Overdrive not working on Polaris since 4.17

[Mel Spektor]

Overclocking works with the AMD drm-next-4.21-wip kernel and RX560.

root@ryzenpc:/home/xfce# echo manual > /sys/class/drm/card0/device/power_dpm_force_performance_level
root@ryzenpc:/home/xfce# cat /sys/class/drm/card0/device/power_dpm_force_performance_level
manual
root@ryzenpc:/home/xfce# echo 2 > /sys/class/drm/card0/device/pp_dpm_mclk
root@ryzenpc:/home/xfce# cat /sys/class/drm/card0/device/pp_dpm_mclk
0: 300Mhz
1: 625Mhz
2: 1750Mhz *

that's not how it works pp_dpm_mclk should increase the clock above 1750 by 2%, it's not working either. check

cat /sys/kernel/debug/dri/0/amdgpu_pm_info

[Mel Spektor]

1750 is the stock clock overdrive is specified in the bios it could go up to 2000mhz.

[Mel Spektor]

Where that is documented, max memory clock is 1750MHz, see https://videocardz.net/asus-radeon-rx-560-2gb-dual-fan/

Overclocking low end GPU cards has no benefits, only more heat is produced and heat will shorten the lifetime of microchips.. You will get more speed by using using a non debug 1000Hz timer kernel and do not not use the CPU frequency scaling feature of the Linux kernel. Use the bios of the motherboard to control the CPU.

pretty sure that was not what happened with 4.16. And yes it has benefits in gaming and mining.

[Mel Spektor]

look here https://www.maketecheasier.com/overclock-amd-gpu-linux/

[Mel Spektor]

I also tried the way is documented in the archwiki ( I think that's more similar than what kernel documentation says)

https://wiki.archlinux.org/index.php/AMDGPU but when I commited the changes

 echo "c" > /sys/class/drm/card0/device/pp_od_clk_voltage

the card loose signal and I had to reset the bios to get signal again, so yes it's bugged .

[Mel Spektor]

one question since the kernel fusion driver has been upstreamed the roc-smi utility is like the "official" amd app for overclocking, right?
because that's not working as well.

(or to use the roc smi utility you need the full rocM stack?)

[Mel Spektor]

maybe Michael can investigate this or bridgman can clarify? I would say at least that behaviour has changed, I'm not completely sure if it's broken, I might be missing something.

[Mel Spektor]

Ok I found the way! finally great is working. Archwiki is right first you have to edit the pstates and also you have to boot with
amdgpu.ppfeaturemask=0xffffffff Clock speeds, voltages are supposed to change via overriding GPU's P-states in /sys/class/drm/card0/device/pp_od_clk_voltage and committing
it with "c".

first check the pstates

# cat /sys/class/drm/card0/device/pp_od_clk_voltage  
OD_SCLK:
0:        214MHz        800mV
1:        481MHz        821mV
2:        760MHz        825mV
3:       1020MHz        900mV
4:       1102MHz        981mV
5:       1138MHz       1025mV
6:       1172MHz       1068mV
7:       1212MHz       1125mV
OD_MCLK:
0:        300MHz        800mV
1:       1750MHz        850mV
OD_RANGE:
SCLK:     214MHz       1800MHz
MCLK:     300MHz       2000MHz
VDDC:     800mV        1150mV

then increase the overdrive limit

 echo "4" > /sys/class/drm/card0/device/pp_mclk_od

# increasing mem overclocking limit by 4%

now edit the pstates and commit format is "<s/m for core/mem clock> <p-state num> <clock in MHz> <voltage in mV>

 echo "m 1 1805 850" > /sys/class/drm/card0/device/pp_od_clk_voltage

commit

echo "c" > /sys/class/drm/card0/device/pp_od_clk_voltage

and voila

cat /sys/kernel/debug/dri/0/amdgpu_pm_info
Clock Gating Flags Mask: 0x3fbcf        
Graphics Medium Grain Clock Gating: On        
Graphics Medium Grain memory Light Sleep: On        
Graphics Coarse Grain Clock Gating: On        
Graphics Coarse Grain memory Light Sleep: On        
Graphics Coarse Grain Tree Shader Clock Gating: Off        
Graphics Coarse Grain Tree Shader Light Sleep: Off        
Graphics Command Processor Light Sleep: On        
Graphics Run List Controller Light Sleep: On        
Graphics 3D Coarse Grain Clock Gating: Off        
Graphics 3D Coarse Grain memory Light Sleep: Off        
Memory Controller Light Sleep: On        
Memory Controller Medium Grain Clock Gating: On        
System Direct Memory Access Light Sleep: Off        
System Direct Memory Access Medium Grain Clock Gating: On        
Bus Interface Medium Grain Clock Gating: Off        
Bus Interface Light Sleep: On        
Unified Video Decoder Medium Grain Clock Gating: On        
Video Compression Engine Medium Grain Clock Gating: On        
Host Data Path Light Sleep: On        
Host Data Path Medium Grain Clock Gating: On        
Digital Right Management Medium Grain Clock Gating: Off        
Digital Right Management Light Sleep: Off        
Rom Medium Grain Clock Gating: On        
Data Fabric Medium Grain Clock Gating: Off  
GFX Clocks and Power:        
1805 MHz (MCLK)        
1110 MHz (SCLK)        
214 MHz (PSTATE_SCLK)        
300 MHz (PSTATE_MCLK)        
868 mV (VDDGFX)        
39.56 W (average GPU)  
GPU Temperature: 57 C
GPU Load: 100 %  
UVD: Disabled  
VCE: Disabled

(commiting didn't work before maybe I entered the wrong values)

[Mel Spektor]

How many fps you gain with GPU overclocking, test with the Tomb Raider 2013 benchmark?.Arch Linux wiki: Warning: Double check the entered values, as mistakes might instantly cause fatal hardware damage! 150Mhz of more memory clock speed does not have any effect.

on Unigine Superposition I get an increment of 30 points more or less. On mining an increment of 0.6 MH. But what makes the difference is unlocking computing units on this cards with a modded bios, instead of 14 you unlock two more.

[Mel Spektor]

I think that rx560 already has 16 compute units, except for some cut down cheaper versions. But if you have a rx460 you can transform it to a rx560 basically.