NVIDIA 418.52.05 Linux Driver Brings Vulkan Ray-Tracing To Non-RTX GPUs

[starshipeleven]

I don't know if it's shitty\immature drivers to begin like starshipeleven says

I'll just say that AMD is known for this on Windows too. Their cards usually ran at around 20% disadvantage when first reviewed by the usual suspects, then 6 months later boom, the driver is actually mature and the card does not suck anymore but the damage is done.

Meanwhile NVIDIA does the opposite, at launch the performance is at maximum (sometimes they do screw up too mind me), then as time goes they decline, and it's not the first time I've encountered people that claim that they are using old drivers for their NVIDIA cards as with the newer ones they lost performance.

[skeevy420]

I'll just say that AMD is known for this on Windows too. Their cards usually ran at around 20% disadvantage when first reviewed by the usual suspects, then 6 months later boom, the driver is actually mature and the card does not suck anymore but the damage is done.

Meanwhile NVIDIA does the opposite, at launch the performance is at maximum (sometimes they do screw up too mind me), then as time goes they decline, and it's not the first time I've encountered people that claim that they are using old drivers for their NVIDIA cards as with the newer ones they lost performance.

That's kind of also what I meant by long term part with AMD -- we know they usually get better and stay good. NVidia cards, OTHO, the driver version really starts to matter more and more since any driver after 321.23 has a 10% performance drop with 980 TIs (numbers pulled from you know where).

[Stefem]

I'd expect some games to get more FPS, better frame times, or bugs fixed...noticeable performance improvements. 7 years in and my 260x still got noticeable improvements with every driver update. I've had two major Mesa updates since getting the 580 and each one brought improvements in my games and workflow.

I don't know if it's shitty\immature drivers to begin like starshipeleven says with or if it's the results of just keeping up with the times, but I've come to expect my AMD cards to improve over time.

Comparing a release to the next could show fluctuation for both AMD and NVIDIA while in the log term both improves, of course the more the driver is inefficient, the more you are likely to see improvement and vice versa. But if after 7 years your 260x still got noticeable improvements with every driver update, then now should be something like 3 times faster... so there are 3 options, your perception does not agree with factual reality (no offence), the driver was an awful crap, which is possible considering AMD's OpenGL performance in the past or with "noticeable" you meant something that can be measured, which may be just 1% variations.

[skeevy420]

Comparing a release to the next could show fluctuation for both AMD and NVIDIA while in the log term both improves, of course the more the driver is inefficient, the more you are likely to see improvement and vice versa. But if after 7 years your 260x still got noticeable improvements with every driver update, then now should be something like 3 times faster... so there are 3 options, your perception does not agree with factual reality (no offence), the driver was an awful crap, which is possible considering AMD's OpenGL performance in the past or with "noticeable" you meant something that can be measured, which may be just 1% variations.

With some games I was able to change my 30fps cap to 60fps near the end of its life -- Some games would fluctuate between 45-60 fps, vsync dropped it to 30 with a touch of lag, so I'd libstrangle (when possible) it to 30 to remove the vsync drop lag; eventually that wasn't needed and I could just play the game at 60fps. That applied to Tomb Raider, Metro 2033, Battlefield 4, Wolfenstein NO/OB, Alien Isolation, Shadow of Mordor, random Assassin's Creed games. With other games I was able to increase the graphical settings without losing FPS where I would before...sometimes it was just the game getting a Vulkan version that allowed better settings, like with Mad Max.

I consider being able to play at full FPS from half FPS or increasing graphics settings without losing FPS to be noticeable improvements.

A lot of stuff happened between 2012 and 2019. I went from Catalyst>Radeon>AMDGPU with that card. There were a lot of ups and downs during all of that. When it was all said and done, my 260x was a much better card due to driver improvements and AMDGPU. I think a lot of us that went through the entire driver transition phase will have similar anecdotal experiences. It gives me a that AMD still hasn't moved the CIK cards to AMDGPU by default now since I knew how well it worked for me.

I also went from a Q6600 with the 3GHz FSB mod to dual x5687 Westmeres (the fastest Westmere not including the rare x5698). I mean, I only quadrupled the thread count and put 6x more memory in the next PC. That also helped, but not as much as you'd think for games since 8GB ram and 4 threads is what most are designed around as well as I was still walled by the GPU. Driver updates, adding more OpenGL extensions, NIR/Sisched, etc helped more in regards to games.

With my 580 in the same dual Westmere system, I'm now walled by the CPUs...but it's good enough for 1080p60.

I'm just happy that I can play all of my games @ 1080p60 with mostly high settings now...damn you low to medium shadows

[loganj]

does this drivers work on kernel 5.1.2? seems that after upgrading kernel to 5.1.2 on arch i lost graphic desktop
same after replacing 4.15 with 5.1.3 on linuxmint. i've tried to reinstall it (dkms/or not) and it failed to build it.

[loganj]

the beauty of amd is that you have so many drivers that you probably can find one to work.
the problem with nvidia is that you have only one driver that works.
quick question: what's so wrong with nvidia proprietary drivers? they work does it not?
if you gonna argue with the current kernel version incompatibility than i can say the same for some open source drivers like for example the driver for my usb wifi adapter that it stoped worked after a new version was release 4.X to 4.X+1. not to mention the same open source driver that works on kernel 5 is a disaster compare to the old version. so yeah nice work for open source.

[loganj]

you forgot one extremely important thing: not every user is a programmer.
as for nvidia proprietary drivers i had no issue if i don't adventure in the "bleeding edge" kernel thing. the problem is that i'm trying to find out if the sound card issue was solved with the kernels 5+
it seems that kernel 5.0 works with nvidia so i was able to test those changes. unfortunately it doesn't fix anything.
the ctxfi driver is still as buggy as before for me (and so many others)

[loganj]

and again: not every person that use linux whants to be a programmer. most don't give a shit about programing. they want to linux just like windows. an OS that works without having to compile your own system of parts of it. with programs that can be installed without building them/etc.
and your advice to sell items that works...i'm speechless