all 8 comments

sorted by: hot top controversial new old
[–] 13 points 7 months ago (5 children)

Noteworthy its for AMD graphics cards only.

  • source
  • hideshow 5 child comments
  • [–] 2 points 7 months ago (4 children)

    Are you sure? Nothing in the dev's blog post alludes to that, aside from being tested on AMD.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 5 points 7 months ago (2 children)

    GTT (gpu swap) is handled by the gpu driver, so only nvidia can see if they can add it to their closed source drivers. radv is the amd vulkan driver.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 5 points 7 months ago (1 child)

    I searched a bit and you seem to be correct, lol. Nvidia never implemented GTT (15 years after it was introduced in the kernel). But the feature from the article is still not exclusive to AMD, Intel should work as well.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 3 points 7 months ago

    RADV is a driver for AMD cards. The linked blog post https://pixelcluster.github.io/VRAM-Mgmt-fixed/ title is "Fixing AMDGPU's VRAM management for low-end GPUs" and the word nvidia does not appear once in it. Unless I am misunderstanding this, it looks like AMD only. If anyone has direct confirmation that this works on Nvidia too (proprietary or open source drivers), I will update my previous replies to reflect that.

  • source
  • parent
  • [+] 5 points 7 months ago* (last edited 6 months ago) (1 child)
  • 31 comments in c/pcgaming@lemmy.ca view that post
    [–] 55 points 7 months ago (15 children)

    Valve are cool, love that they do this.

    I know it helps them, but they have done so much for Linux gaming it’s incredible.

  • source
  • hideshow 15 child comments
  • [–] 2 points 7 months ago (14 children)

    Just wish it was coming from a non-commercial entity. Puts a sour note on the status of Linux gaming that a for-profit entity is the only one out there making meaningful progress.

  • source
  • parent
  • hideshow 14 child comments
  • [–] 53 points 7 months ago (9 children)

    Lots of for-profit commercial entities contribute to open source projects.

    The code they're contributing is covered by the same license as the code contributed by volunteer developers.

    I understand why we should be cautious about these things, but the current situation is that Valve is contributing a lot and their contributions are open source. Yeah, they're doing it for a profit motive, but not to the point where they're trying to kill open source projects or hide the updates behind proprietary binaries.

    Valve is, currently, not being evil. GabeN has plenty of yacht money.

  • source
  • parent
  • hideshow 9 child comments
  • [–] 17 points 7 months ago* (last edited 7 months ago) (1 child)

    Yup, long as it's copyleft (GPL) open source, I don't care if it's microslop paying.

    That said, watch out for a new wave of EEE (embrace, extend, extinguish) using unmaintainable AI code, and be ready to fork.

    Valve, not so much.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 5 points 7 months ago (4 children)

    Gabe really is one of the exceptions that not all CEOs are evil bastards.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 5 points 7 months ago (3 children)

    Nah. No one, not even Gabe Newell, deserves a billion dollars. The very act of having that much wealth when there are still hungry and homeless people is in itself evil.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 7 months ago (2 children)

    I guess. Does he not donate to any charities or anything?

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 7 months ago (1 child)

    Right, but we look at examples like Chromium and we can see where there is still so much potential for things to go sideways. GabeN and his yacht could sink to the bottom of the sea and his estate sells control of Valve to someone less benevolent.

    A commercial entity that has enough control over a project pushes the direction of that project in their favor. And sure you can fork a FOSS project at any time, but once the commercialized version has enough saturation, user inertia and lack of experienced developers to take that initiative often prevents alternatives from achieving success.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 4 points 7 months ago

    I think the big difference is Valve isn't really in control of many of the projects they're funding, they're mostly just bringing in existing maintainers as contractors and letting them work on what they want.

    Chromium on the other hand has always been something Google has explicitly been in direct control of.

  • source
  • parent
  • load more comments (1 reply)
  • [–] 20 points 7 months ago

    Seemingly 2 other people worked on it as well, one from Red Hat and one from Intel.

    Also bear this in mind:

    All the user-space utilities hard-depend on systemd. Without systemd, you’d need to write your own utilities that make use of my kernel patches. Something needs to manage cgroups in your system, and that something needs to enable the right cgroup controllers and set the right limits (see also the long-winded explanation about how this works).

  • source
  • [–] 6 points 7 months ago (5 children)

    So this is just for AMD cards, correct?

  • source
  • hideshow 5 child comments
  • [–] 5 points 7 months ago (2 children)

    RADV is driver for AMD. So I assume it is for AMD cards only.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 4 points 7 months ago

    She answered that in her blog post that the Phoronix article links to:

    Which GPUs does this work with? Is it only AMD GPUs?

