Blender 4.1 Will Further Expand Linux's CPU Rendering Performance Lead Over Windows

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • Gipr
    Junior Member
    • Jul 2013
    • 14

    #21
    >> Linux devs implement Windows NT kernel APIs in the linux kernel to speed up windows games in wine.

    > This is so funny, in a sad way,

    > Not to mention it's complete nonsense, just a string of technobabble that says nothing.

    > LOL!


    qarium

    "LOL"... Now, when I think of sophisticles I can't avoid imagining him with brown skin and long curly hairs.

    Comment

    • sophisticles
      Senior Member
      • Dec 2015
      • 2591

      #22
      Originally posted by Gipr View Post
      >> Linux devs implement Windows NT kernel APIs in the linux kernel to speed up windows games in wine.

      > This is so funny, in a sad way,

      > Not to mention it's complete nonsense, just a string of technobabble that says nothing.

      > LOL!


      qarium

      "LOL"... Now, when I think of sophisticles I can't avoid imagining him with brown skin and long curly hairs.
      So you think the Ancient Greeks had brown skin and long curly hair?

      Comment

      • qarium
        Senior Member
        • Nov 2008
        • 3441

        #23
        Originally posted by sophisticles View Post
        This is so funny, in a sad way,
        Not to mention it's complete nonsense, just a string of technobabble that says nothing.
        LOL!
        it just proof 1 thing: linux does everything for performance and microsoft and apple does nothing for performance.
        Phantom circuit Sequence Reducer Dyslexia

        Comment

        • qarium
          Senior Member
          • Nov 2008
          • 3441

          #24
          Originally posted by Panix View Post
          Most ppl don't use the CPU for Blender - you know that. You can buy an expensive Threadripper....
          You know I was talking about HIP performance vs Optix - since, Nvidia has moved from using CUDA (which you could still use, in theory) to Optix - so, using the ray tracing acceleration. Optix w/ even previous gen Nvidia cards smoke the current gen flagship AMD gpus in Blender right now. AMD / Blender was striding to 'officially' implement ray tracing for AMD gpus since Blender ver. 3.5 and here we are, talking about Blender version 4.1 now. Pretty sad.
          you can only compare the performance of something if the result is the same.

          you can not compare HIP blender output with OptiX blender output because the result is not the same.

          and HIP produce the correct result and OptiX is not the correct result.

          "since, Nvidia has moved from using CUDA (which you could still use, in theory) to Optix"

          this statement is plain and simple wrong because the CUDA output is the mathematical correct output and exactly the same as the CPU output
          and OptiX is not the same output and is mathematically not the same.

          Nvidia did not move from CUDA to OptiX in blender ... OptiX is just a option for people who do not need the correct mathematical output.

          "Optix w/ even previous gen Nvidia cards smoke the current gen flagship AMD gpus in Blender right now."

          grow up just grow up. you can not compare performance if the result is not the same.
          Phantom circuit Sequence Reducer Dyslexia

          Comment

          • Panix
            Senior Member
            • Sep 2007
            • 1555

            #25
            Originally posted by qarium View Post

            you can only compare the performance of something if the result is the same.

            you can not compare HIP blender output with OptiX blender output because the result is not the same.

            and HIP produce the correct result and OptiX is not the correct result.

            "since, Nvidia has moved from using CUDA (which you could still use, in theory) to Optix"

            this statement is plain and simple wrong because the CUDA output is the mathematical correct output and exactly the same as the CPU output
            and OptiX is not the same output and is mathematically not the same.

            Nvidia did not move from CUDA to OptiX in blender ... OptiX is just a option for people who do not need the correct mathematical output.

            "Optix w/ even previous gen Nvidia cards smoke the current gen flagship AMD gpus in Blender right now."

            grow up just grow up. you can not compare performance if the result is not the same.
            Yes, you can compare - because there's no alternative to just using HIP right now - while with Nvidia - you can use CUDA OR OptiX. It's not just the two features or performance differences, it's about an option that isn't available with the AMD side - one that they've been trying to optimize and implement for a while - but, it's still not working. Even if it does, a large gap in performance (still) will be a major disappointment.

            But, the unique negative here isn't just the performance gap but how long they've been working on it for - and the fact, it's still not an option - or at least, nothing that has been measured or tested in a thorough, fully scrutinized way - since, this site and others would definitely do so if it was 'ready' and "complete." Which, it is not - it's not about mathematical output or not - it's about features which are available and work or not.

