I wrote to netbsd-users mailing list but not getting any engagement. I also see your hardware compatibility thread here but it is contrary to my experiences below regarding the GPU support. Hi all I am looking for help understanding (and setting up properly?) support for various AMD graphic cards. I am trying to get back to tinkering and planned to use my modern-ish desktop with NetBSD again. In the past weeks I have tested various generations of AMD graphic cards and unfortunately could not make any of them work fully (in regards to older cards, my 2560x1440 screen may add complexity here). Let me summarize my experience with each of the cards and perhaps anybody can throw ideas at me to try something else. All of this has been tested on a clean NetBSD 10.0RC2 installation with modular-xorg, xfce4 and firefox from pkgsrc-2023Q4. The screen is Lenovo P27h-20 with HDMI, DisplayPort and USB-C input. 1. AMD Radeon RX 6600 I understand this is the least supported one. X works only with modesetting driver, no drm in console (I also booted it with amdgpu driver enabled in my own kernel and radeon disabled. No luck. With "gop 6" in bootloader I manage to get native 2560x1440 resolution over DisplayPort. Screen looks crisp but very slow which is almost unusable. Audio doesn't work over DisplayPort but something does get detected with audio0@hdafg0 (see attached audio-rx6600 file). I discarded this card since it is definitely too new. 2. AMD Radeon HD 8570 When it is the only card in the PC (at some points I had both RX 6600 and others plugged in) it gets correct resolution detected in console (2560x1440) with radeon driver. X works with both radeon and modesetting drivers BUT: radeon provides direct rendering and glxgears shows: GL_RENDERER = AMD OLAND (DRM 2.50.0, 10.0_RC2, LLVM 13.0.1) GL_VERSION = 4.5 (Compatibility Profile) Mesa 21.3.9 GL_VENDOR = AMD however running any compositor (picom or enabling compositing in Xfce4) messes the screen totally. modesetting driver of course is software with llvmpipe and a bit slower, but works OK. Audio works over HDMI with hdafg0. That's the card I am using right now, with llvmpipe as it works the best. I guess I should wait for amdgpu driver to be fully available. 3. AMD Radeon HD 7450 This one seemed promising but was a disappointment. When two cards ware in the PC (RX 6600 and this one) it didn't detect console resolution at 2560x1440 but it went down to 1920x1080. X with no config works fine with hardware acceleration, glxgears report: GL_RENDERER = AMD CAICOS (DRM 2.50.0 / 10.0_RC2, LLVM 13.0.1) GL_VERSION = 3.1 Mesa 21.3.9 GL_VENDOR = X.Org Everything seems smooth BUT some applications like Firefox do not show all graphics correctly - like images missing on web pages, especially the ones with transparency. Some Firefox tabs also do not have the correct shape (missing rounded corners etc. also on web pages). This is very odd. Firefox has also been compiled from pkgsrc-2023Q4 after setting X11_TYPE=modular in /etc/mk.conf. Problem here is that I can't make the card to output 2560x1440 over HDMI even though it should support it. I tried with xrandr and modelines from either cvt or gtf (btw. what is the difference here?) I also tried with modelines taken from Xorg.0.log that I had from Radeon HD 8570. Still nothing - screen gets black and monitor goes to standby. The modelines from HD 8570 were read via EDID so should be correct: (...) [ 272.320] (II) modeset(0): Disabling kernel dirty updates, not required. [ 272.391] (II) modeset(0): EDID vendor "LEN", prod id 25065 [ 272.391] (II) modeset(0): Using EDID range info for horizontal sync [ 272.391] (II) modeset(0): Using EDID range info for vertical refresh [ 272.391] (II) modeset(0): Printing DDC gathered Modelines: [ 272.391] (II) modeset(0): Modeline "2560x1440"x0.0 241.50 2560 2608 2640 2720 1440 1443 1448 1481 +hsync -vsync (88.8 kHz eP) (...) 4. AMD Radeon HD 6670 The situation with this one is the same as with HD 7450 but after some digging I saw that 2560x1440 should be supported only with DVI-D dual link port that it has, not with HDMI. Today I bought such adapter (DVI-D dual link -> HDMI) but also it doesn't work with my screen. When both HD 7450 or HD 6670 are the only cards in the PC, display resolution seems to be correctly detected in console because at the point where drm is enabled and framebuffer is created with the resolution set - the screen again goes blank (on HDMI port on HD 7450 and on both HDMI and DVI-D on HD 6670). In all situations I made sure that /dev/dri/card0 has correct permissions (660 and my user is also in "wheel" group) - since the drivers often go with 640 when creating the device node automatically (something that should be fixed perhaps?) So my question is: Am I doing here something obviously wrong? Is anybody using any other supported card with 2560x1440 resolution over HDMI/DisplayPort that you can recommend? Do you have any ideas why the screen goes blank even in console after setting the resolution and later in X when using the EDID provided modeline? I don't think I can go with older cards than these ones since I risk not supporting resolutions higher that 1920x1080. The only other idea I may have is getting a higher end HD 6xxx or a HD 7xxx card that has DVI-D dual link (mine has D-SUB and HDMI only). Thank you in advance!

