AMD investor relations slides offer a glimpse of chips to come

Can you really call something a leak if the company released it on purpose? AMD's just released a slide deck for its investors, and buried in those slides are a few tiny nuggets of interesting information. Let's take a quick peek into the red team's path ahead.

First up is this slide, which notes that "Pro" versions of the second-generation Ryzen mobile chips are on the way here shortly. This slide also puts down the fascinating third-generation Ryzen desktop chips for "mid-year." In the strictest sense, that isn't news, but it's still really nice to see AMD—a company with a history of difficulty in keeping deadlines—on track with its new hardware, especially given the unusual nature of the design.

That same slide also indicates that the Threadripper chips are coming this year, although the company didn't narrow it down any further. We'd expect them to follow along shortly after the desktop chips. Once again, it's not really a surprise that Threadripper the Third will emerge in 2019, given the close relation of Threadripper and EPYC CPUs. We already knew the Zen 2-based EPYC chips would be coming around before long, after all.

There's also this slide above. It's not nearly as interesting as the first slide, but it does seem to confirm another piece of news we already knew: Navi is on the way this year. We really don't know much about AMD's next graphics processors, but scuttlebutt says that AMD could launch as many as three separate cards, possibly as soon as E3 2019 this summer.

This slide also notes that AMD intends to launch a "next-gen" graphics part next year. Prior rumors had stated that AMD planned to launch low-end Navi cards this year, with a more powerful variation on the same theme coming in 2020. Given this slide, we're now more inclined to believe that AMD is going to launch Navi as a mid-range part this year, and then follow up with high-end cards based on "Arcturus" next year. That's not too far from what the company attempted with Polaris and Vega, really. Let's hope Navi and Arcturus can revive competition in the graphics market.

Comments closed
    • Srsly_Bro
    • 5 months ago

    Intel is coming out with Cascade lake next month and tsmc is in big trouble. I’m worried for AMD now. Raja had better move to his time line before it’s too late.

      • chuckula
      • 5 months ago

      That defective batch of photoresist was supposed to go to Intel’s 10nm fab!

      Raja needs to be careful with the address form auto-fill in his web browser since he changed jobs!

        • Wirko
        • 5 months ago

        This comment made it to the Top Comments list before getting any upvotes. Your exactly timed attack on TR’s speculative-predictive sorting algorithm has fully succeeded!

          • K-L-Waster
          • 5 months ago

          Just call him the Spectre.

    • ronch
    • 6 months ago

    Sounds like Navi is yet another iteration of GCF (Graphics Core Forever).

      • tipoo
      • 5 months ago

      I recently had a somewhat confusing chat that changed my mind a bit on this. GCN, as the ISA, isn’t imposing the limits that makes people wish for a move past it, that’s architectural. You could keep the GCN ISA and remove the four shader engine limit, as that’s on the uArch and not GCN.

      We’ll have to see if Navi takes the chance to do that, but afaik we have no indication it’s going higher end than VII.

        • ronch
        • 5 months ago

        All this Time we thought of GCN as the fundamental graphics architecture, like how the K10 is fundamentally the K7 tweaked and evolved over the years. Up until this time I don’t know anyone who said that GCN is the ISA. GCN is RISC-based though, having departed from the VLIW5 and later VLIW4 ISAs of the HD2000 up until the HD6000 series. Correct me if I’m wrong though.

          • ET3D
          • 5 months ago

          The question is, what’s fundumental to the GCN architecture? There were changes to the scheduler, the cache architecture, addition of tiled rendering, an attempt to add primitive processing…. I don’t know all the changes, because I couldn’t find any detailed documentation of GCN beyond the initial release.

          What is it apart from the ISA that makes CGN CGN?

          If, for example, a CU will have not 4 SIMD units but, say, 8, and everything else remains the same, would that still be considered GCN or not?

            • tipoo
            • 5 months ago

            ^That’s what I was getting at. At first GCN was both the name of the ISA and the architecture, but key architectural changes have been made while remaining GCN. If the architecture is changing, but the ISA isn’t, I guess we’re referring to the collective family by the GCN ISA, but some of the limits assumed to be GCN side are architectural.

