AMD-powered Chromebooks are on the way

Along with the reveal of the second-generation Ryzen Mobile processors, AMD also announced a couple of new chips bound for Chromebooks. The long-time x86 underdog hasn't traditionally had any presence in that market, but now it seems that the company is trying to get its foot in the door with the "AMD A-Series Processors for Google Chromebook PCs."

Stoney Ridge

for Chromebooks

TDP Modules

/ Cores

CPU clock

(base / boost)

GPU core

configuration

GPU peak

clock

Process

Node

VP9 / HEVC

decoding

AMD A6-9220C 6W 1M / 2C 1.8 / 2.7 GHz 3 CUs (192 SP) 720 MHz 28nm Yes
AMD A4-9120C 6W 1M / 2C 1.6 / 2.4 GHz 3 CUs (192 SP) 600 MHz 28nm Yes

If those specifications don't look terribly familiar, they perhaps should. AMD confirmed to us that these processors are new additions to the Stoney Ridge family of low-power APUs that were initially introduced in 2016. That means they're based around a single Excavator module—giving them two "cores"—and three GCN 1.2 (Tonga/Fiji) compute units. The CPU cores have seen some tuning for low-power operation, and the IGP's video block has been updated for modern formats, but otherwise they're unchanged.

Click for a larger version. Source: AMD processor specifications database

In fact, there are already low-power APUs with very similar names out there: the A6-9220 and A4-9120. Those are 15W chips that came out in the middle of 2017. These variants ending in "C" are a little different, even besides the lower TDP. While the A6-9220C's CPU base clock rate drops to 1.8 GHz, it actually gains 120 MHz of GPU clock. Meanwhile, the A4-9120C gains a whole GCN compute unit compared to its higher-powered version, but again sacrifices a good bit of CPU clock rate to hit the lower TDP.

AMD's emphasis for this launch is actually on the performance of its new chips. I expect some gerbils are already cringing at the idea of using a Stoney Ridge APU, but keep in mind that Chromebooks are by no means desktop-class machines. Some Chromebooks are even based around smartphone hardware. Still, AMD keeps its comparisons to other x86-64 chips and in fact directly pits the new APUs against Intel's Apollo Lake-based Pentium N4200 and Celeron N3350.

Source: AMD

Like smartphone operating systems, Chrome OS is designed primarily to handle one task at a time, so the limited core count on these APUs isn't much cause for concern. In benchmarks like PCMark for Android, WebXPRT, and Speedometer, the company's own tests of its new little chips imply decisive leads over the Intel competition. AMD also tested the Unity-engine free-to-play FPS game Bullet Force and claims to have put up a commanding lead there, too.

AMD says it's already in talks with "leading global OEMs" to release "several" AMD-powered Chromebooks throughout 2019. The company specifically named Acer's Chromebook 315 and HP's Chromebook 14 as models that will be getting AMD-powered variants. If you're as curious as we are to see what could be the end result of AMD evolving its "clustered multi-threading" architecture, keep an eye out.

Comments closed
    • MOSFET
    • 9 months ago

    Weren’t there some Jaguar parts that got this low in TDP? And Bobcat before that? (according to CPU-World, yes there were.) Would they not be a better fit? I suppose this has much better GPU.

    ps Quite the reaction to this news post. Great entertainment!

    • ronch
    • 9 months ago

    I honestly don’t know why AMD didn’t just cram a couple of Zen cores in there. Is it because this will be built on 28nm because of lack of 14nm capacity and they think porting Zen to that old node isn’t worth it? And if this is 15w, doesn’t AMD have existing U-series parts like the Ryzen 2500U and 2700U that they can perhaps downclock or something??

      • BurntMyBacon
      • 9 months ago

      To answer one of your (implied) questions: 6W. AMD does not (to my knowledge) have an existing Zen based part at this power level.