@"bbartlomiej"#p16287 Can you post a) the `dmesg` output b) what `gop` lists in the bootloader, and c) the `Xorg.0.log` for each of the cards?

Thanks for reaching out. All outputs (one card at a time) are stored in: https://cloud.krawczyk.it/s/5GfL2NCyskw4Lcy I am not that much interested in RX 6600 as I understand this is not supported and something may change if new KMS from Linux is imported eventually with amdgpu driver. Output for HD 6670 was gathered blind since screen goes dark after DRM loads 2560x1440 resolution - X is dark as well although the log seems to decode EDID information from the screen correctly. X for HD 7450 is working but some images are messed up as I described and can't get the resolution to go above 1920x1080 - screen goes dark with the same modeline that the screen is presenting.

@"bbartlomiej"#p16307 First things first: The resolutions listed by the `gop` command in the bootloader depends on 2 things: 1. firmware implementation (BIOS or Card-ROM). Some only present a few std. modes. 2. whether the system firmware is running in BIOS-compatible (CSM) mode or native (UEFI) mode. The max. resolutions are generally not available in CSM mode. FYI on the cards: Cards recognized by the Radeon DRMKMS driver can be seen [here](https://github.com/NetBSD/src/blob/trunk/sys/external/bsd/drm2/dist/include/drm/drm_pciids.h). For the AmdGPU DRMKMS driver, the list is [here](https://github.com/NetBSD/src/blob/trunk/sys/external/bsd/drm2/dist/drm/amd/amdgpu/amdgpu_drv.c#L739). So, for your cards: ```text $ egrep '(radeon|genfb)0 at .+ ATI ' netbsd-radeon/dmesg-* netbsd-radeon/dmesg-hd6670:[ 1.043392] radeon0 at pci9 dev 0 function 0: ATI Technologies Radeon HD 6670/7670 (rev. 0x00) netbsd-radeon/dmesg-hd7450:[ 1.043464] radeon0 at pci9 dev 0 function 0: ATI Technologies Radeon HD 6450/7450/8450 / R5 230 OEM (rev. 0x00) netbsd-radeon/dmesg-hd8570:[ 1,000702] radeon0 at pci9 dev 0 function 0: ATI Technologies Radeon HD 8570 / R7 240/340 / Radeon 520 OEM (rev. 0x00) netbsd-radeon/dmesg-rx6600:[ 1,025369] genfb0 at pci11 dev 0 function 0: ATI Technologies product 73ff (rev. 0xc7) $ $ fgrep PCI: netbsd-radeon/Xorg.0.log-hd* netbsd-radeon/Xorg.0.log-hd6670:[ 93.998] (--) PCI:*(9@9:0:0) 1002:6758:1458:2545 rev 0, Mem @ 0xe0000000/268435456, 0xfce20000/131072, I/O @ 0x0000e000/256, BIOS @ 0x????????/131072 netbsd-radeon/Xorg.0.log-hd7450:[ 53.086] (--) PCI:*(9@9:0:0) 1002:6779:1642:3a65 rev 0, Mem @ 0xe0000000/268435456, 0xfce20000/131072, I/O @ 0x0000e000/256, BIOS @ 0x????????/131072 netbsd-radeon/Xorg.0.log-hd8570:[ 104.197] (--) PCI:*(9@9:0:0) 1002:6611:1b0a:90c4 rev 0, Mem @ 0xe0000000/268435456, 0xfce00000/262144, I/O @ 0x0000e000/256, BIOS @ 0x????????/131072 $ ``` we can see: ```text Card PCI-ID Radeon AmdGPU ---- ------ ------ ------ HD-6670: 1002:6758 Yes No HD-7450: 1002:6779 Yes No HD-8570: 1002:6611 Yes Yes RX-6600: 1002:73ff No No ``` that: a) the HD-6670 and HD-7450 are supported only by the Radeon DRMKMS driver. b) the HD-8570 is supported by both, and, c) the newer RX-6600 doesn't have a DRMKMS driver at all (which is why it uses `genfb`). For the monitor, `edid-decode` (`wip/edid-decode`, or [here]()) shows: ```text $ cat /tmp/edid-len.txt 00ffffffffffff0030aee96100000000 0a200103803c22782e6665a9544c9d26 105054a1080081809500b300d1c0d100 a9c081c08100565e00a0a0a029503020 3500615d2100001a000000fc00503237 682d32300a2020202020000000fd0032 4c1e5a22000a202020202020000000ff 0056393039585759480a202020200190 02031ff14c01020304051413901f120e 0f23090f078301000065030c00100001 