The first two Windows 10 computers with ARM-based chips are already available, and the third should go on sale any day now. But early reviews suggest that the Asus NovaGo, HP Envy x2, and Lenovo Miix 630 are hampered by software limitations and relatively slow performance.

At least part of the issue is that they all ship with Qualcomm Snapdragon 835 processors, which may have been a state-of-the-art smartphone chip when it launched in 2017, but it’s relatively pokey by PC standards… especially since it has to emulate x86 architecture to run some apps.

This year PC makers are expected to launch the first Windows 10 devices with Snapdragon 845 chips, and it looks like the next-gen processor could bring a big performance boost.

Recent entries in the Geekbench database suggest computers with Snapdragon 845 chips could score around 25 higher in multi-core CPU tests, and about 40 percent higher in single-core CPU tests.

Update: This might actually be the rumored Snapdragon 850 chip, which is expected to be a Snapdragon 845 with a higher clock speed that’s designed specifically for Windows on ARM devices.

Geekbench (click for more details)

It’s always a good idea to take synthetic benchmarks with a grain of salt, since they’re not always representative of real-world performance. That’s especially true when we’re talking about unreleased products: it’s possible that software and hardware changes that roll out before these devices hit the streets (if they ever do) could have a big impact on performance.

That said, as WinFuture noted when it first spotted these benchmark listings, it looks like not only is the Snapdragon 845 chip more powerful than last year’s Snapdragon 835, but it looks like PC makers are running it at higher clock speeds than smartphone makers.

Most phones that use the chip have top speeds of 2.8 GHz. An unannounced Lenovo computer has the octa-core chip running at speeds up to 2.96 GHz.

While a 40 percent performance boost sounds exciting, it’s worth keeping in mind that Qualcomm’s chips have a long way to go before they catch up to Intel and AMD’s latest processors, at least in terms of raw horsepower.

I went looking for another processor that offered similar Geekbench scores, and the closest I could find was the Intel Pentium N4200, which is a low-cost, low-power quad-core processor based on Intel’s Apollo Lake architecture. It was released a few years ago and is regularly used for low-end laptops like the Lenovo IdeaPad 120S, which has a list price of $250, but which often sells for even less.

Geekbench (click for more details)

Windows 10 on ARM computers, by comparison, currently have prices that start at $599 and go up from there.

There are some advantages to Windows on ARM. These computers tend to be very energy efficient, offering long battery life and thin-and-light designs. They’re also considered “always-connected” PCs, since they have integrated support for 4G LTE data networks, allowing you to stay online when you leave the comfort of your home or work WiFi network. And, like smartphones, they can enter a low-power state when the screen is off, while continuing to receive notifications and other updates.

They can’t run 64-bit apps designed for x86 architecture though. Microsoft recently released an SDK that makes it easy for developers to port those apps to ARM64 architecture so they can run natively on Windows 10 on ARM. But it’ll be up to developers to decide whether to do that… which creates a bit of a chicken-and-egg problem: there’s not a lot of incentive for Windows developers to ensure their apps are compatible with ARM unless there are a lot of people using these new computers. But there’s not much reason for people to use these computers unless they support all the apps users want to run.

That’s a problem that’s less serious today than it was a number of years ago when Microsoft trotted out Windows RT, an operating system that could only run apps that were specifically compiled for ARM architecture (and which could only be downloaded from Microsoft’s app store). Supporting Win32 x86 apps through emulation helps reduce the size of the app gap. But emulation uses more resources than running native apps, 64-bit x86 apps aren’t natively supported, and in the end I still have to wonder if you might not be better off just buying a computer with an Intel Celeron or Pentium processor for less than half the price of a Windows 10 on ARM PC?

Or maybe Microsoft and Qualcomm have a few more tricks up their sleeves that we can’t see from benchmark tests alone. I’d like to be pleasantly surprised.

Perhaps we’ll hear more from the companies soon… the annual Computex computer show is just over a week away.

Support Liliputing

Liliputing's primary sources of revenue are advertising and affiliate links (if you click the "Shop" button at the top of the page and buy something on Amazon, for example, we'll get a small commission).

But there are several ways you can support the site directly even if you're using an ad blocker* and hate online shopping.

Contribute to our Patreon campaign


Contribute via PayPal

* If you are using an ad blocker like uBlock Origin and seeing a pop-up message at the bottom of the screen, we have a guide that may help you disable it.

Subscribe to Liliputing via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 9,543 other subscribers

37 replies on “Windows 10 on ARM PCs could get up to 40 percent faster with Snapdragon 845 (or Snapdragon 850)”

  1. What really would get the platform going would be to have a virtualbox-like interface and support for ARM based OS’s running various OS, most of which are already ARM ported. Then the not-free x86 architecture could be ditched entirely, which only gives 3 licenses – Intel, AMD, and Mediatek, i.e. a virtual CISC oligopoly. We all know RISC has more promise. Of course Oracle isnt going to do it, but it’d be awesome to see a group of dedicated people get behind something like this.

