Well, upgrading made sense: With the most recent upgrade I can finally run my RAM at 3600MHz. Before that I could only go up to 3200. And I can't detect any problems with the CPU.

I've got 3 different Radeon cards that work with amdgpu, not one of them could generate a functional mutli monitor setup without a massive hassle. None of them work ootb with a vga monitor I've got. I just got a 5700 and it took me hours and hours to get a working multi monitor setup. I had to hook my Acer monitor up to my NVidia system and export an EDID file and then figure out the horribly documented kernel switches needed to get the EDID file read so the monitor would work at over 800x600. I also had to include the edid file in the initrd (not 100% sure this was necessary due to the aforementioned bad documentation but what I read seemed to indicate that amdgpu uses early KMS and requires the EDID in the initrd). The end result is great but a massive amount of work needs to be done to get EDID reading and multi monitor setups to work. I shouldn't have to have a NVidia box around to generate edid files so that my AMD video card can display to a monitor properly.

I saw some edid progams available in Ubuntu (read-edid, parse-edid???) but couldn't get them working in Fedora. Getting properly functioning monitors shouldn't be this complicated.

Last edited by m2mg2 on 4 April 2020 at 3:04 am UTC

For completeness' sake: Not always but every now and then I end up with this

[    6.954609] amdgpu: [powerplay] failed send message: PowerDownJpeg (46)  param: 0x00000000 response 0xffffffc2
[    6.954653] [drm:amdgpu_dpm_enable_jpeg [amdgpu]] *ERROR* Dpm disable jpeg failed, ret = -62. 
[    9.171323] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[    9.171364] [drm:amdgpu_dpm_enable_uvd [amdgpu]] *ERROR* Dpm disable uvd failed, ret = -62. 

"PowerDownJpeg"?

[   13.262325] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   15.730846] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   17.949442] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   20.427404] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   22.908628] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   25.392032] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   27.875584] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   30.355106] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   32.571936] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   35.052594] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   37.531160] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   40.013514] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   42.117738] amdgpu: [powerplay] Msg issuing pre-check failed and SMU may be not in the right state!
[   42.117740] amdgpu: [powerplay] SMU11 attempt to set divider for DCEFCLK Failed!

...and then it boots as if nothing ever has happened.


Well, the problem has been known for months now:

https://gitlab.freedesktop.org/drm/amd/issues/929

I have a 5700 XT using one HDMI at 24hz 1080p and one DP at 144hz 1080p. I don't have that issue. Have you tried generating edid files for both monitors and forcing them to be used instead of the built in edid detection? Are you using xorg.conf file/files? I had to create an xorg.conf and force an edid file for things to work right. The built in edid detection is garbage, especially with multi monitor. The thread seemed to indicate it could be related to metamodes, which should get resolved with the correct forced edid.

Last edited by m2mg2 on 15 April 2020 at 5:03 pm UTC

Just installed RX 5700 TI on my computer (Debian testing) all seems to be fine.

uname -r
5.5.0-1-amd64


glxinfo | grep "OpenGL renderer\|OpenGL version"
OpenGL renderer string: AMD Radeon RX 5700 XT (NAVI10, DRM 3.36.0, 5.5.0-1-amd64, LLVM 9.0.1)
OpenGL version string: 4.5 (Compatibility Profile) Mesa 19.3.3


ls /lib/firmware/amdgpu/ | grep navi10
navi10_asd.bin
navi10_ce.bin
navi10_gpu_info.bin
navi10_me.bin
navi10_mec2.bin
navi10_mec.bin
navi10_pfp.bin
navi10_rlc.bin
navi10_sdma1.bin
navi10_sdma.bin
navi10_smc.bin
navi10_sos.bin
navi10_ta.bin
navi10_vcn.bin

Last edited by nolddor on 15 April 2020 at 9:57 pm UTC

Quoting: nolddorJust installed RX 5700 TI on my computer (Debian testing) all seems to be fine.