1d007251d01e206e2855005550210000 1e8c0ad08a20e02d10103e9600555021 00001800000000000000000000000000 00000000000000000000000000000000 00000000000000000000000000000000 00000000000000000000000000000094 $ ./edid-decode -c /tmp/edid-len.txt edid-decode (hex): 00 ff ff ff ff ff ff 00 30 ae e9 61 00 00 00 00 0a 20 01 03 80 3c 22 78 2e 66 65 a9 54 4c 9d 26 10 50 54 a1 08 00 81 80 95 00 b3 00 d1 c0 d1 00 a9 c0 81 c0 81 00 56 5e 00 a0 a0 a0 29 50 30 20 35 00 61 5d 21 00 00 1a 00 00 00 fc 00 50 32 37 68 2d 32 30 0a 20 20 20 20 20 00 00 00 fd 00 32 4c 1e 5a 22 00 0a 20 20 20 20 20 20 00 00 00 ff 00 56 39 30 39 58 57 59 48 0a 20 20 20 20 01 90 02 03 1f f1 4c 01 02 03 04 05 14 13 90 1f 12 0e 0f 23 09 0f 07 83 01 00 00 65 03 0c 00 10 00 01 1d 00 72 51 d0 1e 20 6e 28 55 00 55 50 21 00 00 1e 8c 0a d0 8a 20 e0 2d 10 10 3e 96 00 55 50 21 00 00 18 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 94 ---------------- Block 0, Base EDID: EDID Structure Version & Revision: 1.3 Vendor & Product Identification: Manufacturer: LEN Model: 25065 Made in: week 10 of 2022 Basic Display Parameters & Features: Digital display Maximum image size: 60 cm x 34 cm Gamma: 2.20 DPMS levels: Off RGB color display Default (sRGB) color space is primary color space First detailed timing is the preferred timing Color Characteristics: Red : 0.6611, 0.3300 Green: 0.2978, 0.6152 Blue : 0.1494, 0.0644 White: 0.3134, 0.3291 Established Timings I & II: IBM : 720x400 70.081663 Hz 9:5 31.467 kHz 28.320000 MHz DMT 0x04: 640x480 59.940476 Hz 4:3 31.469 kHz 25.175000 MHz DMT 0x09: 800x600 60.316541 Hz 4:3 37.879 kHz 40.000000 MHz DMT 0x10: 1024x768 60.003840 Hz 4:3 48.363 kHz 65.000000 MHz Standard Timings: DMT 0x23: 1280x1024 60.019740 Hz 5:4 63.981 kHz 108.000000 MHz DMT 0x2f: 1440x900 59.887445 Hz 16:10 55.935 kHz 106.500000 MHz DMT 0x3a: 1680x1050 59.954250 Hz 16:10 65.290 kHz 146.250000 MHz DMT 0x52: 1920x1080 60.000000 Hz 16:9 67.500 kHz 148.500000 MHz DMT 0x45: 1920x1200 59.884600 Hz 16:10 74.556 kHz 193.250000 MHz DMT 0x53: 1600x900 60.000000 Hz 16:9 60.000 kHz 108.000000 MHz (RB) DMT 0x55: 1280x720 60.000000 Hz 16:9 45.000 kHz 74.250000 MHz DMT 0x1c: 1280x800 59.810326 Hz 16:10 49.702 kHz 83.500000 MHz Detailed Timing Descriptors: DTD 1: 2560x1440 59.950550 Hz 16:9 88.787 kHz 241.500000 MHz (609 mm x 349 mm) Hfront 48 Hsync 32 Hback 80 Hpol P Vfront 3 Vsync 5 Vback 33 Vpol N Display Product Name: 'P27h-20' Display Range Limits: Monitor ranges (GTF): 50-76 Hz V, 30-90 kHz H, max dotclock 340 MHz Display Product Serial Number: 'V909XWYH' Extension blocks: 1 Checksum: 0x90 ---------------- Block 1, CTA-861 Extension Block: Revision: 3 Underscans IT Video Formats by default Basic audio support Supports YCbCr 4:4:4 Supports YCbCr 4:2:2 Native detailed modes: 1 Video Data Block: VIC 1: 640x480 59.940476 Hz 4:3 31.469 kHz 25.175000 MHz VIC 2: 720x480 59.940060 Hz 4:3 31.469 kHz 27.000000 MHz VIC 3: 720x480 59.940060 Hz 16:9 31.469 kHz 27.000000 MHz VIC 4: 1280x720 60.000000 Hz 16:9 45.000 kHz 74.250000 MHz VIC 5: 1920x1080i 60.000000 Hz 16:9 33.750 kHz 74.250000 MHz VIC 20: 1920x1080i 50.000000 Hz 16:9 28.125 kHz 74.250000 MHz VIC 19: 1280x720 50.000000 Hz 16:9 37.500 kHz 74.250000 MHz VIC 16: 1920x1080 60.000000 Hz 16:9 67.500 kHz 148.500000 MHz (native) VIC 31: 1920x1080 50.000000 Hz 16:9 56.250 kHz 148.500000 MHz VIC 18: 720x576 50.000000 Hz 16:9 31.250 kHz 27.000000 MHz VIC 14: 1440x480 59.940060 Hz 4:3 31.469 kHz 54.000000 MHz VIC 15: 1440x480 59.940060 Hz 16:9 31.469 kHz 54.000000 MHz Audio Data Block: