Quoting: Grabby

Quoting: 3zekiel

Quoting: DoctorJunglist

Quoting: BeamboomVK_VALVE_mutable_descriptor_type still missing though :(

/*-That's what I'm waiting for as well, it's needed for Cyberpunk 2077.

I hope it doesn't take Nvidia too long to incorporate this extension into their drivers.

Quoting: ikirutoRTX 2060 and Ryzen 1700 works without problems on Proton-Experimental with the settings in the screenshots and hex fix.

https://ibb.co/bJ0Z315 [External Link]
https://ibb.co/M75Ndt4 [External Link]
https://ibb.co/p3mMvqw [External Link]

gamemoderun %command% --launcher-skip

Use a hex editor on the Cyberpunk2077.exe.
Replace "75 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08" with "EB 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08".
Replace "55 48 81 ec a0 00 00 00 0f 29 70 e8" with "c3 48 81 ec a0 00 00 00 0f 29 70 e8".

It seems the fix should be in exe for Nvidia. (second hand info though)

Doesn't that hex edit only affects the core scaling of the game ? Vk3d3d devs were saying that the engine is relying on undefined behavior in DX12, which is the cause of the crashes. I doubt core scaling has anything to do with that.
Besides, the crash are very random, so just because he didn't experience them after applying the fix doesn't mean the issue is solved.

Quoting: ikiruto

Quoting: GrabbyDoesn't that hex edit only affects the core scaling of the game ? Vk3d3d devs were saying that the engine is relying on undefined behavior in DX12, which is the cause of the crashes. I doubt core scaling has anything to do with that.
Besides, the crash are very random, so just because he didn't experience them after applying the fix doesn't mean the issue is solved.

Research is not a matter of faith, if there is something that needs verification, it should be verified, and not refer to your doubts. Crashes of the game are corrected not by the first line, but by the second.
 Signature of AVX bug is as follows:
Game crashes at and beyond these locations:
Take a seat for Corpo
Exit bar for Street Kid
After cutscene for arrival at city for Nomad   

You can do this mod yourself trivially. Simply use a hex editor on the Cyberpunk2077.exe,
and replace "55 48 81 ec a0 00 00 00 0f 29 70 e8" with "c3 48 81 ec a0 00 00 00 0f 29 70 e8". 
This simply turns the bad AVX call into a return.
DISCLAIMER: May be breaking things under the hood, but seems stable.

what he said I guess ? (once again only second hand, don't quote me, talk to the source :) )

View this comment - View article - View full comments

Quoting: pbUbuntu? No thanks. I tried it numerous times and there was always something broken. Wifi or hibernation, or touchpad, or drivers etc. etc. It's jut not worth the hassle.

I feel you, on work laptop, wifi disconnect, then if trying to reconnect flat out hang my laptop ... If I use my dual booted Fedora, no such issues.
I did solve many other issues by removing snap and its packages, and reinstalling with deb/flatpaks. I got application that are styled the sanme way as host, less disk/ram usage and overall some sw that used to crash a lot (slack between others) got a lot more stable with flatpak. So yeah, my love for that distro is very limited now, to say the least ...

View this comment - View article - View full comments

Quoting: Spirimint

Quoting: 3zekiel

Quoting: DoctorJunglist

Quoting: BeamboomVK_VALVE_mutable_descriptor_type still missing though :(

That's what I'm waiting for as well, it's needed for Cyberpunk 2077.

I hope it doesn't take Nvidia too long to incorporate this extension into their drivers.

Quoting: ikirutoRTX 2060 and Ryzen 1700 works without problems on Proton-Experimental with the settings in the screenshots and hex fix.

https://ibb.co/bJ0Z315 [External Link]
https://ibb.co/M75Ndt4 [External Link]
https://ibb.co/p3mMvqw [External Link]

gamemoderun %command% --launcher-skip

Use a hex editor on the Cyberpunk2077.exe.
Replace "75 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08" with "EB 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08".
Replace "55 48 81 ec a0 00 00 00 0f 29 70 e8" with "c3 48 81 ec a0 00 00 00 0f 29 70 e8".

It seems the fix should be in exe for Nvidia. (second hand info though)

would you say it works now like amd does?

As said, it is second hand info, I just fwd you info I read hoping it helps, so can't tell. It just seems some users are having an experience without problem with that patch, which seems to indicate there is an issue in exe.

And as @ikiruto pointed out, you would then have to have similar level cards to compare performance. Which I am in even less of a position to help with.

View this comment - View article - View full comments

Quoting: ToriI assume no AMD card handles this yet? Open and Closed Sourced drivers included.

Seems no indeed. Is only beta here too. Wait and see. I read this morning in Discord from Guy that DLSS is worked upon too: https://discord.com/channels/277857463384932353/653300722003214348/788105568812138498 [External Link]

That seems real good, because RT without DLSS is quite a hit on performance anyway.