  2. Pretty good news, I really want an 8″ ARM tablet with pen and LTE support and long battery life, there’s a review of the HP one on a Windows blog where the user is very happy with it. They just need to price it right.

    1. Personally, I’d like an 8″ WoA (Windows on ARM) + LTE device with a physical keyboard. At least to me, a cramped physical keyboard is still better than an onscreen one. I’d settle for an attachable keyboard like the ones on the MS Surface and its clones.

      It’d be nice to if WoA gets OEMs to build these types of machines. The initial 3 WoA devices seem to have squandered ARM’s advantages creating devices that are about just as heavy and large as typical notebooks in the, also, high price range.

      Hopefully, there’s a big enough market for smaller form factors for these larger OEMs to consider entering/re-entering it.

      1. For a physical keyboard, a 10″ is the minimum for a full keyboard, like the Surface 3 and the replacement MS is working on. I am hoping for 10″ chassis horizontally with 11″ screen size diagonally, and very little bezels.

        That is the minimum form factor required for productivity with a keyboard/mouse input. After that I am hoping the OEMs create Nintendo switch style 8″ gaming tablets, with detachable controllers, they can play steam and Xbox games, and be quite powerful with Nvidia Tegra X3.

        MS will create the Surface Andromeda with 6″ screen that unfolds to become a 9″ tablet. And the OEMs like HP can create windows 10 coreOS 6″ smartphone form factor device.

        So that covers 6″, 8″, 10″ sizes for windows devices. And of course the surface Andromeda design can be applied to larger device by an OEM too, like the Lenovo yoga halo.

        1. Well “minimum”, “required” and other similar words/phrases are subjective or, at least, not universal. In that regard, I certainly hope MS and OEMs aren’t listening to you because my “minimum”/”required” sizes and form factors for productive physical keyboard use is completely different from you.

          1. Fair enough, I think we can both agree that we need various devices of different form factors for windows, ranging from 6-11″ categories

  3. > maybe Microsoft and Qualcomm have a few more tricks up their sleeves that we can’t see from benchmark tests alone.

    This is the key. Right now, there is absolutely no reason to invest in these machines while running Windows 10.

    It’s bad enough that Windows remains a hybridOS running store apps, legacy apps – which leads to separate control panels, installation methods, etc… With ARM, they’re limiting x86 64-bit, tossing in emulation (which has always been bad on any system), doubling down on the hybrid framework. Messy… a nightmare.

    I’m thinking that the only way this works is that they slowly get consumers/devs interested. Store apps (given their better performance) eventually gain a foothold. A few years later, Windows on ARM becomes a store-only computer (RT revisited). So these devices are long-term strategy to make MS store viable. This may eventually lead to a stronger mobile ecosystem.

    1. It’s not ARM specific to get it run only in the MS Store. Windows 10 S Mode and Windows Polaris is that strategy, we already are going to see a lot more S Mode PCs coming to market. ARM PCs already run S Mode, same with a number of Intel based PCs.

      We will see the PC market agressively shift towards Windows 10 S Mode. More S Mode PCs available and more people use it, more applications and this therefore will benefit Windows Polaris which is thier true locked down Chrome OS competitor.

      Windows 10 S Mode = Runs MS Store only – UWP, Centennial Win32, PWA are native and has legacy functions. you can also opt out of S Mode

      Windows Polaris = Runs MS Store only – UWP, PWA are native whereas Centennial Win32 is not (Virtualised), has no legacy functions and can’t opt out.

      So Microsoft only needs to shift the market to S Mode and they’ll be good.

    2. > Microsoft recently released an SDK that makes it easy for developers to port those apps to ARM64 architecture so they can run natively on Windows 10 on ARM. But it’ll be up to developers to decide whether to do that…

      …and at the same time they did not port their own flagship application – MS Office to ARM. And I mean the full Office (the one that runs on Windows i386 or MacOS, not the light version for UWP or Android or iOS)

      1. I read one article and only reason why they don’t distribute arm based version is problem with plugins. As all plugins need to be recompiled for arm.

  4. For the pricing on a lot of these snapdragon windows PC’s you could just buy something with a core i3/i5 and run everything perfectly fine.

    1. It’s the battery life and weight advantages. For my work laptop, I would prefer one of these as it is easier to transport and has significantly longer battery life. The only software requirement my company has is office 360.

      1. Based on reviews, these ARM devices don’t have significantly more battery life and aren’t much lighter than similarly priced or similarly performant devices.