Linear PCM: Max channels: 2 Supported sample rates (kHz): 88.2 48 44.1 32 Supported sample sizes (bits): 24 20 16 Speaker Allocation Data Block: FL/FR - Front Left/Right Vendor-Specific Data Block (HDMI), OUI 00-0C-03: Source physical address: 1.0.0.0 Detailed Timing Descriptors: DTD 2: 1280x720 60.000000 Hz 16:9 45.000 kHz 74.250000 MHz (597 mm x 336 mm) Hfront 110 Hsync 40 Hback 220 Hpol P Vfront 5 Vsync 5 Vback 20 Vpol P DTD 3: 720x480 59.940060 Hz 3:2 31.469 kHz 27.000000 MHz (597 mm x 336 mm) Hfront 16 Hsync 62 Hback 60 Hpol N Vfront 9 Vsync 6 Vback 30 Vpol N Checksum: 0x94 Unused space in Extension Block: 60 bytes ---------------- Warnings: EDID: Base EDID: Some timings are out of range of the Monitor Ranges: Horizontal Freq: 28.125 - 88.786 kHz (Monitor: 30.000 - 90.000 kHz) Failures: Block 0, Base EDID: Basic Display Parameters & Features: sRGB is signaled, but the chromaticities do not match. Block 1, CTA-861 Extension Block: Missing VCDB, needed for Set Selectable RGB Quantization to avoid interop issues. EDID: CTA-861: Native progressive timings are a mix of several resolutions. EDID conformity: FAIL ``` So, the EDID isn't clean, but, none of the failures should be fatal. The resolutions chosen by the kernel are: ```text $ fgrep 'framebuffer at' netbsd-radeon/dmesg-* netbsd-radeon/dmesg-hd6670:[ 7.990347] radeondrmkmsfb0: framebuffer at 0xe0363000, size 2560x1440, depth 32, stride 10240 netbsd-radeon/dmesg-hd7450:[ 7.910612] radeondrmkmsfb0: framebuffer at 0xe0363000, size 1920x1200, depth 32, stride 7680 netbsd-radeon/dmesg-hd8570:[ 7,723635] radeondrmkmsfb0: framebuffer at 0xe05d8000, size 2560x1440, depth 32, stride 10240 netbsd-radeon/dmesg-rx6600:[ 1,025369] genfb0: framebuffer at 0xe0000000, size 2560x1440, depth 32, stride 10240 $ ``` and the Xorg server with either the `radeon` or `modesetting` display-drivers use those as well: ```text $ fgrep 'using initial mode' netbsd-radeon/Xorg.0.log-hd* netbsd-radeon/Xorg.0.log-hd6670:[ 94.490] (II) RADEON(0): Output DVI-0 using initial mode 2560x1440 +0+0 netbsd-radeon/Xorg.0.log-hd7450:[ 53.575] (II) RADEON(0): Output HDMI-0 using initial mode 1920x1080 +0+0 netbsd-radeon/Xorg.0.log-hd8570:[ 104.738] (II) modeset(0): Output HDMI-1 using initial mode 2560x1440 +0+0 $ ``` To get 2560x1440 resolution on the HD-6670 and HD-7450, try a setting like this: ```text $ cat /etc/X11/xorg.conf.d/display.conf Section "Device" Identifier "Card0" Driver "modesetting" # next, try "radeon" # Option "IgnoreEDID" "true" # for "radeon" EndSection Section "Monitor" Identifier "Monitor0" Option "DefaultModes" "false" Option "PreferredMode" "2560x1440" EndSection ``` and, make sure that: 1. you use the correct cables: Dual-Link DVI-D and/or Category 2 HDMI cables ("High Speed"). The DVI-D -> HDMI adapter is capable of handling 2560x1440. Some passive adapters may not work. 2. there's only one card in the system (otherwise, you'll have to disable the others in BIOS, or nail down the card using the `BusID` directive--and maybe specify the `Option "kmsdev" "/dev/dri/cardN"` for the `modesetting` driver as well) 3. the card is connected directly to the monitor w/o any KVM-switch in-between. 4. try a VGA cable. That should go up to 2048×1536@85 Hz (388 MHz pixel clock). For the display glitches you mentioned, I have no solution except that you try out both the `modesetting` and `radeon` Xorg display-drivers to see if which one works better and then look through their man-pages to see if fiddling with any `Option`s makes a difference.

