Announcement

Collapse
No announcement yet.

AMD CES 2019 Keynote: Say Hello To Radeon 7 As 2nd Gen Vega

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

  • #31
    Originally posted by theriddick View Post
    I don't think this Radeon 7 card is really that great, lets see why.

    300W design, no chance of small form factor.
    $699USD is what the 2080RTX sells for.
    ...
    If this card was $500usd I'd be impressed.
    impossible the memory allone costs ~380$ for 16GB HBM2

    Comment


    • #32
      Yeah way too much for HBM2, I think AMD is crazy sticking to that stuff, just phase it out unless the supplier can change price. Isn't DDR6 less then half that price?

      Maybe not, 16GB of 3500mhz DDR6 is $416 or so... damn memory prices are still insane.
      Last edited by theriddick; 01-09-2019, 08:14 PM.

      Comment


      • #33
        Originally posted by tildearrow View Post
        Also, oh no... Why does Google want to provide *everything*??? They provide a search engine, image hosting, a calendar, a contacts list, mail, video calls, instant messaging, phone-associated IM, a social network, analytics, advertisements, a video platform, note-taking software, office suite, storage, hosting platform, home assistant, web browser, operating system, app store, bookstore, music store, movie store, phones, tablets, computers, dictionary, Usenet, domains, maps, blog platform, wallet, traffic status, a car, DNS, internet... just... Why?!
        Because once you have millions of users with an account and a default credit card for payments , AND you have tons of custom data centers with tons and tons of bandwidth, you do everything you can to maximize the customer relationships and the capital investment.

        That's why.

        Same reason Amazon, Microsoft, Facebook, Apple, etc. want to do the same thing. You have lots of users and lots of capacity. Load it up.

        Comment


        • #34
          Originally posted by nukedathlonman View Post
          I certainly think it could make an interesting alternative to the RTX 2080 depending on where the performance metrics fall. Has me curious for sure.
          On windows it matches the RTX 2080. On Linux it will fall short since Nvidia has faster Linux Vulkan drivers.

          AMD in Linux is performing great on Polaris which is fantastic for Linux gamers. But when you move to the higher end GPUs you feel the impact of the less effecient vulkan driver from the fragmented ecosystem of amdgpu pro, amdvlk and radV.

          Comment


          • #35
            Interesting, fix they current GPU bugs? (in the code we have no some workaround for Vega with effect to performance)

            Comment


            • #36
              Originally posted by Michael View Post

              AMD hasn't communicated with me about Radeon 7 (yet; sure hope to be receiving review sample...). But assuming it's the new Vega IDs I reported back in December, if so that was just PCI IDs added and no other code changes. If it's not those PCI IDs, then it wasn't introduced yet, but I assume it's those IDs especially with this Vega part shipping next month and so otherwise there will not be Linux driver support in time.

              If you will have the opportunity to review this card, please do not forget to check which is the speed of the double-precision computations.

              For the first time after many years, i.e. after Hawaii, AMD offers a GPU with high-speed (1/2 of SP) double-precision computations.

              We do not know if the 1/2 ratio is still true for the consumer product. If Radeon 7 still has the 1/2 ratio or even 1/4 or 1/8, it could have the best performance per dollar for DP computations.

              Unless you can afford a supercomputer, NVIDIA GPUs are useless for DP computations, because their cost per flop is higher than for Intel CPUs, so it is worthwhile to use them only if you are constrained by the electrical power consumption.

              A cheap AMD GPU able to do DP computations could change everything.

















              Comment


              • #37
                I've been on the AC Odyssey beta on Google Stream, and the performance isn't that great on my Comcast connection, and on a mid-range ASUS router. There is regular momentary graphics quality degradation, which I don't think can be solved without much better QoS guarantees from the router and network. I wouldn't want to give up playing on Steam/Consoles.

                Will be interesting to see what the I/O and processor dies have in 3rd gen Ryzen chiplets and how it's organized compared to the CCXs. If AMD is willing to let go of some profits at the low-end, I hope we see Vega 30/40+8C+I/O or 8C+8C+I/O or just the 8C+I/O on chiplet configs on AM4. AMD is going to kill it with 7nm though, I almost feel sorry for Intel, Nvidia.

                Comment


                • #38
                  AMD in Linux is performing great on Polaris which is fantastic for Linux gamers. But when you move to the higher end GPUs you feel the impact of the less effecient vulkan driver from the fragmented ecosystem of amdgpu pro, amdvlk and radV.
                  I know AMDGPU has performing quite well on Polaris - I've been running an MSI RX 570 Gaming X for a year and a half. :-)

                  Comment


                  • #39
                    Originally posted by theriddick View Post
                    Yeah way too much for HBM2, I think AMD is crazy sticking to that stuff, just phase it out unless the supplier can change price. Isn't DDR6 less then half that price?
                    No. DDR6 is 70% more than DDR5. It's the same shit, different day. There is no solution for AMD, other than to stay on old DDR5 tech. But they can't do that with their architecture, because the frequency bump comes from the 4GB HBM2 memory dies. The fewer they have, the slower the clocks. It had to be 4 X 4. The last generation is 2 X 4.

                    And while you guys may be crapping on AMD for the price, let's not forget their $800/$1300 competitor. Thirteen Hundred Fking Dollars.

                    What is important/remains to be seen is how 7nm deals with the voltage required for these cards. That is the primary problem. Turning your computer into a glorified toaster oven is no bueno.

                    One last thing people need to keep in mind.

                    For 4K gaming, if you actually looked at the fine print of those as yet unreleased titles- they REQUIRE 11GB of Video Memory. So, while you may think Nvidia is a more compelling option because it has features you will never fucking use(Ray Tracing), sooner rather than later, those 8GB cards are going to be literally spec'd out of the game. So do keep that in mind as well.

                    My only practical concern with Radeon 7 are the watts. The rest seems to be where it's supposed to be. Nobody is running RT on a 2080 at frame rates that are playable. ONLY the 2080ti can do that. So that "I gitz xtra shit wiff Nvidia!" is a nonstarter in total.
                    Last edited by ThoreauHD; 01-10-2019, 04:47 AM.

                    Comment


                    • #40
                      And AMD still target shrinking desktop market first and mobile last. In fact as it is I doubt I'd be getting any AMD powered laptop any time soon.
                      Intel meanwhile would probably present 8-core 45w laptops CPU's soon.

                      Comment

                      Working...
                      X