uname -r
5.5.0-1-amd64


glxinfo | grep "OpenGL renderer\|OpenGL version"
OpenGL renderer string: AMD Radeon RX 5700 XT (NAVI10, DRM 3.36.0, 5.5.0-1-amd64, LLVM 9.0.1)
OpenGL version string: 4.5 (Compatibility Profile) Mesa 19.3.3


ls /lib/firmware/amdgpu/ | grep navi10
navi10_asd.bin
navi10_ce.bin
navi10_gpu_info.bin
navi10_me.bin
navi10_mec2.bin
navi10_mec.bin
navi10_pfp.bin
navi10_rlc.bin
navi10_sdma1.bin
navi10_sdma.bin
navi10_smc.bin
navi10_sos.bin
navi10_ta.bin
navi10_vcn.bin

What does dmesg say? No powerplay in red? As far as firmware goes - hashes are more interesting, since the files don't vary in name and hardly vary in size. Next point: How many displays attached? Finally: It works most of the time, which means I can get 10 boots in a row without any hiccups, then 2 boots with a plethora of errors during boot, then back to normal.

Quoting: m2mg2I have a 5700 XT using one HDMI at 24hz 1080p and one DP at 144hz 1080p. I don't have that issue. Have you tried generating edid files for both monitors and forcing them to be used instead of the built in edid detection? Are you using xorg.conf file/files? I had to create an xorg.conf and force an edid file for things to work right. The built in edid detection is garbage, especially with multi monitor. The thread seemed to indicate it could be related to metamodes, which should get resolved with the correct forced edid.

No I haven't generated any edid files. And for f*ck's sake I don't want to. Because...

• I have had this displays for about two years now and


• They never had any issues with my NVidia setup


• They don't have any issues with kernel 5.3/Kisak's mesa (apart from that the displays are swapped when compared to my kernel 5.6 on Ubuntu's 20.04 mesa)


• boot is pretty smooth most of the time, which would make testing of my edid setup just a PITA

I am completely aware that maybe a custom edid file could solve my problems. But this was the intention of this thread all along: That you don't get a "mildly pleasant" OOTB experience with an RX 5700. Sure, I can setup edid files, try these and that mesa versions, throw in some BIOS files, pick a mainline kernel, do some distro hopping, or draw a pentagram and sacrifice my hamster (redheads are way too scarce nowadays). It's just not something I expect to do in 2020.

Quoting: Tuxee

Quoting: m2mg2I have a 5700 XT using one HDMI at 24hz 1080p and one DP at 144hz 1080p. I don't have that issue. Have you tried generating edid files for both monitors and forcing them to be used instead of the built in edid detection? Are you using xorg.conf file/files? I had to create an xorg.conf and force an edid file for things to work right. The built in edid detection is garbage, especially with multi monitor. The thread seemed to indicate it could be related to metamodes, which should get resolved with the correct forced edid.

No I haven't generated any edid files. And for f*ck's sake I don't want to. Because...

• I have had this displays for about two years now and


• They never had any issues with my NVidia setup


• They don't have any issues with kernel 5.3/Kisak's mesa (apart from that the displays are swapped when compared to my kernel 5.6 on Ubuntu's 20.04 mesa)


• boot is pretty smooth most of the time, which would make testing of my edid setup just a PITA

I am completely aware that maybe a custom edid file could solve my problems. But this was the intention of this thread all along: That you don't get a "mildly pleasant" OOTB experience with an RX 5700. Sure, I can setup edid files, try these and that mesa versions, throw in some BIOS files, pick a mainline kernel, do some distro hopping, or draw a pentagram and sacrifice my hamster (redheads are way too scarce nowadays). It's just not something I expect to do in 2020.

If you look at my previous post on this thread before the one quoted, I had many of those same complaints. The OOTB experience with amdgpu is garbage, unless you're really lucky and have the right monitor on the right display connector. This is deeper than just the 5700. I had these issues on all 3 of my amdgu compatible cards with three different monitors. How can amdgpu not do something as basic as read edid settings? On my primary gaming set up with 2 monitors neither one had a functional display OOTB.