Thanks @"rvp"#168 for awesome analysis - I will be testing it all in following days.

@"bbartlomiej"#p16335 do report us how it goes.

I tested HD 7450 since once HD 6670 picks up the resolution during boot - screen goes blank with no signal so I would have to do it blind. - with suggested updates to xorg.conf I still get black screen on startx using either modesetting or radeon driver. I also tried with 1920x1080 in PreferredMode but the same result. I have added respective Xorg.0.logs to the link above. Seems even with no config, the modesetting driver again reads EDID from the screen and tries the 2560x1440 resolution but it fails to set it correctly. Could the modeline be tuned using information from HD8570 which displays correctly? - with no Monitor section with radeon driver it works but again in wrong resolution and visible artifacts in Firefox. I have uploaded a sample web page with artifacts and the same web page with no artifacts (when used on HD8570 and modesetting driver) - my screen does not have a D-SUB input and I don't see a suitable VGA-HDMI adapter that would support higher than FullHD resolutions. I tested HD 8570 with amdgpu (I built myself a GENERIC kernel with uncommented "amdgpu" lines and commented "radeon" lines, but no drm/kms stuff is loaded on boot. This is a fresh build of 10.0_RC3 sources. dmesg added to the same place. Any suggestions? Is anybody using a different card supporting 2560x1440? I think I may have a HDMI-DisplayPort adapter somewhere. I'll test it as well.

@"bbartlomiej"#p16375 Just to make sure: you don't have [these](https://github.com/NetBSD/src/blob/netbsd-10/sys/arch/amd64/conf/GENERIC#L474) two set in your kernel config. I hope?

Problems with various AMD GPU generations and 2560x1440 screen

bbartlomiej

I tested HD 7450 since once HD 6670 picks up the resolution during boot - screen goes blank with no signal so I would have to do it blind.

with suggested updates to xorg.conf I still get black screen on startx using either modesetting or radeon driver. I also tried with 1920x1080 in PreferredMode but the same result. I have added respective Xorg.0.logs to the link above. Seems even with no config, the modesetting driver again reads EDID from the screen and tries the 2560x1440 resolution but it fails to set it correctly. Could the modeline be tuned using information from HD8570 which displays correctly?
with no Monitor section with radeon driver it works but again in wrong resolution and visible artifacts in Firefox. I have uploaded a sample web page with artifacts and the same web page with no artifacts (when used on HD8570 and modesetting driver)
my screen does not have a D-SUB input and I don't see a suitable VGA-HDMI adapter that would support higher than FullHD resolutions.

I tested HD 8570 with amdgpu (I built myself a GENERIC kernel with uncommented "amdgpu" lines and commented "radeon" lines, but no drm/kms stuff is loaded on boot. This is a fresh build of 10.0_RC3 sources. dmesg added to the same place.

Any suggestions? Is anybody using a different card supporting 2560x1440?
I think I may have a HDMI-DisplayPort adapter somewhere. I'll test it as well.

rvp

bbartlomiej I tested HD 8570 with amdgpu (I built myself a GENERIC kernel with uncommented "amdgpu" lines and commented "radeon" lines, but no drm/kms stuff is loaded on boot.

Yeah, I see that the Southern Islands GPUs need to get built with #define CONFIG_DRM_AMDGPU_SI 1 in external/bsd/drm2/dist/drm/amd/amdgpu/amdgpu.h, but, this isn't set anywhere. (Similar to the AMDGPU_CIK in external/bsd/drm2/amdgpu/files.amdgpu enabling #define CONFIG_DRM_AMDGPU_CIK 1)

Don't know why it isn't enabled. Maybe amdgpu is still experimental? Ask on the kernel mailing-list.

bbartlomiej with suggested updates to xorg.conf I still get black screen on startx using either modesetting or radeon driver. I also tried with 1920x1080 in PreferredMode but the same result.

Hmm. The man-page says the radeon driver should understand IgnoreEDID, but, the log-file says otherwise:

$ fgrep IgnoreEDID netbsd-radeon/Xorg.0.log-hd7450-radeon
[   457.353] (WW) RADEON(0): Option "IgnoreEDID" is not used
$

I'll have to go through the code here (both X and DRM) to figure out what's going on...

bbartlomiej Any suggestions? Is anybody using a different card supporting 2560x1440?

Intel Graphics 630 cards seems to work. You should also get the Intel firmware.

bbartlomiej

Thank you @rvp. The desktop PC though doesn't have any Intel GPU as it's a Ryzen 9 box. I will try again later with the HDMI-DisplayPort adapter on HD 7450.
I will report if anything changes.

I also got myself a T480 since it should be well supported (unfortunately with additional dGPU, so we'll see how that goes) since my current GPU problems on desktop make it unusable for daily use and web browsing.

rvp

bbartlomiej Just to make sure: you don't have these two set in your kernel config. I hope?

bbartlomiej

rvp I was using GENERIC with enabling amdgpu only (and disabling radeon) so these were commented out. Earlier I on HD 6670 when I noticed console is not in native resolution, I uncommented that and set to 2560x1440 but it did not help. Now it is commented out by default.
I haven't had time yet to test again with HD 7450 and DisplayPort adapter - I will update this thread once I do.

In the meantime I got a ThinkPad T480 with Intel iGPU and Nvidia dGPU. Display works on Intel even with USB-C video output to the same monitor - getting 2560x1440 perfectly.

rvp

bbartlomiej Can you post the Xorg.0.log from the Thinkpad so that I verify that the EDID retrieved is the same?