View this comment - View article - View full comments

Quoting: DoctorJunglist

Quoting: BeamboomVK_VALVE_mutable_descriptor_type still missing though :(

That's what I'm waiting for as well, it's needed for Cyberpunk 2077.

I hope it doesn't take Nvidia too long to incorporate this extension into their drivers.

Quoting: ikirutoRTX 2060 and Ryzen 1700 works without problems on Proton-Experimental with the settings in the screenshots and hex fix.

https://ibb.co/bJ0Z315 [External Link]
https://ibb.co/M75Ndt4 [External Link]
https://ibb.co/p3mMvqw [External Link]

gamemoderun %command% --launcher-skip

Use a hex editor on the Cyberpunk2077.exe.
Replace "75 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08" with "EB 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08".
Replace "55 48 81 ec a0 00 00 00 0f 29 70 e8" with "c3 48 81 ec a0 00 00 00 0f 29 70 e8".

It seems the fix should be in exe for Nvidia. (second hand info though)

View this comment - View article - View full comments

Quoting: ikiruto

Quoting: 3zekielI see. For SMT, is not so rare to disable its use to be fair, not just on AMD. Some code just does not work well with it (bad load balance between both logical cores can be worst than just using the physical core). Well, it's not so common anymore though. Maybe they had some issues with zen1/zen1.5 ?

https://www.tomshardware.com/news/cyberpunk-2077-amd-ryzen-performance-bug-fix-testing [External Link]

To protect your privacy, external media requires approval to load.

Source: cdn.mos.cms.futurecdn.net

View cookie preferences.
Accept & Show Accept All & Don't show this again   Direct Link

Oh I know AMD does better than Intel nowadays, it is not what I meant. I meant is some cases, SMT in general can cause performance glitches, on both Intel and AMD. And just make hypothesis either they hit theses issues in some of their testing. Or specifically on some older ryzen, and then brutally cut it on every AMD CPUs, which would be dumb but yeah .....

Having a better arch, does not mean you will never hit theses issues. Sometimes is really an issue with the code itself (which opens the question of why not fix instead of doing that, but that's really hard to tell).

Now in general, I doubt SMT has much of an effect on a game. To make full use of it, you need to be able to feed both logical pipelines, but with instructions that do not overlap too much. Where it really works best is when you have threads mostly bound by IOs, in which case you avoid needless context switching between worker thread and io threads.
But if you already have 8 physical cores, the need is already much less.

View this comment - View article - View full comments

Quoting: Trias

Quoting: 3zekiel

Quoting: ikirutoRTX 2060 and Ryzen 1700 works without problems on Proton-Experimental with the settings in the screenshots and hex fix.

https://ibb.co/bJ0Z315 [External Link]
https://ibb.co/M75Ndt4 [External Link]
https://ibb.co/p3mMvqw [External Link]

gamemoderun %command% --launcher-skip

Use a hex editor on the Cyberpunk2077.exe.
Replace "75 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08" with "EB 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08".
Replace "55 48 81 ec a0 00 00 00 0f 29 70 e8" with "c3 48 81 ec a0 00 00 00 0f 29 70 e8".

Do you know what this changes ? (the exe patch)

There are reports that Cyberpunk is not utilizing SMT [External Link] (Simultaneous Multi-Threading) in Ryzen CPU's (i. e, it using only half cores of the CPU, ignoring all "logical" cores). Didn't try it myself, but first string in Hex fix above is supposed to fix it. Not sure about what second string do...

I see. For SMT, is not so rare to disable its use to be fair, not just on AMD. Some code just does not work well with it (bad load balance between both logical cores can be worst than just using the physical core). Well, it's not so common anymore though. Maybe they had some issues with zen1/zen1.5 ?

View this comment - View article - View full comments

Quoting: Arehandoro

Quoting: 3zekiel

Quoting: FauconNoirWhy everyone is talking about Pop!_OS ? What is so cool about it compared to others like Ubuntu or Manjaro for example ?

Ubuntu took / is taking a strange direction again (snap packages) which is bothering to say the least. It relies on proprietary server, is pretty much canonical locked in and is technically not that good. It is - like often with canonical - a rushed solution trying to reinvent what was already ongoing (flatpak, called xdg app at that time) but cutting every possible corners to go faster. Result is bloated install size, unclear sandboxing capacities, very slow first (but not always only the first) startup times, and overall opacity on what you install on your computer since it comes from a locked in source.

Flatpak is from Red Hat/IBM, it's also pretty much Red Hat locked in FlatHub (and nothing stops them from locking it even more like they've done with plenty other products), flatpak does not have confinement unless you use SELinux, which it isn't so easy to install in Debian based systems, you can't install different channels of the same application, runtimes don't get updated when apps depending on them are updated. None of them are perfect.

I use fedora because it's pretty up to date and for gaming, specially with AMD and in need ot Windows games, it's pretty needed, but it's definitely not the distro for new users.

