Custom Windows OS Claims LOWER LATENCY and MORE PERFORMANCE – Atlas OS

You’d be pretty crazy to say you DON’T want more performance and lower latency from your gaming PC, so the rather bold claims from AtlasOS sound awfully appealing. They say Windows is slow, clunky and unreliable, and more potently, “F**k Windows, Upgrade to Atlas”. They say it’s “Designed for gamers, with improved in-game framerates and lowered latency”. Those are some bold claims, but luckily I built an open source latency testing tool, so we can put those claims to the test!

Installing AtlasOS involves you fresh-installing Windows – yeah, this isn’t a custom Linux based OS, or even a custom Windows image, this is stock Windows with some packages removed. Once Windows is installed, you’ll need to download both the Playbook and the AME Wizard. Run the wizard and import the AtlasOS Playbook. The first thing it’ll tell you to do is fully disable Windows Security. This is a pretty big red flag, but we will come back to that. Once all security is disabled, you can continue the install. It’ll ask you to configure some options, starting with if you want to keep Windows Defender enabled, or disable it such that you’ll need to use Atlas’ tools to re-enable it. They say keeping it on is recommended, and I’d agree. Next is disabling security mitigations, namely things like Spectre and Meltdown. Again, they now recommend leaving them as-is. Next is their recommendation to disable core isolation. Now this is more relevant to Windows 11 as core isolation is enabled by default, and is considerably more feature rich. Atlas themselves list out the possible features, including Memory Integrity, Credential Guard, the Vulnerable Driver Blocklist, Firmware protection, Memory access protection and Kernel-mode Hardware-enforced Stack Protection. It’s a pretty useful set of features, although if you’re running Windows 10, there’s a very good chance you already aren’t using it, so it shouldn’t be too big a deal to switch off on Windows 11 for potentially a decent chunk of performance. Next is their recommendation to remove edge – a sound choice – disable bluetooth – much less sound – and disable power saving. At least that one makes sense. Lastly, you need to pick a browser. Brave is the default choice, then Waterfox, a Firefox fork, then Chrome. I find these options quite funny – they picked Waterfox over Firefox because “privacy”, then give you the option to use Chrome. Not Chromium, Chrome. That’s funny.

Anyway, once you’ve picked your browser, it begins the install. Well, I should say “clearout” actually, as almost all this is doing is nuking a bunch of Windows packages. Once it’s done, it restarts Windows and the installation is complete. It’s worth noting that if you ever want to revert back to fully stock Windows, you will need to reinstall the OS. You can undo a few of their changes with their included scripts, but you can’t revert everything without a fresh install. Still, it’s hard to argue that it doesn’t feel snappier instantly. I think they disable the animations, which means everything feels a fair bit more abrupt and jarring, but undeniably more rapid. So, now we have it installed, let’s run some tests. I’m going to use Rainbow 6 Siege here, and I’m testing stock Windows 10 against the AtlasOS afflicted version of the same image. Both are fresh installs – I did want to use a “dirty” copy I’ve used for benchmarking for a while, but there’s a bug somewhere in it that means Siege won’t run in Vulkan mode, which as I found out in my Rainbow Six Siege latency guide – video in the cards above for you – makes a massive difference – so it’s fresh versus fresh. That’s a little more fair anyway! 

So, what are the results? Well, Siege’s built in benchmark reckons that there is functionally no difference. Atlas is running at 444 FPS and 403 FPS in the 1% lows, versus stock Window running at… 444 FPS average and 401.5 FPS in the 1% lows. Hardly the deal of the century! But, what about the latency? Well, I strapped my Open Source Latency Testing Tool to the display and got to testing. Those results were functionally identical too. AltasOS averaged 11.3 milliseconds, and stock Windows scraped ahead at 11.2 milliseconds. Literally nothing in it. So, how can they claim that stripping out a load of “bloat” from Windows will give you better performance and lower latency? In short, it mostly depends on your hardware. I’m using a relatively modern system here. It’s an i5 13600K and an RTX 3070 Ti, so of course removing a couple of things from running in the background won’t make much of a difference, because even in this more CPU limited game, it still isn’t actually CPU limited. On lower end or older hardware, you are at least a little more likely to see a benefit from disabling these background tasks. It’s also worth noting that a lot of their claimed performance figures – which to their credit they lay out exactly how they tested it – were tested on Windows 11. 11, thanks to the increased security features like Core Isolation, can take a much bigger toll on your performance than on Windows 10. Essentially, Windows 11 is generally worse for gaming, and makes tools like AtlasOS more of a necessity. 