Can you see if, say, FreeBSD 12.x (which should end up bringing in a DRMKMS of the same vintage as NetBSD (5.6) or maybe 5.10, now) manages to achieve the full resolution? Just install onto a spare USB flash drive or somesuch, then: pkg install xorg drm-kmod; sysrc kld_list="radeonkms"

bbartlomiej

rvp I've uploaded Xorg.0.log from T480 for both intel driver and modesetting. By default it starts with intel but I've dropped in

Section "Device"
        Identifier "intel"
        Driver "modesetting"
EndSection

and only the modesetting driver prints out EDID data. Seems the modeline printed for 2560x1440 is the same as for HD 8570 however the actual EDID hex data seems different. Weird!

HD 8570:

[   104.732] (II) modeset(0): EDID (in hex):
[   104.732] (II) modeset(0): 	00ffffffffffff0030aee96100000000
[   104.732] (II) modeset(0): 	0a200103803c22782e6665a9544c9d26
[   104.732] (II) modeset(0): 	105054a1080081809500b300d1c0d100
[   104.732] (II) modeset(0): 	a9c081c08100565e00a0a0a029503020
[   104.732] (II) modeset(0): 	3500615d2100001a000000fc00503237
[   104.732] (II) modeset(0): 	682d32300a2020202020000000fd0032
[   104.732] (II) modeset(0): 	4c1e5a22000a202020202020000000ff
[   104.732] (II) modeset(0): 	0056393039585759480a202020200190
[   104.732] (II) modeset(0): 	02031ff14c01020304051413901f120e
[   104.732] (II) modeset(0): 	0f23090f078301000065030c00100001
[   104.732] (II) modeset(0): 	1d007251d01e206e2855005550210000
[   104.732] (II) modeset(0): 	1e8c0ad08a20e02d10103e9600555021
[   104.732] (II) modeset(0): 	00001800000000000000000000000000
[   104.732] (II) modeset(0): 	00000000000000000000000000000000
[   104.732] (II) modeset(0): 	00000000000000000000000000000000
[   104.732] (II) modeset(0): 	00000000000000000000000000000094
[   104.732] (II) modeset(0): Not using default mode "1920x1080" (bad mode clock/interlace/doublescan)
[   104.732] (II) modeset(0): Not using default mode "2048x1152" (bad mode clock/interlace/doublescan)
[   104.732] (II) modeset(0): Printing probed modes for output HDMI-1
[   104.732] (II) modeset(0): Modeline "2560x1440"x60.0  241.50  2560 2608 2640 2720  1440 1443 1448 1481 +hsync -vsync (88.8 kHz eP)

Intel GPU:

[   389.991] (II) modeset(0): Number of EDID sections to follow: 1
[   389.991] (II) modeset(0): EDID (in hex):
[   389.991] (II) modeset(0): 	00ffffffffffff0030aee96100000000
[   389.991] (II) modeset(0): 	0a200104a53c22783e6665a9544c9d26
[   389.991] (II) modeset(0): 	105054a1080081c0810081809500a9c0
[   389.991] (II) modeset(0): 	b300d1c00101565e00a0a0a029503020
[   389.991] (II) modeset(0): 	3500615d2100001a000000fc00503237
[   389.991] (II) modeset(0): 	682d32300a2020202020000000fd0032
[   389.991] (II) modeset(0): 	4c1e5a24000a202020202020000000ff
[   389.991] (II) modeset(0): 	0056393039585759480a202020200127
[   389.991] (II) modeset(0): 	020318f14b9005040302011f13140e0f
[   389.991] (II) modeset(0): 	23090f0783010000662156aa51001e30
[   389.991] (II) modeset(0): 	468f33000f282100001eab22a0a05084
[   389.991] (II) modeset(0): 	1a30302036000f282100001c7c2e90a0
[   389.991] (II) modeset(0): 	601a1e40302036000f282100001c0000
[   389.991] (II) modeset(0): 	00000000000000000000000000000000
[   389.991] (II) modeset(0): 	00000000000000000000000000000000
[   389.991] (II) modeset(0): 	0000000000000000000000000000006a
[   389.991] (II) modeset(0): Printing probed modes for output DP-2
[   389.991] (II) modeset(0): Modeline "2560x1440"x60.0  241.50  2560 2608 2640 2720  1440 1443 1448 1481 +hsync -vsync (88.8 kHz eP)

I will do more tests with the adapter and FreeBSD over the weekend. I need to travel for work tomorrow.

rvp

bbartlomiej Seems the modeline printed for 2560x1440 is the same as for HD 8570 however the actual EDID hex data seems different. Weird!

The monitor being connected via DisplayPort might account for that. However, there's something suggestive to be seen in a edid-decode -n output:

$ edid-decode -n < edid1.txt
[...]
Native Video Resolution if only Block 0 is parsed:
  2560x1440

----------------

Native Video Resolutions if Block 0 and CTA-861 Blocks are parsed:
  129x0                    # this looks like a bug
  1920x1080                # only this over HDMI?
$

Compare against edid-decode -p

bbartlomiej

Not sure what that would mean. The screen supports 2560x1440 over HDMI since it works on HD 8570 (although with software only rendering).