Before recommending any distro to people first we should know their needs and uses, not just blindly recommending one. It's not the same someone that wants to play some Steam games, that someone that wants to play the latest with the latest hardware.

You can't really compare the case of flathub and snapcraft here. For flathub, absolutely everything (packaging wise) is open source and publicly available, same for server side sw. If they try to lock in, you can just pull everything and push somewhere, so the power is much much more in the hand of the community. Also, Fedora themselves (which are not strictly red hat) are doing their remote, gnome has theirs too I think. So the know how exists.

Now, of course, there are some issues. But if you update normally - I mean update all, not just one app -, runtime update will also work. But I agree, some dependency stuff need some improvement.
As for confinement, I'm pretty sure there is an apparmor backend. Confinment does work for me on work PC ubuntu, so not sure there really is an issue here ?

Flatpak is not perfect (well, nothing will ever be perfect) but it is moving forward quite nicely, and has good foundations. It also had a slow start, and some people thinking the pre 1.0 version were production ready and judged as such. Right now however, it has come along very far, and is a very good solution.
For snap, it is really just generalized docker with vendor lock in, so there's not much to be done to save it. Adoption (or its lack thereof for Snap) does show who's winning...

Now the issue is just like Mir/upstart, Canonical caused needless fragmentation because they think they can have influence on GNU/Linux direction by doing lock in / needless forks or new projects. But ultimately, they rush, gain some adoption while they are alone, and then slowly fail. It takes more or less time for each project, but ultimately they still fail, simply because their solution cause many unsolved/unsolvable technical issues, and because they always fail to gain real community traction (since they work against the existing community...). Leaving the issues of supporting their dying mess to others. At least for snap, it will be easy to just erase it, you will not have some "service start xxx" stuff in scripts for years to come.

With Pipewire, flatpak will pretty much have the final piece to its puzzle, and snap will hurry to where most Canonical forked / in house projects are now ... The code cemetery.

Now for the part about not recommending anything to anyone, yes we are clearly talking gaming here. But the overall usability of the system does matter (the very slow (first) startup times for snap apps as an example) as people rarely only just game on their system. Selinux is also an argument for basic random "found off the internet" stuff protection. Having spotify/Onlyoffice is probably something that will become important soon (French administration still sends you docx stuff as an example... Albeit less and less).

View this comment - View article - View full comments

Quoting: ikirutoRTX 2060 and Ryzen 1700 works without problems on Proton-Experimental with the settings in the screenshots and hex fix.

https://ibb.co/bJ0Z315 [External Link]
https://ibb.co/M75Ndt4 [External Link]
https://ibb.co/p3mMvqw [External Link]

gamemoderun %command% --launcher-skip

Use a hex editor on the Cyberpunk2077.exe.
Replace "75 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08" with "EB 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08".
Replace "55 48 81 ec a0 00 00 00 0f 29 70 e8" with "c3 48 81 ec a0 00 00 00 0f 29 70 e8".

Do you know what this changes ? (the exe patch)

View this comment - View article - View full comments

Quoting: FauconNoirWhy everyone is talking about Pop!_OS ? What is so cool about it compared to others like Ubuntu or Manjaro for example ?

Ubuntu took / is taking a strange direction again (snap packages) which is bothering to say the least. It relies on proprietary server, is pretty much canonical locked in and is technically not that good. It is - like often with canonical - a rushed solution trying to reinvent what was already ongoing (flatpak, called xdg app at that time) but cutting every possible corners to go faster. Result is bloated install size, unclear sandboxing capacities, very slow first (but not always only the first) startup times, and overall opacity on what you install on your computer since it comes from a locked in source.
A good example is that sppotify packages, where upstream only updated it like once or twice this year which is reflected in the flatpak, but snap had multiple version for god knows what reasons (most likely only marketing purposes).
Also since snaps are basically full blown containers, not only do they tend to be more bloated, but as most users point out, the window styles and co do not integrate with host at all, which gives a sense of total lack of polish to the experience.
And as a final nail to the coffin, while it is supposed to be universal, it only really works/is adopted on ubuntu ... Whereas flatpak works everywhere, including ubuntu.

So yeah, Pop OS took what is good (PPA etc), and threw away what is bad (snap, in favor of flatpak) hence the recommendation.

As for comparison to Manjaro, that's another story. You lose the PPA (but gain AUR) so it really is a question of preferences. Also Manjaro used to be less stable, although it seems to have gotten better. I would say it has matured a lot from my last tries. But now, would I recommend it over Pop OS for beginners? good question. Probably not yet, as I'm not sure there is as much resources available (Pop profits from all ubuntu too).
For more advances users, I do think I would recommend Manjaro for all its possibilities and bleeding edge sw. Also for being unpatched and thus allowing you to run custom kernels (scheduler/event driven patches) depending on your needs.

View this comment - View article - View full comments