So, let’s assume that you’re primarily a gamer and want to squeeze every last drop of performance possible out of your rig, does Intel’s Hyper-Threading technology hurt the frame rate, or are you better to leave it enabled?
Intel’s Hyper-Threading has been part of Intel’s gaming-focused CPUs since back in the days of the Pentium 4, leveraging the longer CPU pipeline on the Netburst architecture to keep system performance up. Of course, games weren’t exactly designed to take advantage of multiple CPU cores/threads. Even with just two threads available, titles like Quake 3 were in the minority of even acknowledging that they existed.
But, this is late 2019 and times have definitely changed – and CPU core counts have increased drastically. But, let’s say in the case of an I9-9900K which has 8 CPU cores and also Hyper Threading. The CPU is capable of handling 16 threads, but this isn’t the same thing as a ‘real’ CPU core.
So these two logical processors running on the single physical core generally are sharing resources such as cache, execution units and so on. This means that some applications (such as video encoding and 3d rendering) scale really well with HT, and other applications (generally games) don’t offer as much of an improvement.
For our testing then – the methodology is pretty simple. We take an Intel i9-9900K (thanks to Intel for the sample) and crank it to 5GHZ, and then add in a GeForce RTX 2080 Ti and patch the games and windows to their latest relevant version. We then use the BIOS of the MSI MPG Z390 Gaming Pro Carbon motherboard (thanks MSI) to enable or disable HT on the I9-9900K for our gaming benchmarks.
We will run at different resolutions and games but focus mostly on lower resolution testing – because even the GeForce rtx 2080 Ti will bottleneck games such as Gears 5 at higher resolutions. The Intel I9-9900K will not even be slightly pushed at 4K compared to what’s happening to the GPU.
As you can see above, we’re running with the driver version of 436.30, and have the 2080 Ti clocks at default – but cranked the power limit up to 111% which gives a bit of extra headroom.
Also – one final note about our testing – the test system is ‘clean’ – in other words, no Discord in the background, no 100 Chrome tabs, no Spotify and so on. Just the game is running, along with MSI Afterburner (if you’re watching the video and wondering how we captured benchmark footage, we’re using a second machine with a capture card).
The reason that’s important is that the results could be very different if the system is being more taxed, and I had considered throwing in some applications running in the background for separate tests… but decided against it as everyone’s workload is different.
With that all said, let’s take a look at the results:
Assassin’s Creed Origins is both CPU and GPU intensive, at 1440P HT enabled wins, but disabling the I9-9900K’s Hyper-Threading at both 1080P and 4K gives you a staggering 1FPS gain.
I’m still not quite ready to retire Batman Arkham Knight from gaming benchmarks, as it scales so well even on modern hardware. At 1080P, the I9-9900K doesn’t look any different no matter what you do with its Hyper-Threading.Next game is a new one in our test suite, and World War Z benefits universally from running the Vulkan API. For this gaming workload, it’s not really any different. The 9900K scores 228 on Vulkan with either HT on or off. 1440P DX11 and disabling HT wins out, and 1440P Vulkan HT on wins out… but by 2 FPS.
Civilization 6 AI benchmark and in this particular game, we want the lowest possible score in turn time using DX12. HT enabled wins out here. Easily one of my fav games this year, Gears 5 is a graphical showcase for both the Xbox consoles as well as the PC. It’s also very GPU demanding, and the I9-9900K often finds itself twiddling its thumbs because the GPU is just taped out. We’ve disabled minimal frame rates here, or otherwise, the resolution will dynamically adjust. Another gaming benchmark that realistically you wouldn’t have a clue what you had the I9-9900K’s HT set to.
The older two Metro games were fantastically demanding on the GPU, and Metro Exodus keeps up this tradition. The built-in Benchmark gets more taxing the further it gets into the scene. We’ve set things to Ultra here, and ray tracing and so on is disabled. I’ve also included 720P results because of the RTX 2080 Ti holding back the I9-9900K.
At 720P there’s a nod towards leaving HT enabled, but at 1080P when we’re GPU bound… there’s no real difference.
The final game in the built-in benchmarks (next up are manual runs), Shadow of the Tomb Raider. At 720P, HT off takes a lead of 9FPS., but at 1080P and higher the I9-9900K’s thread count isn’t making a blind bit of difference.
And now let’s take a look at some manual runs of games, which may have a very slight margin of error. We’ll start things again with Gears 5, and with once again with or without Hyper Threading, the results here aren’t much different. Even at 1080P, Gears 5 frequently can become CPU bound, but technically speaking HT disabled is a slight win… but not by much.
resident Evil 2 is next, and we’re running at 1080P with DX11, in this particular title DX12 mode consistently provides worst results (something we’ll explore further in the future. But here HT on with our I9-9900K is a slight boost in performance, but it’s well within the margin of error.
Blair Witch is a first person horror title and even at just 1080P once again heavily pushes the GPU, leaving the I9-9900K largely out of the question. We saw a few fewer dips with Intel’s Hyper-Threading enabled when loading into a new area (we’re using an SSD), but it lasted so briefly that it made no difference at all to the average.
Another title that’s fairly old, but not exactly easy to run, Witcher 3. The rtx 2080 Ti is really pushed at 1440P, and essentially we’re left with it being a toss up. The additional threads helped the min FPS (once again, it seemed to slightly reduce stutter on loading when on horse back and galloping fast), but overall its not much of a difference.
I decided to run BF5 twice, once with DX12 and once with DX11 – both times at 1080P. First thing – DX12 really trounces DX11 in this title, whether you’re using HT or not. Secondly, HT enabled does seem to lose out to HT disabled with BattleField 5. If you are playing competitively on a high refresh rate monitor, it might be worth disabling it… but honestly the average FPS wasn’t too much different.
Conclusion time I9-9900K Hyper Threading for Gaming
Well, my advice is that you leave it enabled – unless you’re a competitive gamer and attempting to squeeze every last FPS out of a title that does benefit with it off. For the average person, if gaming with other applications in the background, any extra performance might be vacuumed up by a few Chrome tabs or your antivirus updating in the background.
The other thing we didn’t test is that the i9-9900K does generally run cooler with HT off compared to on, and also consumes less power. This might lead some people to get better overclocks with their i9-9900K if they choose to turn off HT. But obviously, all of this is silicon lottery and also depends upon your particular cooling solution.
Another legitimate question you might have is why not just buy an Intel I7-9700K if you’re going to buy a I9-9900K and then turn off HT. The answer here is that typically (and once again, silicon lottery comes into play), the I9-9900K overclocks better than what the 9700K could. Plus you get additional L3 cache too, which is certainly not going to hurt performance.
We’ll be running similar testing on an AMD system soon, so be sure to check back if you’d like to see that.