Qualcomm 8cx SoC steps up for the next Always Connected PCs

You may remember the Always Connected PCs we talked up a while ago. Those lean notebooks promised competent computing performance paired with constant cellular connectivity and particularly long battery life. Those machines were based on Qualcomm's Snapdragon 835—in other words, a smartphone SoC for laptop work. Qualcomm has now decided to up the ante with its next designed-for-laptops chip, the Snapdragon 8cx.

The company says this is the biggest SoC it's ever produced, and it aims for its partners to use it in always-online machines with multiple-day battery life—quite the bold claim. The 8cx is built on a 7-nm process, and its centerpiece is the Kryo 495 CPU. According to Android Authority, the processor uses four customized Cortex-A76 performance cores along with four tweaked Cortex-A55 power-sipping tickers. The Kryo 495 reportedly has 10 MB of cache in total, and it should hook up to as much as 16 GB of LPDDR4X RAM. There's also support for UFS 3.0 and NVMe SSDs on tap.

Going by Android Authority's reporting, Qualcomm claims the Kryo 495 rivals an unspecified Intel U-series chip for performance in an unspecified workload at a 7-W TDP, but does so at much lower power consumption and without a fan. This comparison is a little difficult to square, as Intel's lowest configurable-TDP-down on a Core i5 U-series part of late is 10 W, and the only parts the blue team rates for 7-W TDPs at all are its Y-series CPUs in their configurable-TDP-up trim. Instead of trying to figure out Qualcomm's vague and buzzy performance claims, we'd just wait until reviewers get devices in hand to see how this SoC actually performs.

Qualcomm claims the Adreno 680 GPU inside the 8cx is 3.5 times as fast as that of the Snapdragon 835 mobile SoC of yore, and twice as speedy as the Adreno 630 in the last-gen, high-power Snapdragon 850. Up to two 4K HDR displays at 60 Hz can run off this new pixel-pusher, too. Otherwise, the 8cx packs what's presumably the same Hexagon 690 neural network processor and X24 modem present in the recently-announced Snapdragon 855. The X24 modem offers download speeds of up to 2 Gbps, again a welcome figure to ensure that "always connected" doesn't mean "always dragging." Additionally, the 8cx also offers support for USB 3.1 Gen2 connectivity. 802.11ad-capable WiFi, aptX HD audio, and Amazon Alexa integration round out the main list of specs.

We figure that these machines will live and die by their software support, and Qualcomm's presentation displayed a fair amount of well-known logos, though the company didn't go into too much detail about its software partnerships other than pointing out that it's worked with Mozilla to make an ARM64 multi-threaded version of Firefox, and, naturally, with Microsoft, who'll bring Windows 10 Enterprise to 8cx-based machines.

Although we'd wager that batteries inside these newfangled notebooks are bound to be relatively small compared to those in conventional laptops, they nevertheless should top up pretty fast thanks to the 8cx's support of Qualcomm's Quick Charge 4+ protocol. Qualcomm didn't spell out loud when we'll see the new generation of Always Connected laptops, but CES 2019 is just around the corner, so you take a guess.

Comments closed
    • JosephParker
    • 10 months ago

    Thanks for sharing this post. I am interested in this topic. They’re inserted right into a smoke chamber which funnels the smoke to the meals compartment and they’re allotted at an interval of 1 each 20 minutes. [url<]https://startbbq.com/best-charcoal-grills/[/url<]

    • kmieciu
    • 11 months ago

    Cool, but i will wait for upstream linux support.

    • tipoo
    • 11 months ago

    No Prime core here then?

      • Lianna
      • 11 months ago

      Maybe they are all Prime cores. Prime cores all the way down.

        • tipoo
        • 11 months ago

        That seems right actually, these Kryo 495’s appear to have more cache than even the Prime core.

        Still, what about a Prime-er core.

      • Anonymous Coward
      • 11 months ago

      I guess its just a yields or marketing trick rather than any difference in silicon.

      • Usacomp2k3
      • 11 months ago

      My first thought as well. I really like the idea of a prime core. I think that could work on the PC too.

        • Anonymous Coward
        • 11 months ago

        Didn’t some of Intel’s chips already know which cores could be turbo-ed the highest?

