Announcement

Collapse
No announcement yet.

Editorial: Using NVIDIA On Linux For The First Time In 10 Years

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Guest
    Guest replied
    Originally posted by gamerk2 View Post
    NVIDIA didn't lie; the card contained 4GB VRAM, with the last .5 simply being accessed at a slower rate. Funny that people forget that architecturally, had NVIDIA NOT done this, they would have had to ship the card with only 3GB instead.
    They lied by omission. They've made no mention of this until hard-proof came from the community, and even ended up settling over a lawsuit on this. Please tell me you don't actually approve of NVIDIA's behavior here?

    Originally posted by gamerk2 View Post
    The NVIDIA drivers tend to be very stable (at least the WHQL certified ones). DX12/WDDM 2.0 seems to be giving them some issues, but not more then I'd expect for a new API/Driver architecture. Anyone remember the initial DX10 driver releases?
    There were a few drivers (WHQL-signed also) that caused temperature issues for users in the past. I have no experience as to how the drivers are nowadays though.

    Originally posted by gamerk2 View Post
    NVIDIA fully supports OpenCL. And before you scream "fanboy", I counter argue "Mantle". Both companies develop APIs on their own dime, and both are free to push them.
    Mantle was short-term and paved the way for the open Vulkan standard, and was also discontinued in-favor of Vulkan. The same cannot be said at all for CUDA, PhysX, Gameworks, and all the other vendor lock-in software NVIDIA pushes.

    Originally posted by gamerk2 View Post
    Gameworks does not gimp performance for AMD. Not optimizing !=Gimping performance.
    Gameworks titles prohibit the optimization and showing of code for competition GPUs (AMD and Intel). So sure, I guess they aren't intentionally gimping the performance directly, but come on, it's pretty clear what their intent is by doing this.

    Leave a comment:


  • starshipeleven
    replied
    Originally posted by cj.wijtmans View Post
    Because AMD sucks performance wise.
    i'm like.... WAT?

    Leave a comment:


  • AsuMagic
    replied
    Originally posted by vein View Post
    But still...the games using gameworks mysterically run slower on AMD cards....*sigh*
    yes because idiots complaining about performance are enabling gameworks features. nvidia is sabotaging hairworks on amd so much that you in fact get better performance on AMD than on nvidia by changing a setting in the AMD panel in TW3.

    Leave a comment:


  • cj.wijtmans
    replied
    Originally posted by vein View Post

    Depends on how you look at it, I supposem for me it is lying when you dont write out on the package of the card that 0.5GB is slower...since without that there is an implication of all of the memory beeing fast...



    Well, there has been a lot fo trouble with the windows 10 drivers...so much that people are switching to AMD. I personally know several people who have done this...but since personal experience shouldn't count as evidence, you can look at the market share... ~10% increase for amd this year...



    The difference is that mantle became vulkan and open... I would be VERY suprised if Cuda becomes open one day...



    But still...the games using gameworks mysterically run slower on AMD cards....*sigh*

    Look, I just reflected on why I don't buy Nvidia and why I don't like them as a company. It is my own preference and I am sorry, but I don't think you will be able change my mind about the matter... You can buy their cards if you want to, go ahead.
    Because AMD sucks performance wise.

    Leave a comment:


  • cj.wijtmans
    replied
    Originally posted by enrico.tagliavini View Post
    For those who might evaluate the same solution as Eric: I think RPMfusion approach for the nvidia driver is better. It's easier to understand (imho), more solid and semi-offically supported by Fedora community. It's not officially part of Fedora, however the maintainer of the package is part of Fedora. Also RPMfusion has a good reputation in the Fedora community, so simply put if you have issues you have more chances of finding help. Not to bash Negativo17, it's just my personal experience.

    If you want to try the nvidia driver from rpmfusion use the official guide as a reference: https://rpmfusion.org/Howto/nVidia

    Also note it mentions the Secure Boot issue and the 32 bit libraries issue. There is no dkms, just akmod version of the kernel module. For an RPM distribution as Fedora this is likely the best solution (also the akmod service comes enabled by default).