    • ronch
    • 9 months ago

    [b<][s<]MEGATRON[/s<] BULLDOZER LIVES!!![/b<]

    • ET3D
    • 9 months ago

    Impressive that AMD has updated Stoney with a newer video block. I find this totally surprising. I didn’t expect construction cores to still be around and kicking by 2019, but hey, I always liked Excavator. If only it had a bit more cache.

    Edit: It’s been pointed elsewhere that AMD advertised these video capabilities when Stoney Ridge was released, so it’s probably the same old Stoney with no change to silicon.

    • Star Brood
    • 9 months ago

    (Barfs in 28nm)

    • Unknown-Error
    • 9 months ago

    Excavator cores? Actually just 1 module with 2 int Cores. Essentially a single Zen core with hyperthreading but without the IPC improvements. Not sure what to say about this. Die Bulldozer, just die!

      • Sahrin
      • 9 months ago

      Bulldozer >>> Goldmont

        • NTMBK
        • 9 months ago

        Yeah, because Bulldozer running at 1.6GHz is going to be SO GREAT

        • chuckula
        • 9 months ago

        That’s the first and only factually accurate post your’ve ever made!

        Just one minor typo: you forogt to put “power consumption” after Bulldozer.

    • Rakhmaninov3
    • 9 months ago

    Why the vitriol before it’s even had the chance to suck majorly for you in some way? Pessimistic predictions are one thing, but damn.

      • JustAnEngineer
      • 9 months ago

      [url<]https://www.youtube.com/watch?v=-y2mqoDjQXI[/url<]

        • albundy
        • 9 months ago

        scroogled. lol

      • Redocbew
      • 9 months ago

      I’m not sure if I should applaud you for your rigorous approach, or offer you the asking price of this chip to lick a frozen signpost just to see if you’d do it.

    • Laykun
    • 9 months ago

    [quote<] Chrome OS is designed primarily to handle one task at a time [/quote<] When was the last time you used Chrome OS? There's all sorts of multi-tasking gestures and tools in Chrome OS. This isn't Android.

    • LocalCitizen
    • 9 months ago

    some smart guy once said: “I love luxury. And luxury lies not in richness and ornateness but in the absence of vulgarity.”

    i guess amd is not luxury

      • ronch
      • 9 months ago

      I need to spend some time to ponder those words one of these days.

        • Pwnstar
        • 9 months ago

        Today is not that day.

      • ronch
      • 9 months ago

      “AMD is not luxury.”

      Jerry disagrees.

    • derFunkenstein
    • 9 months ago

    Just when I thought I was out, they pull me back in!

    • Anonymous Coward
    • 9 months ago

    Did they really make silicon changes to a 28nm design, and then have it freshly fabbed, for the market in 2019?

    Is Stoney Ridge a native 1-module, 3-CU product? Nothing disabled? If so, I guess it must be the cheapest x86 CPU anywhere.

    • chuckula
    • 9 months ago

    You hear that Intel? That’s the sound of AMD beating you across the board in 2019!!

    Our 7nm Epyc Miracle parts will destroy Xeon!

    Our 16 coar Ryzens will nuke your silly i9s.

    And our 28nm “classic” Chromebook parts will show the world that your pathetic Atoms can’t even win at sucking!!

    • DancinJack
    • 9 months ago

    lol no way

      • willmore
      • 9 months ago

      Yeah, I’ve got a Celeron 1007U based chrome book. 2 real cores up to 1.7GHz. And they’re IVB generation cores at that. I bought them because they’re refurbs–which means they’re already at EOL. And they still would out perform these new chips. So, why?