          • DancinJack
          • 11 months ago

          [url<]https://www.intel.com/content/www/us/en/architecture-and-technology/turbo-boost/turbo-boost-max-technology.html[/url<]

          • tipoo
          • 11 months ago

          Yes, but here the Prime cores are physically different, with twice the L2 cache for one, possibly more optimized transistors along critical paths for higher clocks.

            • Anonymous Coward
            • 11 months ago

            I saw that about twice the L2 and assumed they just disabled parts of the L2 for the other cores to make a point. Sounds strange that they would tweak transistors on just one of 4 very nearly identical cores. I’m going with 100% of the R&D having been carried out in the marketing department (but I’ll graciously accept a correction if there is solid evidence to the contrary).

            I’ll grant that its not impossible to actually put more L2 on one of the cores than the others, but it seems like a dumb idea from the perspective of yields. Assigning the best core as the [b<]MEGA SUPER CORE[/b<] sounds like the smart move.

            • tipoo
            • 11 months ago

            I’ll be curious to see a die scan. Disabling other cores L2 cache to make a point sounds out there, unless they were having yield problems, and the 855 die is pretty small.

            • Anonymous Coward
            • 11 months ago

            Mostly its the question of yield that bothers me with the claim of a physically distinct core. Perhaps they have an alternative product to absorb those devices that fail to have the needed top clocks (and efficiency) from that particular core.

    • End User
    • 11 months ago

    “The X24 modem offers download speeds of up to 2 Gbps”

    Hang on. I have a 6GB mobile data plan. Is 6GB equal to 48Gb? Could I burn through my entire data plan in 24 seconds with that modem?

      • Beahmont
      • 11 months ago

      Short answer: Yes.

      Long Answer: Not unless your wireless provider actually has a spot where you can actually achieve 2 Gbps download speed, but if they do yes you would be able to.

    • End User
    • 11 months ago

    [url=https://www.theverge.com/2018/12/6/18129456/qualcomm-snapdragon-pc-firefox-web-browser-64-bit-native-mozilla<]Firefox is making each browser tab run on individual cores in the CPU[/url<] That does not sound great.

      • sweatshopking
      • 11 months ago

      The verge thinks it is

      [url<]https://www.theverge.com/circuitbreaker/2018/12/6/18129678/qualcomm-8cx-mozilla-firefox-windows-10-64-bit-arm[/url<]

        • End User
        • 11 months ago

        You got played. Watch their video. Their own website exhibits poor performance.

        You really don’t pay attention to anything. At all. Ever.

          • sweatshopking
          • 11 months ago

          You should read more carefully. I didn’t say I thought it was fine. I made no claims at all. I said the verge thinks it’s not bad, AND THAT’S WHAT THEIR ARTICLE SAYS.

          • NTMBK
          • 11 months ago

          That’s just the Verge, it runs like ass on an i7.

      • tipoo
      • 11 months ago

      One tab per core is strange, especially in a big.LITTLE environment? If you had 8 tabs 4 would be running on the slow cores, rather than heterogeneously matching the load?

      • synthtel2
      • 11 months ago

      That reads like someone was trying to say “it multithreads well” and the message got mangled somehow. FF already does far better at multithreading than that, and it’d be very weird if the ARM/Windows version didn’t.

    • ronch
    • 11 months ago

    I used to want ARM to make it into PCs and servers but I think there’s really no advantage right now especially given how aggressive AMD is these days. Just Google what the Ryzen 3700X plans to be for $330. Good grief I’m glad I didn’t bite into Ryzen just yet.

      • Anonymous Coward
      • 11 months ago

      Amazon is pushing ARM, anyone can rent a “a1” instance right now and run their standard Linux stuff on them. At this moment I have 64 of their cores chewing up log files for me, the price/performance ratio is fine, but it takes 4 ARM cores to match one Xeon in this task. They claim to have lots of network IO on these, useful for some people.

      So its not particularly compelling yet, but also they say this AWS machine is based on clusters of A72 cores, lots of room for improvement with little effort on their part.

        • freebird
        • 11 months ago

        Not sure I agree with the “lots of room for improvement with little effort on their part.” statement…

        The question remains what foundry process were the chips built upon? If 7nm, then a redesigned chip and associated masks costs quite a bit, regardless of the “effort” needed. That being said, Amazon has the revenue to do it.