            Comment

            • qarium
              Senior Member
              • Nov 2008
              • 3441

              #26
              Originally posted by Panix View Post
              Yes, you can compare - because there's no alternative to just using HIP right now - while with Nvidia - you can use CUDA OR OptiX. It's not just the two features or performance differences, it's about an option that isn't available with the AMD side - one that they've been trying to optimize and implement for a while - but, it's still not working. Even if it does, a large gap in performance (still) will be a major disappointment.

              But, the unique negative here isn't just the performance gap but how long they've been working on it for - and the fact, it's still not an option - or at least, nothing that has been measured or tested in a thorough, fully scrutinized way - since, this site and others would definitely do so if it was 'ready' and "complete." Which, it is not - it's not about mathematical output or not - it's about features which are available and work or not.
              ROCm/HIP Blender backend is a alternative to CUDA not optiX...

              HIP-RT is the alternative to OptiX why not compare this and of course AMD works on HIP-RT...

              remember i told you about RDNA4 release and i said to you 2 months and you said no way and now there are news out there about RDNA4 the linux kernel already has driver pushing upstream for RDNA4 and also MESA ans so one.

              and i am pretty sure that HIP-RT will have a full featured release together with the RDNA4 hardware release.

              so your commend is only month away of being obsolete. so enjoy the months you have left.
              Phantom circuit Sequence Reducer Dyslexia

              Comment

              • qarium
                Senior Member
                • Nov 2008
                • 3441

                #27
                Originally posted by Gipr View Post
                >> Linux devs implement Windows NT kernel APIs in the linux kernel to speed up windows games in wine.
                > This is so funny, in a sad way,
                > Not to mention it's complete nonsense, just a string of technobabble that says nothing.
                > LOL!

                qarium
                "LOL"... Now, when I think of sophisticles I can't avoid imagining him with brown skin and long curly hairs.
                i just got him ... linux does all for performance and apple and microsoft does nothing for performance and then people like sophisticles wonder why macod and windows is slower.
                Phantom circuit Sequence Reducer Dyslexia

                Comment

                • Panix
                  Senior Member
                  • Sep 2007
                  • 1555

                  #28
                  Originally posted by qarium View Post

                  ROCm/HIP Blender backend is a alternative to CUDA not optiX...

                  HIP-RT is the alternative to OptiX why not compare this and of course AMD works on HIP-RT...

                  remember i told you about RDNA4 release and i said to you 2 months and you said no way and now there are news out there about RDNA4 the linux kernel already has driver pushing upstream for RDNA4 and also MESA ans so one.

                  and i am pretty sure that HIP-RT will have a full featured release together with the RDNA4 hardware release.

                  so your commend is only month away of being obsolete. so enjoy the months you have left.
                  Are you really that obtuse?!? HIP - is slow in Blender - look at the Opendata site on Blender and compare to OptiX w/ Nvidia and any Nvidia cards. CUDA was chosen before but now that Optix fully works, it allows for much faster renders. The other thing about AMD gpus in Blender - are tons of complaints of crashes, stutters and the render not finishing. So, even with HIP - which is the only/main option for AMD gpus at the moment, it's barely usable for a lot of ppl. I showed you the developer forums and what ppl are saying. So, stop shilling for AMD and just read what I am saying and check for yourself.

                  I don't care about RDNA 4 when RDNA 3 - the current gen. is so poorly optimized and is only good for gaming - in Linux and in Windows - both. It looks like AMD is prioritizing AI so I guess they decided on another area in which to concentrate on - so, yeah, maybe there might be a coincidence and other programs might benefit from some improvements - but right now, AMD is not very good at the productivity realm including in Linux.

                  When ppl on here rave about AMD gpus in Linux - practically all of them are talking about gaming - not Blender, not ROCm, not anything outside the gaming sphere. That's pretty bad when their flagship cards are well over $1000 in many countries.
                  Last edited by Panix; 09 February 2024, 03:46 PM.

