Updated: Apple touts upcoming Radeon Pro Vega option for MacBook Pros

Apple's event in New York City is ongoing as I write this, and a new MacBook Air and a new Mac Mini have both debuted. We'll be covering those as soon as we can, but I wanted to specifically point out an interesting tidbit of news that might have passed by in an instant on stage. In its press release for the new MacBook Air, the company also noted that some form of Radeon Pro Vega graphics processor is coming to its MacBook Pros as an option "next month."

Source: AMD

Thing is, the only notebook-friendly discrete Vega part that AMD has announced so far—or teased, really—is the Vega Mobile GPU it showed at CES earlier this year. That package integrates a single stack of HBM2 RAM alongside its pixel-pusher. We still don't know much, if anything, concrete about Vega Mobile many months after AMD introduced it, but the company promised "high frame rate AAA gaming in compact form factors," as well as low power usage, low Z-height, a small footprint, and VR-readiness.

UPDATE: On its YouTube channel, AMD confirmed that two Vega Mobile graphics processors are coming to MacBook Pros: the Radeon Pro Vega 16 and the Radeon Pro Vega 20. We're digging for more details and will update this post as we get them.

Source: AMD

Z-height and footprint are two areas that tend to mark Apple-first products from AMD, in our experience, so perhaps it's not so surprising that Vega Mobile is making its debut in MacBook Pros. We'll be keeping an eye on those machines to see whether they do, in fact, get Vega Mobile parts and how much they add to the cost of a given machine as soon as we can find out.

Comments closed
    • tipoo
    • 1 year ago

    I did think the “X” GPU updates of the 2018 MBP revision were a concession for Vega not being ready in the right perf/watt yet.

    I wonder if it’s more efficient than Polaris mobile by now with node refinements. Because if it isn’t, I also wonder if the cooling is any different, because an i9 with the same cooling as every model since 2016 didn’t go so well…

    • Kretschmer
    • 1 year ago

    Nice GPU; it’s a pity that it’s ending up in Macs that are on what…OpenGL 1.0?

      • ermo
      • 1 year ago

      OpenGL 3.3 if my memory serves me correctly.

      That said, Metal will likely be plenty capable of extracting the full performance potential of these chips (not to mention the Molten vulkan-over-metal middeware).

      • tipoo
      • 1 year ago

      Deprecated though. I wish they added Vulkan, but at least we’re moving off their ancient slow OpenGL build (3.3).

    • ronch
    • 1 year ago

    I’m just curious. Given the general perception that Nvidia graphics are superior to AMD’s, why does Apple go with AMD on a lot of their computers?

      • NoOne ButMe
      • 1 year ago

      NVIDIA tried something stupid. Apple said no thanks.

      AMD being willing to bend over backwards to work with Apple helps also.

      • Voldenuit
      • 1 year ago

      Probably nvidia burned their bridges with Apple (or vice versa) after Bumpgate.

      • tipoo
      • 1 year ago

      Big egos on both sides…Nvidias 8000 series had solder crack flaws in the MBPs, Nvidia wouldn’t take responsibility for a long while blaming Apples thermals, there were legal fireworks, yada yada.

      And now we’re deprived of the most impressive GPU architectures available under macOS. But I feel like Apple is just going to sidestep it all with their own GPU silicon alongside their custom CPUs in the coming years.

        • NoOne ButMe
        • 1 year ago

        Nvidia still has their drivers for OSX, yeah?

          • tipoo
          • 1 year ago

          Yeah, but they’re kind of flaky and sometimes shaken apart by OS updates, unlike macOS core drivers.

            • NoOne ButMe
            • 1 year ago

            Good to know, thank you.

    • ptsant
    • 1 year ago

    Can anyone guestimate how the 20CU version compares with the previous RX560?

    I never bothered to look up the exact specs (shaders, frequencies) but I’m wondering how much of an improvement it is.

      • NoOne ButMe
      • 1 year ago

      Probably 60-70% faster.

      About a 40% peak clock and 25% more CUs.

