How to get the best performance out of PlayerUnknown's Battlegrounds

PlayerUnknown's Battlegrounds is one of the hottest multiplayer games right now. While it's still in Early Access, with over two million copies sold, people are constantly queuing up to take another stab at being the last man standing. Whether you like to go in hard and fast, or prefer the stealthy (aka cowardly) approach, the game is a constant adrenaline rush.

Being Early Access means that performance is likely to improve over time. Or maybe not—just look at Ark: Survival Evolved for a game that continues to punish moderate PCs. But at least during the past two months, there have been clear improvements in framerates and plenty of patches. If you're a fan of competitive games like Overwatch or Counterstrike, you should know upfront that Battlegrounds is far more demanding of your hardware. It uses Unreal Engine 4, which has a good reputation for image quality, but it can also tax even the best systems.

One thing you don't want to do is to show up running on integrated graphics, and even many budget GPUs are struggling with the current release. CPU performance can also play a lesser role, particularly with minimum frame rates, but as usual having a fast GPU should be your first priority.

Quickly running through the features checklist, even at this early stage Battlegrounds has plenty of graphics options, and it checks most of the right boxes. Resolution support is good, though modifying FOV is no longer allowed. The devs have said they'll add an FOV slider in a future update, but until that happens everyone is locked in to the same FOV setting.

About the only other areas that are slightly questionable are the modding potential and controller support. Game controllers are supported, but last I checked, you were stuck with keyboard and mouse or the game controller—which means most people will want to unplug the controller before launching. (This should get fixed with a future update.) On the modding front, the developers have said they'll support it at some point in the future, but we'll have to wait and see.

A word on our sponsor

As our partner for these detailed performance analyses, MSI provided the hardware we needed to test PlayerUnknown's Battlegrounds on a bunch of different AMD and Nvidia GPUs, multiple CPUs, and several laptops—see below for the full details. Thanks, MSI!

In terms of graphics settings, there's an overall preset, which configures the seven individual items to the selected level—very low, low, medium, high, and ultra. There's also a motion blur option, which thankfully defaults to off in all cases. If you want more information on how the various settings affect performance, head to the bottom where I'll dig into more details.

For the benchmarks, I've used my standard choice of 1080p medium as the baseline, and then supplemented that with 1080p, 1440p, and 4K ultra. Of course, there's a catch with games like Battlegrounds: some people will want to run at minimum graphics quality, except for the view distance, to try and gain a competitive advantage. It looks ugly, but it's easier to spot people hiding in the grass or shadows.

MSI provided all of the hardware for this testing, mostly consisting of its Gaming/Gaming X graphics cards. These cards are designed to be fast but quiet, and the fans will shut off completely when the graphics card isn't being used. Our main test system is MSI's Aegis Ti3, a custom case and motherboard with an overclocked 4.8GHz i7-7700K, 64GB DDR4-2400 RAM, and a pair of 512GB Plextor M8Pe M.2 NVMe solid-state drives in RAID0. There's a 3TB hard drive as well, custom lighting, and more.

MSI also provided three of its gaming notebooks for testing, the GS63VR with GTX 1060 6GB, GT62VR with GTX 1070, and GT73VR with GTX 1080. The GS63VR has a 4Kp60 display, the GT62VR has a 1080p60 G-Sync display, and the GT73VR has a 1080p120 G-Sync display. For testing higher resolutions on the GT-series notebooks, I used Nvidia's DSR technology.

For CPU testing, MSI has provided several different motherboards. I have the X99A Gaming Pro Carbon for Haswell-E/Broadwell-E testing, Z270 Gaming Pro Carbon for additional Kaby Lake CPU testing, X370 Gaming Pro Carbon for high-end Ryzen 7 builds, and B350 Tomahawk for budget-friends Ryzen 5 builds.

Let's hit the benchmarks

Skydiving into our benchmarks, 1080p medium immediately shows how CPU bound PUBG can get. The Nvidia 1080 Ti, 1080, and 1070 all land at the top of the chart, but the 1080 Ti actually falls below both the 1070 and 1080. This is likely an issue with drivers and/or game optimizations, and it should get corrected at some point, but it's still a bit odd to see 1080 Ti underperform. Not that you'd really want to use a $700 GPU for 1080p medium gaming.

Moving down the chart, the 1060 6GB and 1060 3GB come next. Where is AMD? Lounging in the bottom half of the chart. The RX 470 normally outperforms the GTX 1060 3GB, but Nvidia is clearly doing better in Battlegrounds right now. The RX 470 4GB has spotty minimum framerates and stutters, and the same goes for any of AMD's 4GB cards.

And if you want more evidence of AMD driver problems, look no further than the RX 460 2GB, which can't even get out of the teens—at minimum quality, the card still only manages 21 fps. The 4GB 460 does better, though it still underperforms, but the GTX 1050 proves that a 2GB card can actually run these settings without falling flat on its face.

If you're gunning for 60 fps or more, budget GPUs will generally need to drop to the low or very low preset, and some of them will still come up short. I suspect we'll see an updated AMD driver in the near future to improve things, given the popularity of Battlegrounds, and hopefully some game engine tweaks as well.

1080p ultra drops performance on the fastest cards by about 20 percent, while slower GPUs lose about 33 percent relative to medium quality. Most mainstream and above cards are playable, but if you're looking for 60+ fps, the only way to get there is with an Nvidia GPU right now—GTX 1060 6GB (or the 2015 model GTX 980 4GB) barely clear that mark, though minimum fps will still be a concern.