  5. How do Qualcomm’s Snapdragon SoCs compare with chips from other companies? Particularly ones with integrated LTE?

  6. I am looking forward to the MacBook with the A series SOCs. Probably will be the same price as Qualcomm/Win10, and run Microsoft Office. I would prefer the MacBook as my work machine, but this won’t happen for another 2 years.

    1. I hear (haven’t looked at actual numbers) Apple’s A series SoC is pretty powerful. I’m not very particular about the OS I use other than it needs to be a desktop class OS (ie. desktop Linux distro, macOS or Windows 10). Too bad Apple doesn’t make devices in a form factor I prefer nor they integrate LTE into its Macs.

  7. Sooooo… uhhhh…

    What’s the… TDP of the 835 or 845… and would that question make sense… and what would happen if QCOM actually designed chips for laptops with higher TDP?

    1. I believe the TDP is roughly 2-4 watts, the iPad pro chips are 4-6 watts, and Intel core M3 is 4-6 watts also, but they are 3-4 times more expensive than the most expensive ARM chip MS can use, as in the snapdragon 845. Since apple chips are exclusive, and Nvidia is watching from sidelines along with Dell.

      It wouldn’t make sense for Qualcomm to create a separate class of chips, when they can simply use their mass production line.

      1. Actually that’s a bad comparison. A better comparison would be exactly what the articled compared the 845 to which is the older Intel Pentium N4200 not the Core m3. As you saw in the article, the Pentium N4200 performs roughly the same as the 845 in benchmarks. Considering laptops with the N4200 costs around $250, the cost of the N4200 SoC for OEM’s is probably comparable to that of the 835.

        Basically a cheap laptop with a Pentium N4200 will have roughly the same performance as the Snapdragon 845, but it will be capable of running more programs and devices while costing up to 75% less than most WoA laptops so far.

        Not to mention the Pentium N4200 is already made obsolete by the N5000 which costs the same but performs even better than the N4200. The Core m3 is expensive by comparison yes, but its performance is way above both the N4200 and 845. Plus Core m3 devices cost about double N4200 devices. And once you get into the price range of “premium WoA” devices, you can get “premium x86” laptops with full Core i7 CPU’s.

  8. No deal for me until these computers are offered at high value pricing for the buyer, until then the latest 4 cores Atom N5000 which will soon show up for purchase will still perform better and will be found on a variety of laptops for less than $300…

    1. It is worth pointing out that the benchmarks test x86 code but most users will be running native ARM code which is significantly faster. There will be the odd app that requires x86 but for web browsing, email, office etc you will by running native code.

      1. However, Brad’s linked to article show that even native application benchmarks (ie. MS Edge browser) put the ARM powered notebooks to at the level of non-Core Celerons.

        1. That was 835, the ARM processors are becoming 100% more faster roughly every 2.5 to 3 years.

          The current core M3 processors have Passmark benchmark score of 3500, the Pentium silver j5005 is at 3000 roughly, and Celeron dualcore is around 2000.

          Anything above 2500 is more than sufficient for day to day computing. ARM has alot more room to grow, don’t be surprised if the 865 in spring 2020 reaches 3500 core M3 levels, or even 4k.

          1. CPU performance is always a moving goalpost though. I wouldn’t be surprised if the 865 matches current day Core M3’s by 2020. But then by 2020, if all goes as planned, Intel will be shipping their 10nm Cannon Lake chips en masse and thus performance will jump too. Not to mention AMD has plans for chips on a 7nm process.

            You’re right though that low end daily computing today doesn’t exactly require high passmark scores. However eventually daily computing performance requirements will also rise as software and OS’s become heavier. Just look at web browsers for example compared to ones from the early 2000’s. The CPU and RAM requirements have skyrocketed since then and will only continue to do so.

          2. Tobi, even a 2000 Passmark is adequate for most people unless they are encoding video or playing back 4K video files or streams. Something like the ten year old Core 2 Duo 3.16GHz E8500 (which I am typing this comment on) will do the job for the vast majority of computer users that need a desktop computer. For a laptop any processor of an i3 or higher should do nicely. Even the ultra low power Core M3 or M5 should be OK for people if they are willing to pay a premium for power efficiency (avoid any computer with “Atom” in its CPU name). With this in mind I don’t see a place for ARM equipped computers. For $599 (hell, even for $499) I can buy an new Intel CPU equipped laptop that will run 6-8 hours (or more) on a battery charge and run circles around any ARM equipped laptop. On the used market I can buy a good one for less than $250. ARM might work well enough for cell phones but I don’t expect it to be able to perform even adequately in a Windows or Linux computer. I was rooked into buying a (then) brand new, low-power CPU equipped laptop (with an AMD E1-2100 1.0GHz CPU), it is a complete dog that can’t even play YouTube videos even after a complete wipe and re-install of the OS and relevant drivers. About all I can really use it for is playback of video or audio files stored on the HDD. Even the latest Windows Feature Update on it took over eight hours to download and install! Linux distros don’t work much better, the CPU just can’t handle internet use even though it is only about four years old. Never again will I buy a low power CPU unless it Passmarks near or over 2000.

