Announcement

Collapse
No announcement yet.

NVIDIA Comments On Its Linux Driver Architectue

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

  • #16
    Originally posted by droidhacker View Post
    They're looking to hire... time to send in the spys, to capture all their IP and release it for free on the internet

    ** then you poor nvidia users can finally have a decent driver.
    I'll be sure to tell my hardware that.

    "STOP WORKING AS ADVERTISED AND EXPECTED DAMMIT! PLEASE USE THESE OPENSOURCE DRIVERS AND SLOW DOWN!!! I WANT LESS PERFORMANCE PER WATT, FEWER FEATURES, AND NEUTERED DESKTOP EXPERIENCE! WHY ELSE DO YOU THINK I COMPILE MY KERNELS i386! I DON'T WANT NO OPTIMIZATIONS! F*CK SSE SUPPORT AND THE LIKES!"

    Comment


    • #17
      Originally posted by Xake View Post
      Yeah, but that "edge" is totally moot if that means you have to jump thru loops just to get the binary driver to work...
      ...or wait months to be able to use the latest version of a distribution because the binary provider fails to deliver a driver for the updated APIs/ABIs...
      Actually it's the opensource drivers that have been typically behind in supporting updated API/ABI. Just look at openCL, openGL, etc for good examples.

      ...or not being able to suspend your laptop without it crashing...
      NV blobs are working fine here for that.


      Yeah, it is a edge on a system while you are playing or doing a specific task, but if the driver does not handle "daily usage", that edge is still moot.
      I use blobs for daily usage every day and they work fine.

      PS the phase is "Jump through hoops"

      Comment


      • #18
        Originally posted by Xake View Post
        Oh, my. You really do not know your politics, ey?
        So when AMD merged with ATI to cut costs and time, they where pro-violent-regime (i.e. fascistic)?
        It seems like you don't get it. I wont reply to trolling.

        Comment


        • #19
          Originally posted by elanthis View Post
          They don't need huge teams. The drivers are the same between Windows and Linux, it's just a small bit of OS-specific code to handle. The "Windows team" for the driver development includes all the OS-specific stuff plus the core driver. Plus the DirectX support that Linux lacks.
          In truth, there is a minimal difference between "DirectX Support" and OpenGL support. Most of the functionalities are supplied via ARB extensions either as or before the next Direct3D version ships.

          By the way, there IS something to the remarks they made- just so you know. If you don't have enough people to fix the code core between OSes and APIs, you're not going to have as quality a result- regardless of the OS layer code being a small precentage. (We won't get into the fact that oftentimes it's that "small" code piece that's the source of the problems...)

          If they were so poor, why is it that two of the three "Next Gen" consoles are powered by AMD's silicon on the GPU side of things? It's not cost that motivated that.

          Those OSes are unsupported by any modern GPU.

          No kidding.

          It is, actually. Look how much people are clamoring for OpenGL 4 on Linux. The new APIs are better, more powerful, and developers actually want to use them. I am very much waiting for OpenGL 3 on the Linux FOSS drivers, in fact.
          Indeed.

          This makes no sense. They work on totally separate drivers for totally different hardware. The proprietary drivers are closed for business reasons and not a lack of manpower reason.
          Yes, they're closed for a business reason. The variable quality? That's mostly a manpower reason, coupled with some worldview issues at one of the companies in question.

          So far as the open drivers, NVIDIA doesn't even have a team doing those. Nouveau is all hobbyist/volunteer work, unlike the AMD driver.
          NVidia doesn't have anyone working on FOSS drivers. AMD doesn't have a team dedicated to the FOSS drivers- they're helping, but it's the community, from my understanding, that's doing the work on that one.

          Yes, AMD could put more people on the FOSS driver, but there's a business decision to be made there. They aren't making any significant extra income by having a FOSS driver. Hence, there's no reason to put more money into it. And I don't blame them for that; if there were a large enough number of desktop Linux users who actually demanded FOSS drivers, the market pressure would change. As is, there are a small number of desktop Linux users and most of them are either happy with weak drivers (or just weak hardware *cough* Intel *cough*) or are happy with proprietary drivers.
          In truth, I wish they'd put more effort into their drivers in general- FOSS would be perfect, but I'll accept better results with their closed drivers while the FOSS drivers gel. You indicate that they're not making significant money with the FOSS story, but that's more a chicken and egg problem. Need the drivers to have a demand for them...and there's going to be a bit more Linux in the future for us.

          As opposed to the FOSS community which has hundreds of thousands (if not millions) of developers... and yet there still is a tiny handful of graphics stack developers. Number of employees is irrelevant; number of skilled, experienced, motivated developers is what matters. I'm willing to bet NVIDIA has AMD trounced in this area, comparing the relative quality of their proprietary drivers (on any OS).
          Ah...there you go. You've hit upon the real issue. Somewhere along the line before AMD bought ATI, ATI had slashed it's staffing considerably to try to survive when the downturn from the 2000/2001 bubble burst impacted the PC industry and their sales suffered. They weren't back at the "good" staffing levels when AMD bought them- and it'll take a bit of time yet to get it back up there as they've got to get justification for the people and then find them.

          While AMD has some of the best and brightest working for them, they don't have enough of them right at the moment and it's...difficult...to find good people that can do the work and it takes quite a bit of learning curve to honestly get some of what needs to be known to be able to work on modern GPUs (Which is why we're seeing slow, steady progress from the FOSS drivers- even if we want and need them to be progressing a bit quicker...)

          Comment


          • #20
            Originally posted by deanjo View Post
            Actually it's the opensource drivers that have been typically behind in supporting updated API/ABI. Just look at openCL, openGL, etc for good examples.
            Well that is one kind of API, but you also got that "new kernel version" or "new Xorg version" kinds of API/ABI...

            Comment


            • #21
              Originally posted by NSLW View Post
              It seems like you don't get it. I wont reply to trolling.
              I think I know what you where trying to get at, but I wanted to be sure about it before starting to bisect on why the whole xorg development or the linux kernel development is communistic by that same logic, and it did not sound like you approved of that "communistic" part (at least not when you paired it with fascism).
              But still this is a *nix forum you are commenting on, so either you are only here to mock the development idea about shared knowledge is good knowledge (and are thereby a troll by default), you are so stupid you do not really know what you are talking about, or you are just so vague a person can interpret you a bit how they like. I hope it is the last case.

              Comment


              • #22
                Originally posted by Xake View Post
                Well that is one kind of API, but you also got that "new kernel version" or "new Xorg version" kinds of API/ABI...
                As far as "new kernel version" or "new Xorg version" goes nvidia has an excellent record with keeping current. Usually the day (sometimes even before) a final kernel or xorg version is out there is also a nvidia blob driver out there to support it. With AMD, I do realize that this is a sore point with their blobs which are often lagging behind in those scenarios.

                Comment


                • #23
                  BTW Michael, you might want to correct the spelling in the title.

                  Architectue

                  Comment


                  • #24
                    They better should start supporting Optimus on Linux (and with support I mean at least make "either or" working), else this gonna be a disaster for Nvidia on Linux.

                    Comment


                    • #25
                      Originally posted by mirza View Post
                      Nice he confirmed my theory that problem with drivers is size of NVIDIA (and AMD) Linux team(s). After all, there must be someone to take care of poor Win98/ME users and how it works for them! And to make sure one of them has DX12 3 days earlier then the other one, thats something! Actually, NVIDIA Linux team and AMD Linux team should merge, then both guys together might actually create something open and useful, in the basement of one of those gigantic companies with thousands of employees. </flame>
                      It will be a cold day in Hell when that happens.

                      Comment

                      Working...
                      X