    • LocalCitizen
    • 1 year ago

    confusing naming convention

    vega 10 chip
    – rx vega 56, vega 64 cards
    vega 12 chip
    – pro vega 16, pro vega 20 mobile cards
    vega 20 chip
    – 7nm
    – instinct card

    then there is the “vega m” which isn’t a vega

    edit: also vega 11, vega 10, vega 8, vega 6, vega 3 gpu modules in ryzen / athlon apus

      • freebird
      • 1 year ago

      This is the world we live in
      And these are the code names we’re given
      Stand up and let’s start showing
      Just where our GPUs are going to.

      [url<]https://www.google.com/search?q=Land+of+Confusion+song&ie=utf-8&oe=utf-8&client=firefox-b-1[/url<]

        • Scrotos
        • 1 year ago

        This post needs more love.

          • ET3D
          • 1 year ago

          You’re right. Ah, the good old days of Spitting Image and Genesis.

        • derFunkenstein
        • 1 year ago

        I heard it in Phil Collins voice even before I clicked the comment and saw the link.

      • ET3D
      • 1 year ago

      It’s only confusing to people who care about chip code names. In terms of consumer message, this is hands down the most consistent naming convention ever seen for GPUs. People who care about chip code names are (or at least should be) intelligent enough to understand the meaning from context.

        • psuedonymous
        • 1 year ago

        Are you having a laugh? You’ve got two things both called “Vega 20” that are completely different, two things called “Vega 10” that are completely different, and “Vega M” which has nothing to do with Vega. You’ve got GPU card names that increment with performance, and GPU chip names that decrement with performance, and both lines cross over each other with intersections that do not correspond to the same product.

          • NoOne ButMe
          • 1 year ago

          As a consumer, wth would you care what **internal** codenames are?

          The naming of the GPUs is messy, and should be changed. But the only names consumer should care about (you know, the ones that consumer products are branded with) are very nice.

      • NoOne ButMe
      • 1 year ago

      It’s a mess… If you follow the development of the GPUs.

      Personally I REALLY, REALLY like the consumer facing names. They’re the best (a very low bar) in a long time. Way better than anything either company has had for a long time, if not ever.

      As someone who follows codenames, I cannot wait for them to stop mixing up inward and outward facing codenames.

      note. Vega 20 only has one chip announced/uncovered afaik:
      Radeon Instinct Vega 7nm

    • Krogoth
    • 1 year ago

    Apple should have gone with Turing! Do they not know that Ray-tracing is the future?

    #PoorNavi
    #PoorVega

      • BurntMyBacon
      • 1 year ago

      You are looking at this all wrong. Everyone copies Apple, like with the notch on their phones and removing useful ports just because they are old enough to have proved their worth. If Apple dismisses Ray-tracing then it is DOA. Sorry nVidia.[/jk]

        • K-L-Waster
        • 1 year ago

        In that case, they should have gone Voodoo [s<]3[/s<] Infinity!! #Make3DfxGreatAgain

    • Jeff Kampman
    • 1 year ago

    AMD’s YouTube channel confirms: we’re getting Radeon Pro Vega 16 and Radeon Pro Vega 20 graphics processors for MacBook Pros. [url<]https://www.youtube.com/watch?v=oDGFyc6fZls[/url<]

      • Mr Bill
      • 1 year ago

      Hmmm, were those simply distortions of wire mesh or were they intended to resemble distortions of star constellations?

    • Chrispy_
    • 1 year ago

    [s<]We bought a couple of Hades Canyon NUCs for mobile large-format display trolleys and they do the job but I'm underwhelmed at best, maybe even a little sour and disappointed. They are slow and there's no way to candy coat it. Zotac Magnus ER51070 were vastly cheaper and approaching 3x the performance. Yes, they're fractionally bigger but the footprint is roughly the same and the NUC still requires a frankly-ridiculous power brick, detracting from its compact dimensions. Raja's baby may be the best APU ever, but all APUs are low-end experiences, period. The NUC was slower than the 960M laptop it replaced :([/s<] Oh, wait - this may not be the RX Vega M GH with 24 compute units after all. I'm going to assume that it's a 7nm product though, since Apple care about performance/Watt and are willing to pay top dollar for the best process tech.