    2. I agree. No deal if it’s overpriced like an Ultrabook! I’m not gonna pay premium flagship prices when really all I need is basic computing at an affordable price.

      1. Personal opinion here. These devices are mostly targeting people who are on-the-go (constantly hoping airports, traveling, meeting rooms, visiting clients, etc.). So you end up paying premium for light machine, long battery life and constant connectivity. The machine only needs to handle normal usage in decent speed (internet browsing, emails, calendar, notes, terminal, etc). Qualcomm is not interested is machines lower than 500 USD, they mentioned that at some point.

        1. Too bad early reviews show that the initial ARM devices don’t have that much longer battery life, not that much lighter and as fast as Atom (or whatever it’s called now) based Celerons when running native applications.

          At this price range, you’re better off getting an x86 device with a built-in LTE modem.

          1. The latest Atom is Gemini Lake, which is kinda nice. It does come in Celeron and Premium brands, but TDP is up to 10W, thus high-ends part might be a bit problematic. It’s also SOC (System-on-Chip) thus helps with saving space (this is why we have NUCs), but none come with LTE modem. Intel does know how to build LTE models (I think, some have been used by Apple in iPhones). It would take some years before such product would be developed (assuming they are not working on it for the last year or two).

            I think, the numbers for marketing they have used were mainly based on video playback.

            AnandTech measured active power for Snapdragon 845 on GPU benchmark and it was <4.5W. Gemini Lake Atoms are 6-10W.

          2. Too bad those TDP numbers didn’t noticeably affect the final outcome of these devices. As mentioned, reviews of these initial Windows on ARM devices don’t provide much advantage over similarly priced Intel Core based devices with built-in LTE modems. Some Core M devices are even fanless like with these ARM devices.

            I don’t know, maybe these first few ARM Windows notebooks were just badly engineered.

          3. Core M is also 3-4 times more expensive than the flagship Qualcomm chips. These first iteration devices were baby steps, to get the ball rolling, and set the foundation for when ARM grows even more powerful, it is currently doubling in performance every three years. Something that took Intel 6-8 years, with 10-15% increases every year, then couple of delays.

            ARM will dominate on low end without a doubt.

          4. On the low end ARM certainly is good competition for Intel, but not if WoA devices are price above $400. There’s been a bunch of sub-$300 x86 devices with similar, if not more powerful processors that also have 4G connectivity and so on. If WoA devices can be priced down to the $200 mark, they’ll be extremely strong competition for Intel on the low end of devices. Particularly on devices that won’t be relying on USB accessories.

        2. Precisely, those are flagship chips that go into flagship luxury build devices. The devices are for businessmen and mobile Enterprise workforce, who can easily throw money at em.

          I think the snapdragon 865 would be the point where ARM devices really shine, and hopefully MS can get some big names to recompile to ARM, like Adobe and Mozilla. We already saw demo of emulated Photoshop running on Snapdragon 820. So Adobe recompiling to ARM64 support is no brainier.

          MS should do same for office, they already did once, so they will have browser, email, Adobe creativity suites, office productivity suites, running natively. And UWP apps like one note which are Architecture agnostic, and the PWAs.

          1. Except these ARM devices are in the same price category as Core based devices…

            All this talk about TDP and per-chip costs are nice but looking at the final product, these devices are terrible from a general consumer’s point of view.

          2. There’s no guarantee Adobe or anyone else will recompile to ARM64 native. It’s not as simple as clicking a button to recompile. If the app in question uses any platform native code or x86 Assembly, it’s going to be quite an undertaking to recompile to ARM64. Not to mention even for the apps that are simpler to recompile, it still requires the recompiled app to go through the entire QA process before it can be approved for distribution. Both those scenarios, especially the former, costs a lot of money and resources. So unless Microsoft starts paying developers again, I don’t see what reason or incentive most developers have to recompile or port their apps to ARM64 when there’s an x86 emulator that’s good enough.

  9. Yeah, cannot wait for Snapdragon 845 devices. There is also a Chromebook (2-in-1 with detachable keyboard and incl. pen) in development with Snapdragon 845. I will one of these Snapdragon 845 machines.

    855 with 7nm should also provide a nice boost, but sadly 5G modem will be a separate chip. I guess, one would need to wait for “865” or “875” to have 5G integrated into SOC.

Comments are closed.