7722        2529        3573          75        1620        4825
Swap:          8191           0        8191

Running a mildly GPU-intensive game, laptop ground to a crawl after a few seconds:

sensors
amdgpu-pci-0500
Adapter: PCI adapter
vddgfx:           N/A  
vddnb:            N/A  
temp1:        +51.0°C  (crit = +80.0°C, hyst =  +0.0°C)

acpitz-acpi-0
Adapter: ACPI interface
temp1:        +51.0°C  (crit = +120.0°C)

amdgpu-pci-0100
Adapter: PCI adapter
vddgfx:       +0.72 V  
temp1:        +44.0°C  (crit = +94.0°C, hyst = -273.1°C)
power1:        6.00 W  (cap =  48.00 W)

k10temp-pci-00c3
Adapter: PCI adapter
Tdie:         +51.8°C  (high = +70.0°C)
Tctl:         +51.8°C  


free
              total        used        free      shared  buff/cache   available
Mem:           7722        1478        2605          20        3638        5921
Swap:          8191           1        8190

Nothing seems out of the ordinary. I must assume it's an issue with a bleeding edge CPU. Hopefully future kernel updates will fix it.

1 Like
#9

Yup.

You might try the 5.2-rc kernel to see if it helps, though that is very much pre-release and might have other issues.

#10

Owner of the same laptop, and having some of the same symptoms.
And yes, both Asus and AMD have some patches incoming on 5.2, it's still very early on development to really warrant any testing on it, update at your own risk.

1 Like
#11

I installed the 5.2 testing kernel but it's unusable as it was impossible to install the wifi driver. I'll wait.

#12

Install the headers maybe? Just tried it and it works (currently replying from 5.2)

EDIT:
That being said, I still don't recommend running that kernel at all, from boot and by only doing the same web-browsing I always do I had a higher fan activity than on 5.1 meaning that there are some performance kinks here and there (expected from a RC kernel)

#13

5.2-rc1 has been a bit of a clunker for me. My wireless USB mouse doesn't work at all, and 5.2-rc1 takes forever to either boot or shutdown. 5.0-rc and 5.1-rc both worked well from the beginning. 5.2 might have to go to rc8 or 9 before it's ready :scream:.

#14

I'm just the opposite. Having no problems with 5.2rc1...humming right along.

#15

Guess, I'll have to try rc2 (it just came out.)

Edit: 5.2-rc2 has same issues for me, I'll stick to 5.1 for now. YMMV

Edit2: Added what should have been an unrelated new cstate baytrail work-around and my mouse started working again.

Edit3: rc5 boots and shuts down properly (quickly) and it seems to be economical with RAM (as is 5.1)

1 Like
#16

I'd say this looks a lot like thermal throttling. If you do a cat /proc/cpuinfo while under load, what do you see?

If you haven't already, I would apply the latest BIOS update. I think Asus released one earlier this month.

#17

Thanks, I'll check that when I get off work. However, you can see my output from sensors further up and the CPU temp during throttling is far below anything critical.

I even did a test with booting into Windows, running some GPU-heavy games for an extended period of time, then immediately booted into Manjaro to see if the higher temp would trigger throttling. Seemed to run as normal.

Seems to me like a bug in the kernel thinks the CPU temp becomes critical under any significant load and shuts it down.

#18

I can confirm that neither a BIOS update nor kernel 5.2 has fixed the issue for me. Kernel 5.2 also brought additional stability issues related to sleep and random freezes.

I was starting to think I bought a defective laptop, but everything seems to be running perfectly under Windows.

#19

my old boss used to say 'it's nice to be at the leading edge but beware the bleeding edge'
good luck!

1 Like
#20

There's hope on the horizon, looks like we'll have to wait until kernel 5.3 though:
https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.3-Platform-x86-Updates

Forum kindly sponsored by