    • qmacpoint
    • 6 months ago

    and perhaps out of nowhere AMD may end up with DXR support!!!

    Not gonna happen, not at those roumoured prices.. 🙂

      • tipoo
      • 6 months ago

      They could support it, it just won’t have especially useful performance without dedicated acceleration. I think it was something like 10 vs 350fps in UE4 on AMD vs Nvidia cards

      [url<]https://www.reddit.com/r/nvidia/comments/atxgoq/directx_raytracing_reflections_in_unreal_engine_4/[/url<]

      • DoomGuy64
      • 5 months ago

      Gamers Nexus did a video with a custom UE engine “game” showing that high end rasterization isn’t much worse than DXR, while the noise and performance loss wasn’t worth it for dxr.
      [url<]https://youtu.be/CuoER1DwYLY[/url<] Then, crytek released a raytraced demo running real time on a Vega 56, and is compatible with any non RTX card. [url<]https://youtu.be/1nqhkDm2_Tw[/url<] Doesn't quite look like raytracing is the "killer app" that it's being touted as. It's more of a developer needing to put work into the engine issue, than a necessary hardware feature. Especially when there appears to be more efficient and higher quality methods available as alternatives. RTX isn't the final solution, it's one available method out of many other methods, and all it really did was inspire everyone else to start inventing their own methods. Raytracing is going to take years before anyone solidifies a standard, and proprietary walled garden solutions are never going to be that standard, although RTX will probably run said future standard faster.

    • The Egg
    • 6 months ago

    Headline had me confused. I read it as “AMD Investor Relations Slide….”

      • KeillRandor
      • 6 months ago

      Me too: “AMD investor relations slide after glimpse of chips to come” is what I initially thought it said…

      • Canuckistani
      • 6 months ago

      I had to skim the post again after reading it thinking I had missed the bad news. Glad to see I wasn’t the only one who misread the headline.

      • Wirko
      • 6 months ago

      Come to chips of glimpse, a offer slides relations, investor AMD!
      (a little out-of-order will hurt no one, especially not Advanced Micro Devices)

      • Oem
      • 6 months ago

      AMD INVESTOR RELATIONS SLIDES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

      . . . SHOW SOME REALLY GOOD NEWS, INVESTOR RELATIONS BETTER THAN EVER

      Thank you for clicking on this [s<]bait[/s<]post.

      • CScottG
      • 6 months ago

      Yup! I blame it on chuckula..

        • chuckula
        • 5 months ago

        AMD can make their own investor relations slide ALL BY THEMSELVES.

    • Chrispy_
    • 6 months ago

    C’mon AMD, we need new architectures.

    2nd Gen Ryzen Mobile? That’s just a rebrand of the 1st gen.
    AMD-Powered Chromebooks? That’s a 2016 product that you’ve underclocked. It’s 28nm ffs!
    Athlon Mobile APU? That’s just a rebranded 2200U

    We’re not idiots. Claiming these as “achievements” is really sad, because these rebrand placeholders aren’t fooling anyone.

    As for the upcoming “non-stop product momentum” (/cringe), the 2nd Gen Ryzen Pro Mobile is just an atificial price/driver segmentation. Nothing new there either. Just give us a release date for Zen2 (AKA 3rd-gen Ryzen) and stop with the fake news.

      • chuckula
      • 6 months ago

      Oh please, like anybody cares about mobile.

      16 Coar RyZen desktop? CHECK!

      And I’ll pay extra for AMD courageously not putting in those dumb “integrated” graphics that those idiotic shills at Intel marketing pretend are used anywhere.

      64 Coar Ripper 3? DOUBLE CHECK!

        • Chrispy_
        • 6 months ago

        I’m in the market for a 3700X if the leaks are accurate. 5.0GHz? 24 threads? Sign me up.

        I just wish they’d stop all this “LOOK AT OUR NEW PRODUCT LABELS” news, when the last new thing they actually made was the WX2990.