I tested with HD 7450 today and my DisplayPort-HDMI adapter but it didn't work at all (even in BIOS/UEFI). I am not getting any output when I use HDMI on GPU and plug it in DisplayPort port on the screen. Maybe it's a one-way adapter (DP -> HDMI, not HDMI -> DP)

Testing now with FreeBSD 12.4.

rvp

bbartlomiej Not sure what that would mean. The screen supports 2560x1440 over HDMI since it works on HD 8570 (although with software only rendering).

Yes, but the question is whether the other 2 cards can drive 2560x1440 over HDMI. Anyway, can you try passing this EDID blob to Xorg?

00FFFFFFFFFFFF0030AEE96100000000
0A200103803C22782E6665A9544C9D26
105054A1080081809500B300D1C0D100
A9C081C08100565E00A0A0A029503020
3500615D2100001A000000FC00503237
682D32300A2020202020000000FD0032
4C1E5A22000A202020202020000000FF
0056393039585759480A202020200190
02031FF14C01020304051413101F120E
0F23090F078301000065030C00100001
1D007251D01E206E2855005550210000
1E8C0AD08A20E02D10103E9600555021
00001800000000000000000000000000
00000000000000000000000000000000
00000000000000000000000000000000
00000000000000000000000000000014

Convert to binary:

$ printf "$(sed 's/../\\x&/g' edid.txt | tr -d '\n')" > HDMI-edid.bin
$ edid-decode < HDMI-edid.bin          # check if correct

Pass this blob to the radeon driver:

Option "CustomEDID" "HDMI-0:/some/path/HDMI-edid.bin:digital"

(Not sure if the :digital bit is reqd.)

Post the Xorg.0.log files.

I've just removed the "native" from the 1920x1080 resolution in the CTA-block. See if the Xorg server swallows this like a good boy.

bbartlomiej

rvp Try starting Firefox in an English locale: env LANG=en_US.UTF-8 firefox or env LANG=en_GB.UTF-8 firefox. And, after you've picked your jaw up off the floor, file a PR.

No, that didn't make any difference. On HD7450 using radeon driver it starts in 1920x1080 and screen has artifacts on XFCE4 desktop, menu entries and within Firefox's window - stuff missing from web pages. Added another screenshot to my Nextcloud link.

rvp Option "CustomEDID" "HDMI-0:/some/path/HDMI-edid.bin:digital"

I can't make Xorg read my custom file 🙁

[   139.286] (WW) RADEON(0): Option "UseEDID" is not used
[   139.286] (WW) RADEON(0): Option "CustomEDID" is not used

Updating to 10.0_RC4 since there were some changes in GPU-related areas (although Intel-related). In the end I will raise a PR since I'm also out of ideas.

rvp

bbartlomiej Everything seems smooth BUT some applications like Firefox do not show all graphics correctly - like images missing on web pages, especially the ones with transparency. Some Firefox tabs also do not have the correct shape (missing rounded corners etc. also on web pages). This is very odd.

Try starting Firefox in an English locale: env LANG=en_US.UTF-8 firefox or env LANG=en_GB.UTF-8 firefox. And, after you've picked your jaw up off the floor, file a PR.

rvp

bbartlomiej I can't make Xorg read my custom file

Ah, hell! I looked at the wrong radeon(4) man-page. Might have to fix this up in the kernel then. Did you try out FreeBSD or Linux on these cards?

bbartlomiej

rvp I started testing with FreeBSD 12.4 but there are no packages available anymore and ports gives warning that it's too old to be supported and no guarantee stuff will compile so I stopped that.

I did test with old Ubuntu with Linux 5.15.0 kernel and it seems to behave the same way on HD7450:

resolution is 1920x1080
Xorg.0.log shows modelines are read from the screen for 2560x1440 but are not used when I try to use xrandr
when I manually add a mode to xrandr using cvt, screen goes blank as on NetBSD.

I've attached the dmesg and Xorg.0.log in the link I shared at the beginning but it looks like this version of DRM doesn't support this card with this screen properly. I tested with Ubuntu using Linux 6.x and it is the same.

So my best shot is to wait for Radeon HD 8670 to get support in NetBSD as this did work on Linux 6.x.
Or can any of you recommend another GPU that's supported by NetBSD and will work in 2560x1440?
Thanks for all the help!

rvp

bbartlomiej I did test with old Ubuntu with Linux 5.15.0 kernel and it seems to behave the same way on HD7450:

Hmm. What about the latest Ubuntu?

bbartlomiej Xorg.0.log shows modelines are read from the screen for 2560x1440 but are not used when I try to use xrandr

I wanted to ask about this before, then plumb forgot: Can you post the xrandr output, and the command you used to set the new mode?

bbartlomiej when I manually add a mode to xrandr using cvt, screen goes blank as on NetBSD.

