It is possible that he is not hitting direct scanout and some of the perf hit is from that. If you launch gamescope as the compositor (i.e., not nested) he should get better performance since direct scanout should happen. It is also possible that some overlay is causing compositing to occur that is killing perf. In general, I don’t think there is anything CachyOS related (beside x86-64-v3/4 optimized kernel and packages) that is going to help here.
Yep this is actually quite possible. Will try gamescope and see if there is any impact.
I have a RX9070 and no real issues at all. Don’t be too worried.
Okay, since you don’t want to test whether the 120 FPS cap works, does the cable meet the minimum DP 2.0 specifications?
In your log file, CachyOS keeps issuing a runtime warning for your GPU. This could have been caused by Overdrive mode, but it might also indicate a poor pin connection on the GPU. Could you please post a new bug report without Overdrive?
I’m surprised the performance is so poor, as I have almost the exact same card. I just have the F3 model, whereas he has the slightly older F1 model.
As I don’t have these issues, it must have something to do with his setup.
As per earlier post - 25d4537
I am unsure if the DP cable is up to spec tbh. He was using hdmi but as Linux doesn’t support full bandwidth hdmi 2.1 he swapped it out.
Have you tried removing the graphics card and then plugging it back in?
Runtime errors are still appearing in the log, and Wayland is reporting DBus errors.
That would be the last hardware-related cause I can think of, unless it’s something really odd like an undersized power supply, etc.
Have you tested how it runs in X11 mode?
And by the way: the HDMI 2.1 standard only supports 4K@120Hz; the cable wouldn’t have been able to handle that at all. The 2.2 standard is expected this year, which is supposed to support up to 240Hz.
yes, I would get a certified cable based on the card’s requirements and check the spec of what the display supports as well.
Higher performance cards to reach the capabilities of the tech can be more particular about the cables and accessories.
I don’t know about modern GPUs but many used to come with cables designed to be compliant with the cards. Did your GPU come with a cable? use that. If not, try another cable. Or if you don’t have another cable does this display have another input? an HDMI rather than DP? swap the inputs to the other option and see if the problem persists.
(apologies if it’s all DP now or something, I don’t know the port setup of your card / displays)
He is using only DP, was my first thought too but:
active: DP-1,DP-2
empty: HDMI-A-1,HDMI-A-2
And my card came without a cable.
yeah, try HDMI and see if it’s some driver bug with DP. so many people are having problems with these cards and wayland that it’s hard to isolate the common factors with all the variables in rig setups.
Could also be the display, depending on it’s age and firmware. DP is a protocol that requires a lot more bandwidth to communicate than the general purpose HDMI needs.
So the display could be choking and failing on the software protocol side of the DP signal, but work fine on the more general media hardware-driven HDMI.
I am not the OP^^
But are you sure about your statement about the Overhead? DP was designed for getting rid of the big Media overhead of HDMI. I was wondered about that and what i cant validate your statement.
Sorry, you are correct. let me use different terms. I used overhead, but I should have used complexity in the layers of negotiation. The overhead of having a flexible feature-rich software layer that manages the connections, rather than the simplistic hardware driven negotiation of HDMI.
As I understand it:
DP depends on GPU and a software driver layer of the process on the source side to manage the connection (rates, linking, etc) where-as HDMI is just plug it in and the HDMI compliant hardware negotiates everything like the timing and refresh and all that stuff between the devices.
So, DP has it’s dominant layer of management moved into the software layer, where HDMI does most of it’s handshaking and management at the hardware layer. GPU driver updates getting pushed out can mess with the stability of display port. Where-as HDMI is mostly driven by the hardware portion and so the various GPU mfgs are all interacting with it in the same way.
Since DP’s management layer is feature rich and accessible by GPU manufacturers (and users to some degree there is more chatter and reasons for things to go wrong between the DP source and target devices. HDMI is mostly driven at the hardware layer out of the hands of the OS/drivers.
DP = More fine tuning, supports higher throughput / bandwidth, greater surface area for negotiation errors
HDMI = Idiot Proof, lower throughput standards (currently)
Sorry you do not find me correcting your “professional knowledge” as helpful.
Many of us have knowledge and some of it may be correct, or based on miss-information, as you mentioned, always happy to learn something new.
In this case, your claims are not based on real broad proven data, it is based on a specific case where a game was killing a very specific model of GPU from a specific vendor, which was found to be the card makers fault, not a wide spreads issue of “More FPS taxes and degrades cards”
That is me giving you new knowledge to learn from so you do not spread potential miss-information you once thought may of been correct.
Am I understanding this correctly? Your statement implies that higher power consumption – which automatically means higher currents and voltages, resulting in more heat dissipation and subjecting the delicate circuitry on PCBs and within chips to greater stress – has no negative effect on a graphics card’s lifespan.