So, you should use AtlasOS then, right? Well……. Probably not – and there’s a few reasons why. First is the security concerns. There are the obvious things like disabling Windows Defender is generally a really dumb idea, and more subtle things like disabling Spectre and Meltdown mitigations which for your average gamer isn’t really likely to be a problem, but is a sign of what’s still on the table in terms of tweaks to get the best possible performance. I think it’s important to mention that the current version of Atlas OS – V0.3.1 – is wildly different to older versions, including the versions you’ll find most of the discussions about AtlasOS are based on. Earlier versions were much more aggressive in removing everything and anything they thought looked like targets for performance or telemetry. They fully removed Windows Defender – irreversibly – they bricked Bluetooth support and a whole host of other things that mostly meant anyone running their tweaks just made their PC very, very insecure and vulnerable. This most recent version is much more refined, and their recommendations seem to have shifted to more sensible choices – Defender is left alone by default, mitigations are left in place, and only Core Isolation is disabled – something that isn’t even enabled by default on Windows 10, and is much more limited there too. Changes like Defender are also reversible using the scripts they provide in the Atlas Folder. You can leave Defender enabled for most of your usage, then disable it to eke out the most performance in games, then switch it back on to be protected again. That’s much better than just nuking it from orbit for good. 

The problem with this tamer iteration is that it then doesn’t stand out from any of the other Windows debloat scripts that have been around for years – the Windows 10 script I use on my systems is dated from 2017, and gives you granular control of every feature and tweak. Want to disable telemetry, the advertising ID and Cortana, but leave Defender alone? No problem. It does things like opens Windows Explorer to This PC instead of Quick Access, a feature I can’t live without. I’ll even install the Linux subsystem – and this is one from 2017! There are a load of scripts and tools like this, with many of them even giving you the option to revert things if you’d like. So, why would you want to use the one that doesn’t give you that granular control over what it’s doing to your system?

The other problem is that Microsoft basically makes a debloated version of Windows already – it’s called LTSC, or Long-Term Servicing Channel, and it has a considerably longer support timeline. It doesn’t come with any bloatware installed, and of course I benchmarked it. Framerates were pretty much the same, at 442 FPS average and a little lower on the 1% lows at 393 FPS, but it was running a slightly different build of Windows to the other two so I’d call that close enough – it’s just 2% slower there. As for the latency, again it’s pretty much identical at 11.4 milliseconds, versus 11.3 and 11.2 on Atlas and stock Windows 10 respectively. Hell, you can still run a powershell script like the one I use on this if you’d like to take the tweaking even further! 

To summarise this lot then, AtlasOS isn’t your only option for tweaking your Windows installation, and I’d argue that it’s more closed off options give you less control than the numerous other options available. AtlasOS markets itself as, well, an OS. It proudly proclaims, “F**k Windows, Upgrade to Atlas”, and that “Windows is slow, clunky and unreliable”, but especially in its most recent iteration, it’s just a fresh install of Windows with a few tweaks. I couldn’t find any performance gains, but if you do, let me know what hardware and games you’re running in the comments below. Personally, if you want the debloated Windows experience, install the LTSC version instead. It comes straight from Microsoft with most of these tweaks pre-done for you, and even comes with a much longer support window too. Better yet, clean up your current install and run one of the debloat scripts to tweak everything to your heart’s content. I’ll link the script I use in the description if you’re interested. Equally, if you’re interested in testing latency of games, displays and peripherals, you can pick up an OSLTT unit at OSRTT.com – I built them by hand at my desk downstairs, so they are actually made in the UK!