Did you try the modeline printed/got from EDID? Is it the same as what's generated by cvt?

bbartlomiej Or can any of you recommend another GPU that's supported by NetBSD and will work in 2560x1440?

Answered in post #8.

bbartlomiej

rvp Hmm. What about the latest Ubuntu?

tried just now. Same thing (Ubuntu now is full Wayland).

rvp I wanted to ask about this before, then plumb forgot: Can you post the xrandr output, and the command you used to set the new mode?

root@ubuntu:/sys/class/drm/card0-HDMI-A-1# xrandr --newmode 2560x1440 312.25 2560 2752 3024 3488 1440 1443 1448 1493 -hsync +vsync root@ubuntu:/sys/class/drm/card0-HDMI-A-1# xrandr --addmode HDMI-0 2560x1440 root@ubuntu:/sys/class/drm/card0-HDMI-A-1# xrandr Screen 0: minimum 320 x 200, current 1920 x 1080, maximum 16384 x 16384 HDMI-0 connected primary 1920x1080+0+0 (normal left inverted right x axis y axis) 609mm x 349mm 1920x1200 59.95 1920x1080 60.00* 60.00 50.00 59.94 1920x1080i 60.00 50.00 59.94 1680x1050 59.88 1600x900 60.00 1280x1024 60.02 1440x900 59.90 1280x800 59.91 1280x720 60.00 50.00 59.94 1024x768 60.00 1440x480 60.00 59.94 800x600 60.32 720x576 50.00 720x480 60.00 59.94 640x480 60.00 59.94 720x400 70.08 2560x1440 59.96 VGA-0 disconnected (normal left inverted right x axis y axis) root@ubuntu:/sys/class/drm/card0-HDMI-A-1# xrandr --output HDMI-0 --mode 2560x1440 root@ubuntu:/sys/class/drm/card0-HDMI-A-1# xrandr --output HDMI-0 --auto root@ubuntu:/sys/class/drm/card0-HDMI-A-1# uname -a Linux ubuntu 6.5.0-9-generic #9-Ubuntu SMP PREEMPT_DYNAMIC Sat Oct 7 01:35:40 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux

Of course on xrandr output you see 2560x1440 beacause I manually added it to HDMI-0

rvp Did you try the modeline printed/got from EDID? Is it the same as what's generated by cvt?

I remember trying both with modeline from cvt and the one from Xorg.0.log - same issue on HD 7450

rvp Answered in post #8.

Is Intel Arc supported by NetBSD? I haven't seen it in man pages anywhere. It stopped on Iris I think? Since my CPU is AMD Ryzen 9 I can't really use the Intel iGPU. I could only go with discrete Intel Arc GPU.

rvp

bbartlomiej

root@ubuntu:/sys/class/drm/card0-HDMI-A-1# xrandr --newmode 2560x1440 312.25  2560 2752 3024 3488  1440 1443 1448 1493 -hsync +vsync

Hmm. That doesn't match what Xorg prints:

[   104.732] (II) modeset(0): Modeline "2560x1440"x60.0  241.50  2560 2608 2640 2720  1440 1443 1448 1481 +hsync -vsync (88.8 kHz eP)

Can you try with those?

bbartlomiej

rvp
I did try with modelines from Xorg.0.log on NetBSD. The above snippets were from cvt just to show how I set it. Since it's Wayland-based I couldn't get the same logging information.

I tried with the same modeline on Ubuntu:

ubuntu@ubuntu:~$ xrandr --newmode 2560x1440 241.5 2560 2608 2640 2720 1440 1443 1448 1481 +hsync -vsync ubuntu@ubuntu:~$ xrandr Screen 0: minimum 320 x 200, current 1920 x 1200, maximum 16384 x 16384 HDMI-0 connected primary 1920x1200+0+0 (normal left inverted right x axis y axis) 609mm x 349mm 1920x1200 59.95* 1920x1080 60.00 60.00 50.00 59.94 1920x1080i 60.00 50.00 59.94 1680x1050 59.88 1600x900 60.00 1280x1024 60.02 1440x900 59.90 1280x800 59.91 1280x720 60.00 50.00 59.94 1024x768 60.00 1440x480 60.00 59.94 800x600 60.32 720x576 50.00 720x480 60.00 59.94 640x480 60.00 59.94 720x400 70.08 VGA-0 disconnected (normal left inverted right x axis y axis) 2560x1440 (0x6da) 241.500MHz +HSync -VSync h: width 2560 start 2608 end 2640 total 2720 skew 0 clock 88.79KHz v: height 1440 start 1443 end 1448 total 1481 clock 59.95Hz ubuntu@ubuntu:~$ xrandr --addmode HDMI-0 2560x1440 ubuntu@ubuntu:~$ xrandr --output HDMI-0 --mode 2560x1440

Still blank screen with mode out of range and I had to revert blind with:

ubuntu@ubuntu:~$ xrandr --output HDMI-0 --auto