I didn't have to mess with any BIOS versions, jump distro's, deal with kernel or mesa versions. The only things I had to do were generate the xorg.conf, create and force the edid settings (I put the edid in the initrd but I'm not sure that is needed). BUT, it took me way too long to get there and that is way too much!!!

Overall I'm happy with my functionality and that I don't have to deal with a proprietary module every kernel update but Linux can do better. At the very least we need a utility that can read the edid settings properly, put a file in the firmware folder and add the kernel parameters to force them to be used.

I don't know if this might help but I just noticed the AMD firmwares have been updated in the linux-firmware git: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/

Quoting: TuxeeWhat does dmesg say? No powerplay in red? As far as firmware goes - hashes are more interesting, since the files don't vary in name and hardly vary in size. Next point: How many displays attached? Finally: It works most of the time, which means I can get 10 boots in a row without any hiccups, then 2 boots with a plethora of errors during boot, then back to normal.

# sudo dmesg | grep amdgpu


[    1.324728] [drm] amdgpu kernel modesetting enabled.
[    1.324826] fb0: switching to amdgpudrmfb from EFI VGA
[    1.324877] amdgpu 0000:03:00.0: vgaarb: deactivate vga console
[    1.337717] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_gpu_info.bin
[    1.337725] amdgpu 0000:03:00.0: No more image in the PCI ROM
[    1.337759] amdgpu 0000:03:00.0: VRAM: 8176M 0x0000008000000000 - 0x00000081FEFFFFFF (8176M used)
[    1.337760] amdgpu 0000:03:00.0: GART: 512M 0x0000000000000000 - 0x000000001FFFFFFF
[    1.337878] [drm] amdgpu: 8176M of VRAM memory ready
[    1.337879] [drm] amdgpu: 8176M of GTT memory ready.
[    1.338035] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_sos.bin
[    1.338063] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_asd.bin
[    1.338111] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_smc.bin
[    1.338158] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_pfp.bin
[    1.338205] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_me.bin
[    1.338253] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_ce.bin
[    1.338265] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_rlc.bin
[    1.338311] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_mec.bin
[    1.338356] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_mec2.bin
[    1.339181] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_sdma.bin
[    1.339195] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_sdma1.bin
[    1.339326] amdgpu 0000:03:00.0: firmware: direct-loading firmware amdgpu/navi10_vcn.bin
[    2.057728] amdgpu: [powerplay] use vbios provided pptable
[    2.057879] amdgpu: [powerplay] smu driver if version = 0x00000033, smu fw if version = 0x00000035, smu fw version = 0x002a3200 (42.50.0)
[    2.057880] amdgpu: [powerplay] SMU driver if version not matched
[    2.102231] amdgpu: [powerplay] SMU is initialized successfully!
[    2.164631] fbcon: amdgpudrmfb (fb0) is primary device
[    3.429915] amdgpu 0000:03:00.0: fb0: amdgpudrmfb frame buffer device
[    3.445766] amdgpu 0000:03:00.0: ring 0(gfx_0.0.0) uses VM inv eng 4 on hub 0
[    3.445767] amdgpu 0000:03:00.0: ring 1(comp_1.0.0) uses VM inv eng 5 on hub 0
[    3.445768] amdgpu 0000:03:00.0: ring 2(comp_1.1.0) uses VM inv eng 6 on hub 0
[    3.445769] amdgpu 0000:03:00.0: ring 3(comp_1.2.0) uses VM inv eng 7 on hub 0
[    3.445770] amdgpu 0000:03:00.0: ring 4(comp_1.3.0) uses VM inv eng 8 on hub 0
[    3.445770] amdgpu 0000:03:00.0: ring 5(comp_1.0.1) uses VM inv eng 9 on hub 0
[    3.445771] amdgpu 0000:03:00.0: ring 6(comp_1.1.1) uses VM inv eng 10 on hub 0
[    3.445772] amdgpu 0000:03:00.0: ring 7(comp_1.2.1) uses VM inv eng 11 on hub 0
[    3.445772] amdgpu 0000:03:00.0: ring 8(comp_1.3.1) uses VM inv eng 12 on hub 0
[    3.445773] amdgpu 0000:03:00.0: ring 9(kiq_2.1.0) uses VM inv eng 13 on hub 0
[    3.445774] amdgpu 0000:03:00.0: ring 10(sdma0) uses VM inv eng 14 on hub 0
[    3.445774] amdgpu 0000:03:00.0: ring 11(sdma1) uses VM inv eng 15 on hub 0
[    3.445775] amdgpu 0000:03:00.0: ring 12(vcn_dec) uses VM inv eng 4 on hub 1
[    3.445776] amdgpu 0000:03:00.0: ring 13(vcn_enc0) uses VM inv eng 5 on hub 1
[    3.445777] amdgpu 0000:03:00.0: ring 14(vcn_enc1) uses VM inv eng 6 on hub 1
[    3.445777] amdgpu 0000:03:00.0: ring 15(vcn_jpeg) uses VM inv eng 7 on hub 1
[    3.445895] [drm] Initialized amdgpu 3.36.0 20150101 for 0000:03:00.0 on minor 0
[    4.554228] snd_hda_intel 0000:03:00.1: bound 0000:03:00.0 (ops amdgpu_dm_audio_component_bind_ops [amdgpu])

