Anyone had any luck with Helldivers 2? It seems to have the same issues that the 7000 series had before and just crashes on launch.
I’m on arch with 6.13.6, linux-firmware-git and mesa-git. Other games seem to be working fine so far.
Mine also crashes. No luck getting it past the first spinning wheel. I’m also having issues with Forza Horizon 5 and Wreckfest. Forza doesn’t load past the initial company logo videos and Wreckfest launches the process but nothing comes up.
Did this get backported to RDNA 3.5 on the 8700G?
Quick update on my experience with Ubuntu 24.04.2 LTS (Kernel 6.11.0-19-generic) running KDE Plasma:
Pros:
- Remote Play is amazing. I use this feature often, and with my old 1660 Ti, it was always finicky—flickering when switching to Big Picture Mode, poor connection speeds, and occasional lag. With my new GPU, it’s a completely different story. I can play Tekken 8 flawlessly on my phone. Coming from a CRT/Tube TV background, I’m pretty sensitive to input lag, but I couldn’t perceive any. It’s kind of mind-blowing.
Cons & Issues:
- Frequent crashes and system freezes. Certain games cause my system to lock up entirely. I initially suspected a power supply issue and upgraded, but the problem persists. Still troubleshooting.
- Some graphical glitches. Most of my games run great (CS:GO, Tekken 8 althought trying to play online crashed it, Yakuza 5, Bayonetta, Crash Trilogy, Total War: Three Kingdoms, Civilization VI, etc.), but others have quirks.
- Resident Evil 2: Floor reflections are weird or broken; game exits after idling for ~10 minutes.
- Quake 3: Displays strange color bars. Note: I just checked and its fixed now
- Occasional colored static in KDE Plasma. Not a dealbreaker but noticeable at times.
Final Thoughts:
For now, I’m sticking with my current setup and waiting for things to stabilize. Once I can safely run a simple apt dist-upgrade without issues, I’ll update. In the meantime, my backlog is massive, so I’ve got plenty to keep me busy.
Loving the thread, thanks to everyone for sharing their stories.
So I am using Fedora with a 9070. Something odd I’ve experienced is weird GPU hangs on Surviving Mars. I know it’s rather old, but I really like that game. It was the number one most played game in my Steam Library for a very long time, but now it’s been overtaken by Factorio. The game’s Linux port has gotten rather buggy since Abstraction took over development and I would rather run it through Proton except it has this one weird screen flashing issue. That’s a bit off topic; so anyway, the only way I have been able to get past the GPU hangs is by hitting the reset switch. This is something I need to avoid too, now, since I switched my gaming drives over to F2FS on top of LVM.
For crashes and freezes, do try to the run games with
I have not had a single crash since I added the var.
1 Like
Just to make sure :
mesa-git and linux-firmware-git are AUR packages.
Pacman won’t see those packages.
So if you want to install them, you’ll need to use an AUR helper (like yay, paru or others).
I have been using the bazzite testing image that has the Mesa 25 drivers and I have been beyond pleased with the results. The 9070XT gets drastically better performance than my 7800XT, which surprised me.
The best part? Performance should get better and better as the developers learn the ins and outs of Mesa drivers.
The 9070XT has been a home run if you ask me.
True, pacman can’t build aur packages. The standard tool to build aur packages is makepkg commandline tool. For example, here’s how to build my mesa-git package:
git clone https://aur.archlinux.org/mesa-git.git
cd mesa-git
makepkg -sri
As others mentioned, you can use an aur helper, but if you run into issues, you might not get an answer from aur package maintainers. I don’t use aur helpers myself but I try to respond to questions if they seem relevant.
That i in makepkg -sri means install after the build. The actual installation is done through pacman. makepkg -sr only build the package in the current directory. You can find the built package with .zst extension there. If you have the built package, you can use pacman -U built-package-name.zst to install it.
I have had crashes, in several games, some consistently (e.g. at launch) and some sporadically (say after an hour or so). In some cases it might be more stable with these options though.
AMDVLK just released (4-5 hours ago) so you can grab that (build from source or wait until your distro ships it) and try with games you have problems with. Then at least we can know if its a RADV issue or an issue with RDNA4 in general 
Going to try this on the cyberpunk benchmark because the big gpu firmware update didn’t fix that for me.
EDIT: The RADV_DEBUG arguments still didn’t fix the issue preventing the cyberpunk benchmark from running. KWin is still crashing just before the benchmark starts when on DP and it doesn’t reliably reset the desktop environment (or at the very least it doesn’t always bring the GPU’s display controller back up). At this point I’m willing to chalk it up to the combination of effects that Garuda Dragonized uses interact with the gpu in some way that it doesn’t like. I’ve got a completely free night tonight so I might try setting up an arch environment along with a minimally configured kde to reduce some variables here and see if I’m still running into these issues.
Made an account to chime in, I have had issues with CachyOS as well. Crashes back to login screen when attempting Cyberpunk (Game or benchmark). I have tried via HDMI and Display Port. Tried updating Mesa, Kernel etc (basically everything above this comment). And nothing has worked.
Guess will just have to wait for Mesa to get updated.
I’ve just received and gotten set up with my RX 9070 XT (XFX Quicksilver Magnetic Edition). I’m running Arch, having installed linux-firmware-git and mesa-git, and getting great performance. Using LACT to increase the power limit from 304 W to 340 W, I get a 4.5 % increase in frames when running a FurMark benchmark.
corectrl will allow you to increase the power limit and also allow you to undervolt, i was playing with it last night with nice results
1 Like
This post made me question everything. I had a perfectly reproducible crash at the very start of the game, right when you exit the door to the larger area with the helicopters. I had my kernel updated, my mesa-git, my linux-firmware-git, everything that should make this Sapphire Pulse 9070 XT work as well as it possibly can right now.