    Whether or not your GPU can benefit from it depends on the kernel driver - more specifically, whether it sets up the dmem cgroup controller.

    amdgpu and xe both have support for the dmem cgroup controller already. In theory, Intel GPUs running the xe kernel driver should benefit as well, although I’m not sure anyone tested this yet.

    For nouveau, I have sent a patch for dmem cgroup support to the mailing lists. This patch is also included in my development branch, so if you use my AUR package it should work. In other cases, you will need to wait for the patch to be picked up by your distribution, or apply it yourself.

    The proprietary NVIDIA kernel modules do not support dmem cgroups yet, so this won’t work there.

  • source
  • parent
  • [–] 3 points 7 months ago* (last edited 7 months ago)

    Maybe my GTX 1650 (4GB VRAM) gpu is not useless.

  • source
  • [–] 2 points 7 months ago (3 children)

    This is amazing because I have a GTX1660 and just switched to Catchy.

  • source
  • hideshow 3 child comments
  • [–] 1 point 7 months ago (2 children)
  • [–] 5 points 7 months ago (1 child)

    It's on the AUR but you need a patched kernel which you can grab from catchyOS

    Instructions are in the developer's blog post: https://pixelcluster.github.io/VRAM-Mgmt-fixed/

    Q: I use another Arch-based distro! What now?

    The dmemcg-booster and plasma-foreground-booster utilities are available in the AUR as well (plasma-foreground-booster carries the package name plasma-foreground-booster-dmemcg), so you can install them from there.

    For the kernel side, you can either use the CachyOS kernel package on a non-CachyOS system by retrieving the package from their repository, or you can compile your own kernel. Installing linux-dmemcg from the AUR will compile the development branch I used to develop this. Being a development branch, this carries the risk of some stuff being broken, so install at your own risk!

    If you want to apply the kernel patches yourself, you need these six .patch files: [links in blog]

    I’m not sure how easily they apply on specific kernel versions, but feel free to leave a comment if you run into issues and I’ll try to help out.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 2 points 7 months ago

    Thank you for looking into, appreciate the help. I wanted to read the article later to figure out the details. I think a custom Kernel goes a bit too far for my taste, so will then wait until its officially integrated.

  • source
  • parent
  • 21 comments in c/pcgaming@lemmy.world view that post

    Looks like it basically gives the focused application highest priority for VRAM allocations and deallocations, which lower VRAM devices would be doing more often.

    I’m kinda surprised this wasn’t already the default behavior, as the application the user has focused should almost always be given render priority, regardless of what it is.

  • source
  • [–] 52 points 7 months ago (13 children)

    "...limited amounts of video memory. Such as for graphics cards with just 8GB of dedicated vRAM..."

    "...just 8GB of dedicated vRAM..."

    "...just 8GB..."

    "...just..."

    "..."

    Y'all are spoiled if you don't understand how damaging that sentence is.

    "Just stop being poor! Y"

  • source
  • hideshow 13 child comments
  • [–] 20 points 7 months ago* (6 children)

    to be fair, budget 8gb gpus have been a thing since the RX 480 (2016). if somone has less nowadays, id reckon its refusal to buy used, or refusal to buy something thats not nvidias sub 200$ market.

    while yes, there of course are poorer regions. but even regions like brazil who have a ridiculous import tax on tech, still regularly stock the rx 580 as a "budget" gpu option.

    if you go low enough, the user might as well have an igpu, which does not have the same vram limitations discrete cards have.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 23 points 7 months ago* (last edited 7 months ago)

    Seriously. If you think 8GB of VRAM is "spoiled", you fell for NVIDIA's bullshit. You could get a used 1080 Ti with 11GB VRAM for $150 and it would be significantly faster than an RX 580.

    EDIT: you can downvote, but you still fell for it and got ripped off. Do some research before you make a major purchase.

  • source
  • parent
  • [–] -3 points 7 months ago* (last edited 7 months ago) (4 children)

    The *80 part meant high end...

    Edit: Yall either young or dumb. Back then there was only one 90, 290. And it was two 280 chips on one board. Amd didn't sell anything above 80. It was literally high end. Nvidia was like 10% better and 30% more expensive at high end. rx480 was literally the highest end leading flagship latest and greatest amd gpu when it came out.

    Kids these days don't even know what mid tier means...