# sudo sha256sum /lib/firmware/amdgpu/navi10*

af6cadd2f85b7dbeefe5ded35b12f9cc001a3b7e0d9be91ab12db387b521ead8  /lib/firmware/amdgpu/navi10_asd.bin
6da63c3d8f20bdee46ab231637fa5d9a73d6cdcb48cad91c7761ea05e1123b9d  /lib/firmware/amdgpu/navi10_ce.bin
31ce8d3241a5c21b59607c0a456c8396918a47f624343e80a789beac5b4f3395  /lib/firmware/amdgpu/navi10_gpu_info.bin
a2401289e3099a1d6180af6695cbbd132841acdd681e9984205751620b3e5da4  /lib/firmware/amdgpu/navi10_me.bin
bd082f5ea8b597773bd7c6c174d8855eac4b19dd1e8d062110b75f4f89cfc5ff  /lib/firmware/amdgpu/navi10_mec2.bin
bd082f5ea8b597773bd7c6c174d8855eac4b19dd1e8d062110b75f4f89cfc5ff  /lib/firmware/amdgpu/navi10_mec.bin
969c0dd89630e389d840c0beb18d79a2a76cf75f2ae89930555d25c22178bd98  /lib/firmware/amdgpu/navi10_pfp.bin
4b909675ba07e1bff4c36ea2a8e9073bd42b05cb96dc517ef2ef6a61032722da  /lib/firmware/amdgpu/navi10_rlc.bin
3b454b32949970c353b371e5177bf26a4bd8883aa4f38c54f6b4163e7da40ae1  /lib/firmware/amdgpu/navi10_sdma1.bin
3b454b32949970c353b371e5177bf26a4bd8883aa4f38c54f6b4163e7da40ae1  /lib/firmware/amdgpu/navi10_sdma.bin
87f53230f19907a33d3fc0ae3a543ed4a8457713aa5f10f9cdf7b44edab4dd25  /lib/firmware/amdgpu/navi10_smc.bin
794234afa1a6a7367ac5258cdc927dbe21a9dfe492ff331ad16cfa4ebdd13b15  /lib/firmware/amdgpu/navi10_sos.bin
ec35349ccf9e5160f27e5ffda7d66b2a9e53c515d3d25064422b64abdb0e87f1  /lib/firmware/amdgpu/navi10_ta.bin
0fc0aa4e0b10ea8567e9442c21cfa2ffb5c9bcf975aad0793a213612f701d061  /lib/firmware/amdgpu/navi10_vcn.bin

Currently only one monitor (Samsung P2270HD) is connected using HDMI port.

Basically this is what I did to install the graphic card:
https://www.reddit.com/r/debian/comments/fyxdlw/high_cpu_usage_with_xorg_and_gnomeshell_despite/fnmdl3l/