Announcement

Collapse
No announcement yet.

20-Way Radeon Comparison With Open-Source Graphics For Steam On Linux Gaming

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

  • #21
    Originally posted by Kivada View Post
    It's both slow and old. It's one step off of the slowest card of the HD6 era but still only has 64-bit GDDR3.
    I think people mostly think about those 6450 ones, but with GDDR5 memory yeah there was these variants Those can get something like 80fps in CS:S but on Windows .
    Last edited by dungeon; 27 August 2014, 01:13 PM.

    Comment


    • #22
      Originally posted by brosis View Post
      So unfortunate for HD6950. It gets beaten constantly by HD6870. Although the market prices are nearly same, its clear that 6950 architecture is heavily underperforming now.

      It looks like the best bang for the buck is still HD5850/70 and HD6850/70.
      I might just be getting a Sapphire Radeon HD 6950 2GB GDDR5 as it looks like I can get a good deal on it second hand.

      Even with these results, it looks like it would be a decent step up from the Diamond Radeon HD 4670 1GB GDDR3 I am currently using, which has only just recently been showing its age to me.

      Comment


      • #23
        Originally posted by justmy2cents View Post
        just currrious, is the fact that 6450 is so slow bug in mesa or is this card really so bad?
        If you compare the processing power and memory bandwith with the other cards in the HD 6000 series then you will see why the 6450's are that bad. I don't know if Michael is using a DDR3 or GDDR5 card.

        http://en.wikipedia.org/wiki/Radeon_HD_6000_Series

        Comment


        • #24
          I'm sorry but regardless of its limitations an HD6450 and HD5450 should be getting at least 30fps on nearly all of the titles tested. I've owned an HD6450 for 2 years and have played tf2, portal, cs:source all with frames around 40-60+ fps (rarely tested them without vsync). Source engine games are not that punishing at all, almost any low end GPU or Intel graphics should play them at above 30fps, only exceptions being maybe Portal 2 on some settings and Dota 2.

          It's more likely there is a serious regression or some bug in the benchmarking.

          Comment


          • #25
            Originally posted by brosis View Post
            So unfortunate for HD6950. It gets beaten constantly by HD6870. Although the market prices are nearly same, its clear that 6950 architecture is heavily underperforming now.

            It looks like the best bang for the buck is still HD5850/70 and HD6850/70.

            The HD6950 HD6970 and HD6990 are the only 3 GPUs in existence that used VLIW4.

            There are millions more VLIW5 GPUs out there like the HD5870 and HD6870 which means that the code base for those cards is far more optimized. My guess is that the only reason the HD6900 series even works at all is because the architecture is similar enough to the older VLIW5 design that it wasn't too difficult to get it to light up.
            Last edited by Kivada; 27 August 2014, 01:26 PM.

            Comment


            • #26
              Originally posted by ModplanMan View Post
              I'm sorry but regardless of its limitations an HD6450 and HD5450 should be getting at least 30fps on nearly all of the titles tested. I've owned an HD6450 for 2 years and have played tf2, portal, cs:source all with frames around 40-60+ fps (rarely tested them without vsync). Source engine games are not that punishing at all, almost any low end GPU or Intel graphics should play them at above 30fps, only exceptions being maybe Portal 2 on some settings and Dota 2.
              That is correct .

              It's more likely there is a serious regression or some bug in the benchmarking.
              There is no regression, thing is called hyperz which is disabled by default because it is unstable in some situations - cause GPU lockups .
              You can enable it if it is stable for you.

              But there are some new patches on ML about the metter Marek posted recently, which can make hyperz slightly stablier, less artifacts prone and even more faster than it is now .
              Last edited by dungeon; 27 August 2014, 01:42 PM.

              Comment


              • #27
                thanks for reply and... nah, for gaming i use 750Ti. what i care with amd is fluid desktop experience with oss drivers.

                I was just curious since one of my desktop (no gaming) cards is 6450 and for those needs i always buy amd. seems so strange that it is completely beaten by all my other cards when it was the last one i bought now, if i compare price i payed for 750Ti and 6450 price/perf is abysmal and i bought them at same time. note that i don't really care how much slower oss vs catalyst is, nor how good NVidia is compared to it. 6450 card performs really well for what is meant to and oss drivers are a blessing. i just noticed i could probably buy better price/performance amd card if i checked things before buying, which bring it to... it's my own damn fault, lol

                Comment


                • #28
                  Keep in mind, small cards are slow but use the least power as well

                  Originally posted by justmy2cents View Post
                  thanks for reply and... nah, for gaming i use 750Ti. what i care with amd is fluid desktop experience with oss drivers.

                  I was just curious since one of my desktop (no gaming) cards is 6450 and for those needs i always buy amd. seems so strange that it is completely beaten by all my other cards when it was the last one i bought now, if i compare price i payed for 750Ti and 6450 price/perf is abysmal and i bought them at same time. note that i don't really care how much slower oss vs catalyst is, nor how good NVidia is compared to it. 6450 card performs really well for what is meant to and oss drivers are a blessing. i just noticed i could probably buy better price/performance amd card if i checked things before buying, which bring it to... it's my own damn fault, lol
                  If you do not use that desktop for gaming or 3d content creation, the only thing a bigger GPU brings you is more heat and higher power consumption. Even the smallest cards have the same video playback engine as the biggest, and if you edit video here's an upcoming thought: I tested the HD6450 against my HD 6750 in the Movit (GPU accelerated effects) version of Kdenlive that is being experimented with, and did not find the GPU to be a bottleneck even for realtime color correction on 1080p video. Remember that video is 2d, so only one Z value to be involved with and a hell of a lot less data points. The Movit developers used a GTS 660 Nvidia card for some of their work and were able to run multiple GPU accelerated effects on 4K video in realtime, I suspect smaller cards could also do this. The bottleneck for GPU compute seems to be memory copying from system to GPU RAM and back anyway.

                  Comment


                  • #29
                    Originally posted by Kivada View Post
                    The HD6950 HD6970 and HD6990 are the only 3 GPUs in existence that used VLIW4.
                    That's true for discrete GPUs, but the Trinity and Richland APUs also use VLIW4 GPUs.
                    Test signature

                    Comment


                    • #30
                      Originally posted by Kivada View Post
                      The HD6950 HD6970 and HD6990 are the only 3 GPUs in existence that used VLIW4.
                      The list of all such cards was recently posted into a similar thread. There are APUs and Firepros using it.

                      Comment

                      Working...
                      X