  • source
  • parent
  • hideshow 4 child comments
  • [–] 4 points 7 months ago* (2 children)

    not for amd on random generations.

    polaris launched at 200$, and that was their top end gou because they forgone higher end models that generation, and opted for targeting the mainstream market. the 580 was just a refresh of it that allocated more of that stock to 8gb models. it was the best selling amd gpu (by far) because they made a lot of them to essentially flood the market.

    if youre judging high end by name and by market segment, you have already fell into marketing, as names are completely arbitrary.

  • source
  • parent
  • hideshow 2 child comments
  • [–] -1 points 7 months ago* (last edited 7 months ago) (1 child)

    Every 80, even the 480 that is the 580. 580 is just a higher clocked 480, and 470 is 480 with a gew cores shut down.

    And how should I judge whst is high end if not by where it sits by performance, price, and how they market it.

    At the time it was high end, as I said and contrary what he said.

    It wasn't highest end, that was nvidia. But it was definitely, by performance, high end.

    The insane prices (and power usages) now, combined with constant hypes, have made people not really objective.

    1060 was a great mid card when it came out. 6gb vram. 8gb is mid going towards high end now. I play with 3D rendering for a while now and I wish 16gb was the norm, but it just isn't.

    PS Those amd cards worked a bit differently then nvidias. They were harder to optimize for. And nvidia had more money to trow at driver optimizations and working with big game and game engine studios. So the public view of amd cards was worse. That's why they said later for amd gpus to "age like fine wine". When crypto came, they all bought amd cards (easier to write mining kernels since crypto is much simpler then graphics).

    Doom 2016 is optimized for those gpus...

  • source
  • parent
  • hideshow 1 child comment
  • [–] 3 points 7 months ago

    200$, even by 2016 standards was not remotely high end. so you couldnt even use it as price.

    when the AMD card was on the tier of performace of the 1060, and if THAT is considered "high end" then what is the 70 class gpus. what is the 80 class gpus. what is the titan levels.

    when you have 3 distinct tiers above you not even including any possible super or Ti varients of sort, how could you remotely consider it "high end" when you yourself say the 1060 is mid tier.

    at a price and performance, the 480/580 is not high end back when it launched, and definitely not high end anytime soon, hence 8gb isnt "high end" in 2026 unless someone has a very short list of gpus they are willing to use. I specifically use the 580 as an example because its actively still being used in poorer regions as the budget card, that actively has 8gb.

  • source
  • parent
  • [–] 5 points 7 months ago* (last edited 7 months ago)

    Be mad at the manufacturers who royally ripped you off with overpriced e-waste that only had 4GB, 6GB, or 8GB VRAM, instead of getting mad at those who are acknowledging reality honestly. None of these 8GB cards have been worth buying for at least 5 years. 6GB was top tier 11 years ago. It is 2026.

  • source
  • parent
  • [–] -2 points 7 months ago (4 children)

    I absolutely love the replies by people demonstrating the very privilege you're talking about. It's almost like you set a trap and they just keep walking into it.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 4 points 7 months ago (3 children)

    Privilege is when you aren't coping with buyers remorse because you didn't lap up overpriced e-waste that was already obsolete

  • source
  • parent
  • hideshow 3 child comments
  • [–] 0 points 7 months ago (2 children)

    Well, yes. It's literally one of the most commonly cited examples of privilege. Vimes' Boots Theory. A person who can afford to spend £100 on boots that last 5 years spends less than a person who spends £30 on boots that last 1 year.

    (A common misunderstanding being that the more expensive purchase is someone being "wise", when instead the person buying £30 boots didn't have the privilege to purchase £100 boots)

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 7 months ago*

    not exactly the situation, as its more like, there are boots that are 100 that last 5 years, you can buy boots that cost 30 and last 1 year, but the market has used 5 year boots that is on its 3rd year for 30, but one still opted to buy the 30/1 boot.

    a tech situation irl is cars.

    you can buy expensive 70k new car. you can buy affordable 20k new car. but youre also forgetting that you can buy used older 70k car for a lot less. the 20k car is not by default, better than the older car.

  • source
  • parent
  • [–] 5 points 7 months ago (4 children)

    Anyone know where that wallpaper used in the article is from?

  • source
  • hideshow 4 child comments
  • [–] 22 points 7 months ago (3 children)
  • [–] 8 points 7 months ago (2 children)

    specifically the dark mode version. it's called "Sub-Arctic"

  • source
  • parent
  • hideshow 2 child comments
  • [–] 4 points 7 months ago

    Really? Right in front of my GTX 1060 3GB?

  • source