      • chuckula
      • 1 year ago

      It’s not 7nm. It’s the same mobile Vega that was shown at CES in January.

      • Jeff Kampman
      • 1 year ago

      If Vega Mobile is in fact the chip Apple is using, it’s definitely not the Kaby Lake-G package. It’s a discrete graphics chip with HBM, no CPU.

        • chuckula
        • 1 year ago

        I never said it was Kaby-G. AMD never showed off Kaby G at CES or any other show, only Intel directly demoed those syatems.

          • Jeff Kampman
          • 1 year ago

          I wasn’t replying to you 😛

            • chuckula
            • 1 year ago

            Lol! Ok

        • Chrispy_
        • 1 year ago

        Any idea what the TDP is?

          • Jeff Kampman
          • 1 year ago

          No public specs yet beyond compute-unit counts and 4 GB of HBM2.

            • Chrispy_
            • 1 year ago

            AMD are more than happy to make semi-custom solutions for large enough buyers and Apple definitely qualify.

            I’m really curious how a 20CU Vega with only one HBM2 stack performs, because I’m assuming Apple will clock it for power-efficiency rather than overvolting it to within an inch of its limit like AMD did with Polaris and Vega desktop designs.

            • leor
            • 1 year ago

            This is what I’m curious about as well. Is it linear to the Vega 56? That would make it about 40% of the performance which isn’t fast at all, so I don’t imagine that would be the case.

            • freebird
            • 1 year ago

            The Vega Desktop GPUs actually run better when under-volted and they weren’t very aggressive on the HBM clocks as well. The HBM on Vegas can usually be undervolted and overclocked at the same time. I undervolted mine to 1000-1050mv and overclock it to 1100Mhz. When I used them for mining which doesn’t stress the card as much as gaming the HBM could be undervolted down to 925mv and still maintain 1100Mhz clockspeed with proper cooling. Vega can get pretty frugal (compared to the standard desktop part) when underclocked/undervolted.

          • NoOne ButMe
          • 1 year ago

          Probably 35-40W. That is what the Polaris parts today in the MBP approximately draw.

      • NoOne ButMe
      • 1 year ago

      Nope, 14nm.
      But Vega can be very power efficient.

      This is clocked much more akin to the Vega11 iGPU. You know, the one which with less memory bandwidth (same, but IGPU cannot access all) was about on par with the GT1030.

    • Voldenuit
    • 1 year ago

    At least the marketing slide is honest enough to call their *cough* Polaris chip “Vega” in quotation marks, Dr. Evil style.

    EDIT: Wait, this is not the intel CPU + Vegaris EMIB package, so maybe it /is/ Vega?

      • Jeff Kampman
      • 1 year ago

      If this is indeed the same Vega Mobile AMD showed off at CES, it’s a discrete GPU with HBM2 on package.

    • DPete27
    • 1 year ago

    You don’t suppose that “Vega Mobile” would be in reference to the Vega M IGP in the [url=https://ark.intel.com/products/130409/Intel-Core-i7-8809G-Processor-with-Radeon-RX-Vega-M-GH-graphics-8M-Cache-up-to-4-20-GHz-<]Intel i7-8809G[/url<]?

      • Jeff Kampman
      • 1 year ago

      I highly doubt it. Nobody has put a 100-W package power Kaby-G part in a notebook, for one, and the quad-core Kaby Lake part would be a regression in CPU performance for the MacBook Pro.

      Given the system architecture of MacBook Pros today (Intel H-series Coffee Lake CPUs plus discrete Radeon Pro graphics processors), a discrete Vega Mobile chip would seem to make the most sense.

      • Magic Hate Ball
      • 1 year ago

      The package looks different.

      That one has the Intel CPU, AMD GPU, and HBM all on one elongated package. This appears to be a standalone package.

      • freebird
      • 1 year ago

      I object your Honor, This is sheer speculation. I move that it be stricken from the comments.

      Objection sustained.

Pin It on Pinterest

Share This