Windows 10 performance is one of the hardest things to nail down right now. Testing of the new OS, due to ship on July 29, isn’t exactly easy, because the public doesn’t yet have a way to install the RTM version. Microsoft has been pretty mum about any under-the-hood changes. Even worse, Microsoft’s hardware partners have been handcuffed from talking about the OS at all.
One thing we do know: Among the many big changes over Windows 8, Windows 10 may literally change the game in graphics performance as the only Microsoft OS with DirectX 12. Early testing of the OS shows performance gains elsewhere will be far smaller, though.
For testing, I used two identical HP Spectre x360 laptops. One had build 10240 of Windows 10 Home, which Microsoft has officially blessed as “reviewable code.” The other had Windows 8.1 Home.
Both laptops have the same screen, same battery, same 8GB of LPDDR3, same 128GB SSD make and models, the same Intel Core i5-5200U CPUs, and the same BIOSes. Other than OSes, they are exactly the same.
I threw a bunch of different benchmarks at the machines to see if I could coax out any performance differences. I ran compression tests, chess benchmarks, and 3D rendering, as well as a spate of DX9 and DX11 benchmarks.
We were hoping for a dramatic face-off. But what we got was essentially a tie—results so close, they were separated by a margin of error.
For example, here’s the performance in PCMark 8 Creative Conventional. PCMark, for those who don’t know, simulates various “real-world” application loads. The Creative Conventional, for example, tests simulated web browsing, photo editing, video editing, gaming and group video chat. The Home load adds writing and casual gaming.
As the chart above indicates, performance is near identical. Similar results persisted just about everywhere between Windows 8.1 and Windows 10: mostly a tie with some loads giving Windows 10 a very slight edge. I could produce a lengthy page of graphs between the two in WinRar, CineBench, Valve’s old Particle Test, 7Zip, Passmark and 3DMark, but what’s the point? The bars would be almost the same on all of them, and I’d just be wasting Internet bandwidth. If you want to see a graph, just take the chart above and change the name of it to 7Zip… or whatever test you want.
I won’t throw away a day of testing without giving you more numbers, though, so here’s some of the results I saw. Other tests I simply didn’t record because after a while, it felt silly essentially writing down the same number twice.
But what about gaming?
There is more promise for Windows 10 on the gaming side. DirectX 12, in games that implement it, should see healthy improvements. But in games that don’t use DX12, it’s probably going to be a lot closer.
For example, I ran Tomb Raider on our PCWorld zero-point system. It has an Intel Core i7-4770K, 16GB of DDR3/1600, and a GeForce GTX 980. I used the same Nvidia drivers with both OSes. Note that I ran Wndows 10 build 10162 rather than the current build 10240, as that’s the last ISO of Windows 10 preview that Microsoft made available—no amount of coaxing would get Microsoft’s servers to kick down anything newer. My tests show a definite, if small, edge for Windows 10. Here’s Tomb Raider for you to gawk at:
Other tests gave up a little more of a win for Win10, but this won’t set the world on fire like DX12 is expected to once games that support it are out. We’re planning a more in-depth look at Windows 10 gaming performance, so stay tuned.
Here are the takeaways: The first is that despite all the Windows 8 hate out there, the OS is actually quite fast. Anecdotal reports I’ve seen from when Battlefield 4 was released, for instance, attributed many performance improvements to running Windows 8 over the beloved Windows 7. Windows 8 offered improvements in video and audio decoding that made it faster, too. Another takeaway is that if Windows 8 was zippy, Windows 10 will be, too.
This isn’t the last word
There’s a lot my initial testing doesn’t cover. Battery life improvements, file system improvements and other areas may indeed have been buffed up by Microsoft. Once I get a proper ISO of the OS, I can perform clean installs, and I’ll have a better feel for its performance outside of the areas I’ve touched on today.
OS performance testing has other challenges. Many of the benchmarks I ran are designed to test hardware, not the OS. Cinebench R15, for example, is a pure CPU test, though the OS has some impact. Windows Vista famously destroyed USB performance until SP1 was released, and the overhead from the OS can pull down performance elsewhere too.
Windows 10 seems to offer basically no relevant performance advantage over Windows 8 in mainstream tests, but let’s not be too negative—because there’s no reason to be. With Windows 7, Microsoft updated the scheduler for how the OS dealt with CPUs, which promised improvements and battery life savings for both Intel and AMD CPUs. That wasn’t a check-off item for Windows 10. because Windows 8.1 performance was already very good.
And, again, let’s not forget that Windows 10 ushers in DirectX 12, which should very much yield significant performance increases in games that support it.