        • rnalsation
        • 9 months ago

        [url<]https://www.cpubenchmark.net/cpu.php?cpu=Intel+Celeron+1007U+%40+1.50GHz&id=1847[/url<] [url<]https://www.cpubenchmark.net/cpu.php?cpu=AMD+A6-9220&id=3072[/url<] [url<]https://www.cpubenchmark.net/cpu.php?cpu=AMD+A4-9120&id=3093[/url<]

          • auxy
          • 9 months ago

          That site (cpubenchmark.net) isn’t a reliable source because its results are based on PassMark which is absolutely not representative of any desktop workload. (;’ω’)

          Compare an [url=https://www.cpubenchmark.net/cpu.php?cpu=AMD+FX-8370+Eight-Core&id=2347<]FX-8370[/url<] to a [url=https://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i7-3770K+%40+3.50GHz&id=2<]Core i7-3770K[/url<] on that site. We all know how that matchup comes out, and yet if you look at Passmark you'd think they're closely matched. In particular for this comparison I wouldn't use PassMark because it specifically inflates AMD FX value, and these chips are based on that lineage. I banned mention of the site entirely in my hardware IRC channel (##hardware on Freenode!) because too many people were being misled by it. So saying, I think Zak was probably too nice to these chips, but it's not like he can just take a dump on new products I guess. (´・ω・) And I guess they're probably fine for browsing and whatever. I sure wouldn't want to use one though...

            • rnalsation
            • 9 months ago

            Look at the single thread score, that is the metric you want to take from PassMark. I just linked those to dispute the claim a Celeron 1007U will outperform these “new” parts for web browsing on Chrome OS.

            To be clear though I think all these parts are bad even the 1007U.

            • Beahmont
            • 9 months ago

            If I remember their testing methods correctly, even their single threaded testing doesn’t mean anything for desktop app performance because it doesn’t use CPU resources like an actual application it just hammers the CPU with nonsense calculations that put disproportionate pressure on certain parts of the CPU and ignore other parts that actual desktop applications don’t ignore.

            What a test does to max out a CPU matters, especially in artificial benchmarks. We know the real world performance of even these Celeron 1007U’s is significantly better than anything in the construction core line from AMD.

            • rnalsation
            • 9 months ago

            I would legitimately like to see some real world comparisons of the 1007U vs these 1 module Excavator parts. Especially after the point thrashing I have taken.

            For the record when using PassMark scores compared to the experience I have had when using the actual hardware, (I worked in a repair shop for 4 years and got to use many low and high power parts) the score was fairly consistent to my subjective experience.

            Edit: Which on of you gold subscribers is hammering me for genuine curiosity? Shame on you

            • Redocbew
            • 9 months ago

            This is not the chip you’re looking for. You can go about your business. Move along.

          • chuckula
          • 9 months ago

          AMD Fanboy logic backfire:

          1. Intel hasn’t improved its per-core performance at all since Sandy Bridge so Ice Lake parts aren’t any faster than that crappy Celeron.

          2. RyZen 2 7nm miracle parts are meant to compete with Ice Lake! Lisa Su said so!

          3. These 28nm parts are clearly superior to the crappy Celeron that’s clearly equivalent to Ice Lake.

          4. Therefore, since RyZen 2 7nm miracle parts are supposed to be competitive with Ice Lake, these 28nm construction core parts are also competitive with 7nm Ryzen 2 cores by the transitivity property.

          THANKS AMD!

            • BurntMyBacon
            • 9 months ago

            I think there may have been malware in that post.
            My L0g1c un1t$ are @ct1ng funny.
            D0es n0+ c34t9ute!*
            $t@faj4 0v3rf9040w
            M3m02r493 $p@c234 C0ronfp3td

    • sleeprae
    • 9 months ago

    Cringing? That and more. I’m not sure there’s any excuse good enough for AMD to justify launching a processor in 2019 that’s not based on Zen.

      • NTMBK
      • 9 months ago

      How about the fact that a new Zen dual-core die would cost millions to develop, whereas this probably cost about $3.50?

        • Wirko
        • 9 months ago

        Raven Ridge already exists, so it would also cost $3.50 to develop again.

          • NTMBK
          • 9 months ago

          It’s a big chip for these ultra low margin devices. Ideally you’d want a dedicated 2C 3CU Zen die, but it looks like AMD cancelled that.

        • willmore
        • 9 months ago

        Tree-fiddy

Pin It on Pinterest

Share This