          • cynan
          • 6 months ago

          +1. Sounds like a great potential replacement for my ailing Sandy Bridge-E.

          • chuckula
          • 6 months ago

          [quote<] I just wish they'd stop all this "LOOK AT OUR NEW PRODUCT LABELS" news, when the last new thing they actually made was the WX2990.[/quote<] Radeon VII would like to quote Rodney Dangerfield now.

            • Chrispy_
            • 6 months ago

            Hmmm good point; We were sampling an Instinct MI50 a couple of months [i<]after[/i<] the 2990WX. And no, I refuse to call Radeon VII a new product even if they did bolt a 'new' display output onto their MI50 to sell it to consumers 😉

            • Srsly_Bro
            • 6 months ago

            He’s unable to be reached for comment at the moment. Try again later.

          • Srsly_Bro
          • 6 months ago

          Not true.

          AMD 2920x and 2970wx came out months after.

          No need to make stuff up, bro.

        • Srsly_Bro
        • 6 months ago

        DOUBLE CHECK MATE!!!

      • Redocbew
      • 6 months ago

      That’s just AMD continuing to be bad at marketing though, isn’t it? New architectures get released then they get tweaked, patched, rebranded, updated, and flogged until they die. Making it sound like every step along the way will be the BEST THING EVAR is pretty standard practice for a chipmaker.

      • charged3800z24
      • 6 months ago

      Actually, the Ryzen Pro Mobile series just add some security features etc, for use in units like HP EliteBook and ProBook laptops. They use the same drivers as the other mobile just like the Ryzen vs Ryzen Pro desktop series. I own a 2700x Pro and it is even unlocked.

      • Concupiscence
      • 6 months ago

      I just want a Zen 2 ultrabook with decent battery life. Is that so wrong?

      • ronch
      • 6 months ago

      Agree with everything youve said. The only real news here is really just 7nm Zen cores and whichever products they may find themselves in. Anything built on 14nm or ’12nm’ is not something to get excited about. And there’s the next gen graphics architecture but it’s still further out.

      Gotta keep in mind though that slight tweaks are pretty common. Ryzen Mobile is pretty good in terms of energy efficiency even compared to Intel (2500U delivers about 80% CPU performance and 2x GPU performance of an i5-8300H at 15w, 8300H is 35w). Yes 7nm is cool but I’m guessing TSMC ramp up will take a little time. Delivering a completely new GPU architecture is pretty expensive as well and GCN is long due for the scrap heap but there’s no doubt AMD knows this and is working hard on their next gen.

    • chuckula
    • 6 months ago

    Gul Dar Shill: [Pulls out Threadripper] How many dies do you see?

    Picard: THERE ARE FOUR DIES!

    Gul Dar Shill: [pulls out Threadripper 3] No, there are NINE Dies! Muhahaha!

    Picard: Screw this, just put me in the reboot series and hope they respect canon [lake]

      • Krogoth
      • 6 months ago

      Ngreedia Dukat: “Tell me Chuck, why there are no good AMD RTG GPUs?”

      • Waco
      • 6 months ago

      Oh come on people, you may not like chuckula, but this post is perfect.

        • chuckula
        • 6 months ago

        The first and second victims of cults are logic and a sense of humor (in no particular order).

        • DancinJack
        • 6 months ago

        maybe some fools here aren’t Trek fans or something, but yeah, this post is really great.

      • derFunkenstein
      • 6 months ago

      I admit to lol-lygagging on this post. +3

    • chuckula
    • 6 months ago

    Don’t worry about our GPU guize!

    Threadripper 3 will put us out of business before Halloween so there won’t be a GPU!!

    #GameOverManGameOver

      • Krogoth
      • 6 months ago

      On a serious note, I suspect Navi at best will only be able to compete against TU116 a.k.a 1660Ti. Nvidia’s 7nm refresh will solidify their post-Maxwell era lead. A 7nm version of TU116 is going probably kill Navi in both performance and production costs.