                  Comment

                  • qarium
                    Senior Member
                    • Nov 2008
                    • 3441

                    #29
                    Originally posted by Panix View Post
                    Are you really that obtuse?!? HIP - is slow in Blender - look at the Opendata site on Blender and compare to OptiX w/ Nvidia and any Nvidia cards. CUDA was chosen before but now that Optix fully works, it allows for much faster renders. The other thing about AMD gpus in Blender - are tons of complaints of crashes, stutters and the render not finishing. So, even with HIP - which is the only/main option for AMD gpus at the moment, it's barely usable for a lot of ppl. I showed you the developer forums and what ppl are saying. So, stop shilling for AMD and just read what I am saying and check for yourself.
                    I don't care about RDNA 4 when RDNA 3 - the current gen. is so poorly optimized and is only good for gaming - in Linux and in Windows - both. It looks like AMD is prioritizing AI so I guess they decided on another area in which to concentrate on - so, yeah, maybe there might be a coincidence and other programs might benefit from some improvements - but right now, AMD is not very good at the productivity realm including in Linux.
                    When ppl on here rave about AMD gpus in Linux - practically all of them are talking about gaming - not Blender, not ROCm, not anything outside the gaming sphere. That's pretty bad when their flagship cards are well over $1000 in many countries.
                    in germany the 7900XTX is 950€ ...

                    it has nothing to do with beeing obtuse i just only compare stuff what has 100% the same output. if you do not like the mathematical correct output you are free to go with optiX.

                    its very interesting if Nvidia launches a new gpu generation you praise it as gold but if AMD is in the launch of RDNA4 you say it is shit.

                    "It looks like AMD is prioritizing AI"

                    RDNA4 adding raytracing hardware that has nothing to do with AI. you can use this hardware only for games or software like Blender
                    so it is complete insanity from you to claim RDNA4 is focused on AI... CDNA is focused on AI... but RDNA4 is focused on games and also Blender.

                    "When ppl on here rave about AMD gpus in Linux - practically all of them are talking about gaming - not Blender,"

                    this is really nonsense the only AMD hardware who does not care about Blender and also does not care about gaming is CDNA
                    even if you have CDNA mi300 you can not use HIP-RT you are forced to use HIP---

                    but with RDNA4 you will have the option to use HIP-RT similar to OptiX-..
                    Phantom circuit Sequence Reducer Dyslexia

                    Comment

                    • Panix
                      Senior Member
                      • Sep 2007
                      • 1555

                      #30
                      Originally posted by qarium View Post

                      in germany the 7900XTX is 950€ ...

                      it has nothing to do with beeing obtuse i just only compare stuff what has 100% the same output. if you do not like the mathematical correct output you are free to go with optiX.

                      its very interesting if Nvidia launches a new gpu generation you praise it as gold but if AMD is in the launch of RDNA4 you say it is shit.

                      "It looks like AMD is prioritizing AI"

                      RDNA4 adding raytracing hardware that has nothing to do with AI. you can use this hardware only for games or software like Blender
                      so it is complete insanity from you to claim RDNA4 is focused on AI... CDNA is focused on AI... but RDNA4 is focused on games and also Blender.

                      "When ppl on here rave about AMD gpus in Linux - practically all of them are talking about gaming - not Blender,"

                      this is really nonsense the only AMD hardware who does not care about Blender and also does not care about gaming is CDNA
                      even if you have CDNA mi300 you can not use HIP-RT you are forced to use HIP---

                      but with RDNA4 you will have the option to use HIP-RT similar to OptiX-..
                      Wrong again. Show me where I said something praising the 50 series? I didn't make any comment about it since it's not released yet. Also, I am not a fan of Nvidia - the company or its practices - I only look at performance of the hardware with whatever I am (going) to use.

                      I prefer something that works well and especially, if it's expensive. RDNA 4 is redundant in this discussion as a I told you before - it's not released yet. You can speculate and I can as well - but, the current gen - we know about its current status and that's what I am mostly talking about. Nvidia uses OptiX for Blender - and its performance is way better than AMD - if you had a 4070 Ti which is cheaper than any 7900 XT or XTX right now - it'll be much better performance for the $$. That's just one example. Even in AI, Nvidia - it's argued it's a better experience. I share the same sentiments that it's way better to have an open source option and in the Linux ecosystem, I also assert it's a better fit - but, the performance isn't there for these applications. I think it's bad that so many give AMD a free pass on this simply because, 'ya know, open source....bro.'

                      Edit: You don't have any taxes in Germany? I didn't know that. Wow. /s
                      Last edited by Panix; 09 February 2024, 04:35 PM.

                      Comment

                      Working...
                      X