No announcement yet.

NVIDIA GeForce GTX 1060 Offers Great Performance On Linux

  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Originally posted by Passso View Post
    Wow that is so close to 980... for 249$ this is the best deal since a long time ago. Impressive.
    (closed drivers are ok for me, so no flame plz)
    Good luck finding one for $250. It'll be $300 thanks to Founders My Wallet Edition prices. The FE cards are Nvidia's way to condition you to pay more for their graphic cards. And since the 1060 is $300 and performs like a GTX 970, guess what else costs $300 and performs like a GTX 970? A GTX 970... In fact you can probably find them for a little over $200 nowadays thanks to the RX 480. Eat that Founders Edition!


    • #22
      Originally posted by bridgman View Post
      Citation needed. Comparable die size, comparable process, comparable performance (higher on newer apps, lower on older apps), slightly higher power draw but also more VRAM. What makes it "bad" ?

      When you find review sites that don't stick to the reviewers guide you get a more balanced picture.
      Bigger die, more heat, more power consumption, less performance. These are facts so the product is indeed inferior technically. Now you can still manage to sell it, but at the appropriate discounted price point, and for Linux users the open driver is a great bonus.
      About the motherboard, the ref design still uses a lot on the pcie bus, I'm not comfortable with that and would like to avoid it if possible.

      Honestly, I understand your fanboyish reaction in support to your employer, but this is still a bit pathetic.


      • #23
        Originally posted by faldzip View Post
        Aaaaand where is DOTA2 on Vulkan?
        Come on now, we can't have nvidia looking bad for all future games can we?


        • #24
          Originally posted by bokal View Post
          All I see is how amazing are the numbers with mesa on the RX480.

          That will be my next GPU.
          But the Fury is still much better than the rx480 if you can afford 100 bucks more. And if you search for actual prices you can find even a fury x for 420€. I just got a used one for 350€. Thats quit a bargain.

          @Michael: you really should try to include current prices from amazon or NewEgg for the perfomance / dollar comparison. The fury or gtx 980 would come out much better.


          • #25
            Originally posted by remm View Post

            The conclusion is the same everywhere: Polaris is plain bad from a technical standpoint compared to Pascal. Its price point can make up for that, and that's the end of the story. The open driver is very good for Linux however (finally ! as a GCN1 owner, I got nothing but promises back then ...), so that's another selling point. But in no event I would place the RX480 ref card in one of my PCs, it's too easy to run into motherboard issues without actively looking for them.
            LOL, Pascal is a horrible arch(can't even async compute) and Polaris is epic. You can see the DX12/Vulkan benches on windows for that. The linux driver stack still needs work.

            Edit: It is really more that Nvidia is like Microsoft. They wipe their ass with standards and cut corners to improve performance. Some devs have already complained about this with DX12 and Vulkan.
            Last edited by SaucyJack; 07-19-2016, 10:55 AM.


            • #26
              Raw compute performance per slot means something. The 1060/1070/1080 all offer much better performance in terms of real world compute real-estate as the monthly expenses of a machine built to house these cards is often more expensive for data center applications than the insignificant up-front costs of the cards themselves (think $/square foot).


              • #27
                Originally posted by atomsymbol View Post

                Windows benchmarks published by other sites are showing similar performance for GTX 1060 and RX 480 (and R9 390). This increases the probability that RX 480 Linux performance will be competitive to GTX 1060 after additional Mesa optimizations.

                In terms of power consumption, GTX 1060 wins.
                i meant if you test Vulkan doom for example, the RX480(nightmare 1080p TSSAA8) it trade blows closer with the 980TI but the 1060 is having a hard competition with the 980, so in properly coded games(heavy parallel code + vulkan) the RX480 has more muscle to spare but in OpenGL/DX11 the CPU overhead murder the RX480 performance. In DX12 is hard to tell right now because most games are actually DX11 lightly threaded engines with some DX12 code to improve some tasks.

                btw many review sites if you notice avoid tests where the 1060 looks bad, they are forcing the tests to 1440p or 4k where both cards are ROP strained, so wait some time until more serious detailed review come.(this cards are meant for 1080p after all)

                For now we can say the power difference is a bit smaller since the rx480 could be around 15% faster overall(best case scenario ofc) on hardware and the 1060 is probably more expensive too since i suspect the non founder edition will take a while to appear on the market


                • #28
                  Originally posted by jrch2k8 View Post
                  1060 is probably more expensive too since i suspect the non founder edition will take a while to appear on the market
                  Non-founder's editions are already available in some e-shops.


                  • #29
                    Originally posted by AndyChow View Post
                    I have no idea what to do at this point. The gtx 1060 looks great, but the rx 480 has the open-source drivers... Meanwhile, my hd 6850 is getting too old.
                    I'd go with the RX 480. Performance isn't completely where I would like it to be either, but they're in the middle of major changes to their driver model. This is one of the few instances where I would trust that performance is going to become significantly better over time. You also have to look at the tests that were run. Being competitive in the Unigine tests (which has often been hardware agnostic) but uncompetitive in certain game ports shows us that the quality of the ports are a big factor here. Even so, Xonotic shows us that the driver still has a bit of work ahead of it. It's a mix of the ongoing overhaul of the driver and poor game ports that's giving us most of the issues on linux. The older ports likely won't be updated, but every issue caused by the state of development of the driver itself should be cleaned up in due time.


                    • #30
                      Any confirmation if it is actually shipping to actual customers, or if it was just a paper launch?