        Good article on the Graviton processor here:
        [url<]https://www.theregister.co.uk/2018/11/27/amazon_aws_graviton_specs/[/url<]

          • Anonymous Coward
          • 11 months ago

          Well, I mean of course they would need to make a whole new chip, but the cores they used on a1 are pretty easy to beat with improved off-the-shelf cores. Also the caches used are apparently pretty modest, a shared 2MB L2 per 4 cores, hardly server-grade these days. I imagine they just used off-the-shelf memory controllers, ethernet controllers, encryption engines, etc etc as much as possible. So when I say “little effort” I mean the R&D needed to gain a lot of IPC is pretty modest here, they just need to have their engineers put together pretty standard things and they should see a big payoff.

          Also, I really doubt they went for 7nm. I expect these chips are designed to be cheap, and they use the same ones everywhere behind the scenes, and probably have been doing so for some time. They have so much “serverless” infrastructure, bits and pieces moving data around, huge amount of things to implement in the most cost-efficient way they can.

          AWS doesn’t waste money on anything, so far as I can tell, so no doubt they have a totally rational business case for this investment. I expect “a1” is just them fishing to see what the market can do with the tools they already needed.

      • tipoo
      • 11 months ago

      To me the biggest benefit would be breaking the duopoly. AMD is on a roll, but what could the last 10 years have been like with 3-4 competitors duking it out instead of a relaxing Intel and lagging AMD, until only recently.

        • Anonymous Coward
        • 11 months ago

        Its a bit of a stretch to describe recent history as much other than [b<]mono[/b<]poly. But if anyone cared about an alternative, IBM has been there all along with something which is fairly viable, although apparently not so viable that was chosen over Intel. I speculate that IBM's problem is they are in certain highest performance niches, less attractive than the super efficient volume market.

    • ronch
    • 11 months ago

    Executive used to wear a coat and tie when doing presentations like this. Now everyone just wants to be hip and cool and wear a t-shirt and jeans.

    Someone should start the trend of just wearing a barrel with straps on their shoulders. You know, “We’re a lean company bent on putting out great products and minimizing costs.”

      • psuedonymous
      • 11 months ago

      Next year’s GDC, Jensen Huang takes to the stage wearing his trademark leather jacket. Only.

        • ronch
        • 11 months ago

        Don’t worry Nvidia graphics cards will automatically blur some parts of the video on-the-fly.

          • tipoo
          • 11 months ago

          THE MORE YOU BUY
          THE MORE IT BLURS

    • End User
    • 11 months ago

    No benchmarks released. Hmmmmmmm.

      • End User
      • 11 months ago

      A dude from The Verge got his hands on a reference design laptop at the event. He had previously launched a bunch of websites in tabs in Firefox (for ARM), one of which was theverge.com. When he switched back to that tab and then scrolled down, the page load was atrocious:

      [url<]https://twitter.com/StarFire2258/status/1070828688000733184/video/1[/url<]

        • sweatshopking
        • 11 months ago

        Firefox for arm isn’t even remotely close at all to being finished. This proves absolutely nothing.

        But I’m always happy to read one of your posts.

          • End User
          • 11 months ago

          The evidence suggests that the current build is a sluggish turd.

    • End User
    • 11 months ago

    Windows 10 on ARM has no app or CPU advantage over any other platform.

    • homerdog
    • 11 months ago

    Will this Windows on ARM thing run any windows program? The stuff I use for work will remain forever x86 and most of it barely works there.

      • derFunkenstein
      • 11 months ago

      Back when it was first announced, it would only run 32-bit x86 apps. I haven’t seen anything since then that indicates otherwise.

        • sweatshopking
        • 11 months ago

        Yeah, at this stage there are no Official plans to emulate x64.

    • sweatshopking
    • 11 months ago

    Does it make my iTunes quicker

    Qualcomm did specifically state that they expect it to compete in single threaded performance with the 15w i5s.

      • chuckula
      • 11 months ago

      And thus was a meme born.

      #ThanksAMDIMeanQualcomm

Pin It on Pinterest

Share This