        • chuckula
        • 6 months ago

        Hey man, I only said they were going to put Intel outof business this year.

        After they get Raja back from his secret mission, Ngreedia will be the victim in 2020. Especially when Ngreedia can’t get those sweet sweet 7nm chips from TSMC like AMD.

          • Krogoth
          • 6 months ago

          Ngreedia hasn’t unleash their secret weapon….. MULTI-GPU GLUING!

            • K-L-Waster
            • 6 months ago

            Glue is soooo Feb 2019. They’re moving on to… STAPLES!!!

            • freebird
            • 6 months ago

            So AMD is going private equity like STAPLES did in 2017? I thought that was what was happening with Global Foundries… now I’m confused… I must be sniffing too much glue.

            • Srsly_Bro
            • 6 months ago

            Wait and see how that goes when Raja K starts at Nvidia after Intel has been leveled.

          • Srsly_Bro
          • 6 months ago

          Raja will level the playing field, and by that I mean Intel will be completely leveled in his path of destruction.

          Raja’s 2019 World Tour Path of Destruction coming soon.

        • Chrispy_
        • 6 months ago

        Navi is the next-gen console architecture, so how it performs relative to Nvidia isn’t a problem, because Nvidia don’t even have a console GPU.

        My only hope is that it’s a gaming architecture, rather than the 8+ year old GPGPU architecture like GCN. Ever since Maxwell, Nvidia has been creating a more game-specific GPU line with rebalanced ratios of ALUs, ROPs, schedulers etc. It’s one of the main reasons why in games most Geforce cards are more efficient in terms of both performance/Watt and performance/transistor.

          • dragontamer5788
          • 6 months ago

          [quote<]Navi is the next-gen console architecture, so how it performs relative to Nvidia isn't a problem, because Nvidia don't even have a console GPU. [/quote<] [s<]Nvidia Shield[/s<] Nintendo Switch would like to have a word with you.

            • Chrispy_
            • 6 months ago

            Haha, I like the little Switch but it’s not really a console.

            • RAGEPRO
            • 6 months ago

            I sort of agree, but for the purposes of your argument it certainly is. Have you seen the Switch sales numbers?

            • Chrispy_
            • 6 months ago

            For discussions on new architectures and new products, I think we can rule out the Switches’ Nvidia Maxwell-generation Tegra IGP.

            It’s about as “new” as the A6 and A8 chromebook APUs that AMD is proudly claiming as an achievement. I certainly wouldn’t be shouting about a 2015 architecture that was the fourth refresh of Bulldozer, a sunk-cost fallacy and arguably AMD’s worst CPU blunder in the last two decades.

        • charged3800z24
        • 6 months ago

        Why would Navi be slower than Radeon VII?

          • RAGEPRO
          • 6 months ago

          The current rumors are that it isn’t a high-end part, much as Polaris (RX 480) was slower than Fiji (R9 Fury).

          • Krogoth
          • 6 months ago

          It is because it is build to be cheap and compact like its Polaris predecessor. People who are expecting it to be significantly faster than Vega 10/20 are completely delusional.

          The days of massive performance boosts between architecture and processing nodes have been gone since Kepler/Tahiti.

          • Chrispy_
          • 6 months ago

          It’s going to be the GPU in the PS5, which is likely to have a launch price of $400-500 at most, so don’t expect a GPU much more expensive than the Polaris GPU in the PS4 Pro or XB1X.

          I doubt that Sony will be willing to make a console cost more than $500 because they want people to buy it for the margins on games. At the same time, Sony have learned from past experience not to sell the console itself at a loss.

          Regardless of the performance of Navi, it will be a medium-sized, cost-effective die for the sake of yields and pricing – and it’ll very likely use affordable GDDR6 rather than HBM2 for the same reasons, meaning that the architectural improvements are going to need to be humongous in order for an affordable mid-size/bandwidth limited piece of silicon to match bleeding edge GCN products like the Radeon VII. GCN may not be a modern architecture but when you throw enough die area, power, and memory bandwidth at it, it gets the job done.

Pin It on Pinterest

Share This