And there were a few other cases, too – StarCraft 2, Diablo IV, Halo Infinite, Dragon’s Dogma and Skull and Bones are the ones I recall hearing about. Of course, these were isolated incidents and were quickly rectified, but they did happen.
And I just don’t find it helpful when the only contribution is “That particular statement isn’t quite right; it’s not a widespread problem at all.” Yet at the same time, no assistance is offered to the OP. And that runs through all your posts in this thread, even though this thread exists solely to provide the OP with a solution to their problem.
Going over this thread and reading that Windows works fine, but Linux is where performance falls short, this tells me that the hardware is fine and this is a software problem.
But before suggesting anything software wise, I would like a sanity check on something. Your friend is plugging in the DP/HDMI cable directly into the 9070XT GPU right? And not into the motherboard?
Also, did your friend do a CLEAN install of Cachy? Is he dual booting on the same drive? Did you give him an old drive with Cachy already installed on it?
I know these may be stupid questions to ask, but I have seen some crazy shit over the years and your post reminds me of these type of “new user” issues to PC gaming and newly converted Linux users.
As for software troubleshooting, I would do:
- In terminal force reinstall gpu drivers: sudo chwd -f -a
- Reboot, retest games
- In CachyOS Hello, under Apps/Tweaks, do “Reinstall all packages”
- Reboot, retest games
- Lastly/nuke option:
- Download a fresh ISO of Cachy, remake the bootable USB
- Unplug ALL hard drives/ssds from the system, plug in a new/blank/sacrificial ssd into the machine
- Install CachyOS fresh
- DON’T do any tinkering
- Install Steam via the CachyOS Hello Apps/Tweaks via the “Install Gaming packages” option
- In Steam set the Proton compatibility option to be proton-cachy-slr(or just try Proton Experimental).
- Retest games
I have had decent performance with my 9070XT and 3700X. I will say the only problems with performance I have had is in the native Linux War Thunder build. I have to have ray tracing disabled. I think that is specifically to do with that game though.
When comparing side by side you guys need to make sure all graphical settings and resolution are the same. I will fire up CS2 myself and see what that looks like.
Edit:
So for CS2 I was getting 60-120fps with lows in the 40fps. So maybe there is something out of place.
I may have been CPU limited though.
I’m on mesa git “cachyos-v3/mesa-git 26.1.0_devel.220238.c0f1689e117-1”
So I probably need to go back to the normal branch and test that as well for the sake of being thorough.
Edit 2:
Alright so for me enabling re sizable rebar in bios made the most significant change and I’m glad I stumbled on this thread. Enabling that got took me from the above to a pretty solid 180-200 fps in 1080p and got rid of the horrible lows.
Using these launch options
RADV_PERFTEST=gpl,nggc gamemoderun taskset -c 0-7 nice -n -10 %command%
I was able to get it to a stable 220fps with the lows for dust2 around 150-180.
At this point I’m pretty positive I’m cpu bound at this point but I’m happy. So maybe have your friend try out the rebar option in their bios as this made the biggest change for me.
I have a 9800X3D, 9070XT(Taichi) and 64GB(DDR5 6000) @1440p and I have ZERO complaints about FPS with CachyOS. When I installed CachyOS for the first time and that was the only time, I didn’t mess with anything other than updating thru the Terminal. Only issue I had was to get Star Citizen to run took a bit more than it did on Bazzite, but works now without issues. Speed wise nothing comes compares to CachyOS IMO. Windows 11 feels like a turtle in comparison, the time it takes to open Blender in CachyOS is instant vs. 3 to 5 seconds in Windows 11 and that’s on my 9950X3D with a 5090. I would just wipe your friends rig, make sure ReBar is enabled(or whatever your MB calls it) and like a previous user stated just install Steam, the game add-ons and test again. Yes cables do make a difference in this day and age for high FPS.
Noticeable degradation to the point it would impact the life of your card beyond normal usage, no, it does not, I have had cards that 10+ years later still run fine, and some that were used for mining back in the day 24/7 that still operate in systems given to others. Unless you are constantly running said cards out of spec, the impact is not something you could even measure.
I am contributing by correcting information that may lead someone to believe something, that does not hold any actual impact or grounds to be recommended to someone. You suggested something, I contributed by providing information to correct what you recommended with details.
I am sorry you personally do not find that as a contribution to the OP’s original problem, but your post is what caused my responses, since they are not helping the OP either with their problem, because your suggestions was based on a very isolated and specific issue, not applicable here.
It’s always a good idea to cap the FPS—it saves power, improves performance, and reduces wear and tear on the hardware.
`
I have a RX9060 and I get the expected performance in Cachy. No need to worry imho
If anyone happened to read my previous message here, formatting and reinstalling helped me. I must have broken something switching between mesa git and regular mesa months ago. Symptoms included things like mangohud not working correctly and weird framedrops with shaders that only happened on newer game installs.
I have RX 9070 XT same avg fps as windows but around 15% better 1% lows