    Nevertheless it's still not trivial at all for the average user.
    Use a better(user friendly) distro then.

    Leave a comment:


  • enrico.tagliavini
    replied
    For those who might evaluate the same solution as Eric: I think RPMfusion approach for the nvidia driver is better. It's easier to understand (imho), more solid and semi-offically supported by Fedora community. It's not officially part of Fedora, however the maintainer of the package is part of Fedora. Also RPMfusion has a good reputation in the Fedora community, so simply put if you have issues you have more chances of finding help. Not to bash Negativo17, it's just my personal experience.

    If you want to try the nvidia driver from rpmfusion use the official guide as a reference: https://rpmfusion.org/Howto/nVidia

    Also note it mentions the Secure Boot issue and the 32 bit libraries issue. There is no dkms, just akmod version of the kernel module. For an RPM distribution as Fedora this is likely the best solution (also the akmod service comes enabled by default).

    Nevertheless it's still not trivial at all for the average user.

    Leave a comment:


  • vein
    replied
    Originally posted by gamerk2 View Post

    NVIDIA didn't lie; the card contained 4GB VRAM, with the last .5 simply being accessed at a slower rate. Funny that people forget that architecturally, had NVIDIA NOT done this, they would have had to ship the card with only 3GB instead.
    Depends on how you look at it, I supposem for me it is lying when you dont write out on the package of the card that 0.5GB is slower...since without that there is an implication of all of the memory beeing fast...

    The NVIDIA drivers tend to be very stable (at least the WHQL certified ones). DX12/WDDM 2.0 seems to be giving them some issues, but not more then I'd expect for a new API/Driver architecture. Anyone remember the initial DX10 driver releases?
    Well, there has been a lot fo trouble with the windows 10 drivers...so much that people are switching to AMD. I personally know several people who have done this...but since personal experience shouldn't count as evidence, you can look at the market share... ~10% increase for amd this year...

    NVIDIA fully supports OpenCL. And before you scream "fanboy", I counter argue "Mantle". Both companies develop APIs on their own dime, and both are free to push them.
    The difference is that mantle became vulkan and open... I would be VERY suprised if Cuda becomes open one day...

    Gameworks does not gimp performance for AMD. Not optimizing !=Gimping performance.
    But still...the games using gameworks mysterically run slower on AMD cards....*sigh*

    Look, I just reflected on why I don't buy Nvidia and why I don't like them as a company. It is my own preference and I am sorry, but I don't think you will be able change my mind about the matter... You can buy their cards if you want to, go ahead.

    Leave a comment:


  • gamerk2
    replied
    Originally posted by vein View Post
    * They lie to their customers (970)
    NVIDIA didn't lie; the card contained 4GB VRAM, with the last .5 simply being accessed at a slower rate. Funny that people forget that architecturally, had NVIDIA NOT done this, they would have had to ship the card with only 3GB instead.

    * They have bad drivers (on windows 10, there has been only trouble)
    The NVIDIA drivers tend to be very stable (at least the WHQL certified ones). DX12/WDDM 2.0 seems to be giving them some issues, but not more then I'd expect for a new API/Driver architecture. Anyone remember the initial DX10 driver releases?

    * They lock customers with technology (Cuda and almost no support for OpenCL)
    NVIDIA fully supports OpenCL. And before you scream "fanboy", I counter argue "Mantle". Both companies develop APIs on their own dime, and both are free to push them.

    * They deliberately make games run worse on competition cards (Gameworks...example: Witcher 3)
    Gameworks does not gimp performance for AMD. Not optimizing !=Gimping performance.

    Leave a comment:


  • Passso
    replied
    Originally posted by birdie View Post

    then this thread has so many raving linux/open source zealots, that's my last post here.
    champagne

    Leave a comment:


  • cyberwizzard
    replied
    I am slightly surprised at the effort it takes to get the nVidia driver going on Fedora: under (K)Ubuntu its a matter of using a GUI to get all proprietary drivers set up or simply installing it using apt/aptitude. I never had to do more than reboot once for it to start using the binary driver (nowadays).

    Leave a comment:

Working...
X