On a hunch I swapped motherboards from the cheapo Jginyue B550I board I’ve been happily using with my 4070 Ti and went back to my old AsRock B550M one, rock solid.
Edit: On the other hand, Remnant 2 doesn’t even make it in game now, whereas I could play maybe a minute before.
Edit 2: Remnant 2 does run great on AMDVLK 2025.Q1.3 though.
I gave the latest AMDVLK a shot on Arch Linux.
I ran Cyberpunk 2077 for about ten minutes before the below crash:
➜ ~ uname -a
Linux coco 6.13.6-zen1-1-zen #1 ZEN SMP PREEMPT_DYNAMIC Fri, 07 Mar 2025 20:18:44 +0000 x86_64 GNU/Linux
➜ ~ pacman -Qm mesa amdvlk linux-firmware
amdonly-gaming-mesa-git 25.1.0_devel.203127.bdfb478be11.d41d8cd-1
amdvlk-git 2025.Q1.3.r0.b29187c-1
linux-firmware-git 20250307.bd3d8a8b-1
[ 1193.625105] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:40 vmid:1 pasid:32785)
[ 1193.625112] amdgpu 0000:03:00.0: amdgpu: in process GameThread pid 4396 thread vkd3d_queue pid 4454)
[ 1193.625113] amdgpu 0000:03:00.0: amdgpu: in page starting at address 0x00000083da92a000 from client 10
[ 1193.625115] amdgpu 0000:03:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00140A50
[ 1193.625116] amdgpu 0000:03:00.0: amdgpu: Faulty UTCL2 client ID: CPC (0x5)
[ 1193.625117] amdgpu 0000:03:00.0: amdgpu: MORE_FAULTS: 0x0
[ 1193.625118] amdgpu 0000:03:00.0: amdgpu: WALKER_ERROR: 0x0
[ 1193.625118] amdgpu 0000:03:00.0: amdgpu: PERMISSION_FAULTS: 0x5
[ 1193.625119] amdgpu 0000:03:00.0: amdgpu: MAPPING_ERROR: 0x0
[ 1193.625120] amdgpu 0000:03:00.0: amdgpu: RW: 0x1
[ 1236.789775] rfkill: input handler enabled
[ 1241.692888] rfkill: input handler disabled
[ 1254.491262] amdgpu 0000:03:00.0: amdgpu: Dumping IP State
[ 1254.493356] amdgpu 0000:03:00.0: amdgpu: Dumping IP State Completed
[ 1254.503387] amdgpu 0000:03:00.0: amdgpu: ring comp_1.1.1 timeout, signaled seq=10619, emitted seq=10620
[ 1254.503390] amdgpu 0000:03:00.0: amdgpu: Process information: process GameThread pid 4396 thread vkd3d_queue pid 4454
[ 1254.503391] amdgpu 0000:03:00.0: amdgpu: Starting comp_1.1.1 ring reset
[ 1255.128498] amdgpu 0000:03:00.0: amdgpu: Ring comp_1.1.1 reset failure
[ 1255.128501] amdgpu 0000:03:00.0: amdgpu: GPU reset begin!
[ 1259.128510] amdgpu 0000:03:00.0: amdgpu: failed to suspend display audio
[ 1261.239407] amdgpu 0000:03:00.0: amdgpu: MES(1) failed to respond to msg=REMOVE_QUEUE
[ 1261.239410] [drm:amdgpu_mes_unmap_legacy_queue [amdgpu]] *ERROR* failed to unmap legacy queue
[ 1263.239909] amdgpu 0000:03:00.0: amdgpu: MES(1) failed to respond to msg=REMOVE_QUEUE
[ 1263.239912] [drm:amdgpu_mes_unmap_legacy_queue [amdgpu]] *ERROR* failed to unmap legacy queue
[ 1263.273643] amdgpu 0000:03:00.0: amdgpu: MODE1 reset
[ 1263.273645] amdgpu 0000:03:00.0: amdgpu: GPU mode1 reset
[ 1263.273684] amdgpu 0000:03:00.0: amdgpu: GPU smu mode1 reset
[ 1264.283622] amdgpu 0000:03:00.0: amdgpu: GPU reset succeeded, trying to resume
[ 1264.283708] amdgpu 0000:03:00.0: amdgpu: PCIE GART of 512M enabled (table at 0x00000083DAB00000).
[ 1264.283729] [drm] VRAM is lost due to GPU reset!
I tried the Launch Options for RADV mentioned by @sebas but they don’t change anything. The newest AMDVLK isn’t in NixOS just yet but I did make an issue prompting the maintainers to update it.
Any NixOS users:
-
Track this if you want to try the latest AMDVLK: Update Request: amdvlk 2025.Q1.2 → 2025.Q1.3 · Issue #389610 · NixOS/nixpkgs · GitHub
-
Latest
mesa-gitfrom Chaotic-Nyx should be available when this merges: Bump 20250313-1 by github-actions[bot] · Pull Request #995 · chaotic-cx/nyx · GitHub
It wasn’t in arch repo either, took 1h to compile on my machine 
Are you sure AMDVLK is being used? If you use mangohud you can set the vulkan_driver setting which will show which driver is being used to make sure it does not use RADV instead of AMDVLK (since you had both installed)
That error looks very much like the same issue as the RADV driver, so if it’s not it’s not RADV we should be expecting similar problems with AMDVLK. Think you can check?