12 Replies Latest reply on Nov 5, 2017 9:07 PM by gc9

    AMD usage share plummets on Steam, from 17.1% to 10.8%, in October.

    black_zion

      And why is this important? Same reason it was after last month's large drop, AMD touts Vega specific features, but with that usage share, what incentive to developers have to use them? Decidedly not good news with the GTX 1070 Ti taking on Vega 56, with performance nudging the 1080. It's also not good for consumers either, as AMD is enabling prices to stay high.

       

       

      A quick look at the prices shows why. The RX 580 is still $40 higher than launch price (thank you cryptominers), and $20 more expensive than the same power GTX 1060

       

       

      And Vega 64 reference is still more expensive than the GTX 1080 custom, which pulls half the power.

       

        • Re: AMD usage share plummets on Steam, from 17.1% to 10.8%, in October.
          ajlueke

          Actually, you can pick up an 8 Gb RX 580 for $279.00.  While still $10 more expensive than a GTX 1060 and more power hungry, it isn't a horrible deal.

           

           

          Also Vega 64 has fallen back to it's MSRP of $499.

           

           

          But the numbers don't lie.  I think a lot of gamers were waiting to see what Vega brought to the table before purchasing new hardware.  It seems they now have decided they didn't like what they saw, either due to the performance, or the massive price inflation that has lasted since Vega's launch.

           

          With the numbers coming back to parity in time for the holiday season and the threat of Volta pushed into 2018, those numbers may shift back the other direction before the year is out.

          • Re: AMD usage share plummets on Steam, from 17.1% to 10.8%, in October.
            qwixt

            Steam usage isn't really that good of a metric because the user still needs to approve it. Whenever I am asked, I always say no. I don't care about providing statistical data.

             

            The reason why AMD usage might be dropping is that Vega is not a good value. Only now with the prices dropping is it starting to get close. From what I have investigated, I would only consider the vega 56 a good purchase value. I don't see much hope for custom vega 64 cards based on the asus card. I still find it hard to believe that a custom card could be so poor as that asus. The liquid cooled vegas 64s are at 1080 ti cost or more, yet perform more like a 1080+.

             

            If I had to buy one, it would be the 56, and I would install the 64 bios. But the reference cooler ruins it. The 64 LC I would buy since it is a bit quieter, but the cost performance ratio is way out of whack. I still would have thought a custom air solution could get close enough to the 64 LC performance, and at a cheaper price. One really curious thing I saw on youtube is a guy ran a fury or fury x, and vega 64 at the same clock speeds, and there was almost no difference in benchmarks. So that says to me that the drivers are either really bad still, or all those architectural changes were more about frequency improvements, and not IPC in reality.

             

            I hold no allegiance to AMD or nvidia, but I do like to support the underdog if there is a competitive option. Unfortunately, due to pricing and/or performance, AMD really doesn't have an option for me in Vega. The price of the 64 LC has to drop considerably, or a custom card, which we might hear more about in Nov. It's disheartening to say, but I don't see amd putting up much high end competition in the future for nvidia. The market really needs it because nvidia is shameless with its prices without competition.

            • Re: AMD usage share plummets on Steam, from 17.1% to 10.8%, in October.
              gc9

              The attention grabbing thread title is about market share on Steam.  That survey is dominated by midmarket cards and below.  All the discussion so far is about high-end enthusiast cards which make up only about 3% of the Steam surveyed users, so not very relevant to the title.

               

              The "Steam Hardware & Software Survey: October 2017" page also shows

              • a sudden 20% shift from Windows 10 to Windows 7, and
              • a sudden 25% shift to speaking Simplified Chinese.


              A possible conclusion: Some change suddenly opened up Steam to a huge number of Chinese players.

               

              The videocard detail page shows that the cards that increased marketshare are not Vega56 or Vega64 competitors.  Competitors for these cards, such as GeForce 1070, 1080, and 1080 Ti, also lost marketshare.  The cards that gained signficant marketshare include GeForce 750, 750Ti, 950, 960, 1050, 1050 Ti, 1060.  (Many older cards by several manufacturers made insignificant gains.)

               

              A possible conclusion: The huge number of Chinese players had many midmarket Nvidia cards, not competitors to Vega.

               

              RX 460 and RX 480 are listed with very low percentages.  RX 470, RX 470D, RX 550, RX 560, RX 570, RX 580 are strangely missing.  If Mindfactory public numbers are in very rough proportion, I'd expect something more like a 4:1 ratio (or 48:12) of midmarket GeForce (x50, x60) to Radeon RX Polaris, but this Steam report videocard subpage shows something more like 60:1.  RX460 is steady, but RX480 has been declining. 

               

              A possible question: Have almost all 2017 RX Polaris cards been bought up by (large scale not-gaming-on-Steam) compute customers, such as blockchain miners?  (Seems unlikely, especially for RX 550 and RX 560.)

              A possible question: Is something preventing 2017 RX cards from appearing in the Steam survey?   Could the survey be misrecognizing cards?

              What other evidence could be gathered?

              1 of 1 people found this helpful