One of the issues with testing games like Battlegrounds is that performance optimizations typically happen later in the game development process—the opposite of Early Access. Again we see the 1080 Ti fall below the 1080 performance. A few months from now should change things for the better, but regardless, Battlegrounds is one of the more demanding games currently available.

Given the above results, 1440p ultra is clearly going to be a problem for all but the fastest GPUs. You pretty much need a 1080 or 1080 Ti to break 60 fps, though a few minor tweaks will get the 1070 there (or a 980 Ti, which performs similarly). At least the 1080 Ti is finally able to brute force its way ahead of the 1080.

AMD's RX Vega should be out within the next 30 days, and the current rumors are that we'll see three different models. Hopefully AMD will also have drivers that tune Battlegrounds performance a bit more, as the Fury X is clearly not living up to its potential. As for the RX 460 2GB, I couldn't even complete the benchmark sequence, and resorted to spinning in a circle in stop-motion fashion.

Notice that VRAM is also becoming a defining factor, with the 1060 6GB showing much more consistent framerates than the 1060 3GB. Going forward, I would strongly recommend that anyone purchasing a 'mainstream' card ($150 or above) should set their sights on 6GB or 8GB VRAM.

In typical fashion, it should be no surprise that 4K ultra kills performance—even the 1080 Ti falls short of 60 fps. If you want to game at 4K, the 1080 Ti can get there at medium/high quality (give or take), while the vanilla 1080 will need to opt for mostly low/very low settings—see below for details on which settings have the greatest impact on framerates.

There's a darker side to this chart as well. Unreal Engine 4 uses deferred rendering techniques that, generally don't work with multi-GPU. I did check SLI 1080 performance, and it was worse in all cases compared to a single GPU. Unreal Engine 4.15 did add an option for AFR SLI support, but Battlegrounds is using an older build, and it's still not clear if multi-GPU will ever happen. Considering the apparent CPU bottlenecks (see below), other items are likely much higher on the list before the developers even think about tackling multi-GPU.

Not that you really need to game at 4K ultra, right?

CPU performance scaling

Image 1 of 4

Swipe left/right for more images.

Image 2 of 4

Swipe left/right for more images.

Image 3 of 4

Swipe left/right for more images.

Image 4 of 4

Swipe left/right for more images.

What about the CPU side of things—how many cores does Battlegrounds need to run properly? Given the 1080 Ti results (and its price), I've decided to standardize on the GTX 1080 8GB for CPU scaling. I tested that card on four Intel CPUs along with four AMD Ryzen CPUs.

The urban settings in Battlegrounds have far more polygons and other objects to render, resulting in much lower framerates—out in the grassy countryside, performance can be up to 50 percent higher.

I do need to mention a game update that came out between the Intel Core and AMD Ryzen testing, which may be why Ryzen performs slightly better in some cases. Regardless, it's pretty clear that the most important thing with your CPU will be raw clockspeed, provided that you have at least four physical CPU cores. All four Ryzen CPUs were tested at the same 3.9GHz overclock—an overclock I expect nearly all Ryzen chips to reach. While SMT does add 6 fps at 1080p medium, going beyond 4-core/8-thread has a relatively minor impact on performance.

This is even more the case at 1080p ultra and above, where the four Ryzen CPUs are very nearly tied. The 8-core 1700 does come out on top, but not by a huge margin. Meanwhile the 2-core/4-thread Core i3 part struggles, particularly on minimum fps.

Of course, most of these CPU limitations are only visible with an ultra-fast graphics card. Using a slower mainstream card like a GTX 1060 3GB, Core i5 and Ryzen 5 are more than sufficient. I didn't run a full set of benchmarks, but there's less than a five percent difference between the i5-7500 and i7-7700K using a 1060 3GB.

There's also a "-USEALLAVAILABLECORES" launch option that may help some people. I didn't see a difference in limited testing with an i7-5930K, but I haven't checked performance with lower-end CPUs yet.

Mobile Battlegrounds

Image 1 of 4

Swipe left/right for more images.

Image 2 of 4

Swipe left/right for more images.

Image 3 of 4

Swipe left/right for more images.

Image 4 of 4

Swipe left/right for more images.

Shifting gears to notebook testing, the mobile CPUs aren't able to keep the GPUs fully fed with data. Clockspeed matters more in Battlegrounds than core counts, so the 4.8GHz i7-7700K can distance itself from the i7-7700HQ, which runs at around 3.5GHz while gaming.

The GT72VR has a slightly faster i7-7820HQ, plus better cooling to keep it running at higher Turbo clocks, but it still can't touch the desktop 7700K. Thanks to the CPU bottleneck, the mobile GTX 1080 actually falls below the desktop 1060 6GB card at 1080p medium, at least when combining average fps and minimum fps. At 1080p ultra, it's still slower than the desktop 1070. That's not normally the case, but here we're dealing with about a 20 percent CPU clockspeed deficit.

Once the notebooks are running at 1440p and 4K (via DSR on the GT62VR and GT73VR), the CPU bottleneck becomes less of a factor. The 1080 is nearly the same performance as the desktop card, while the 1070 and 1060 are 10-20 percent slower due to clockspeeds and thermal constraints. Still, it's nice to see gaming notebooks that are mostly within striking distance of desktop PCs—albeit at substantially higher prices.

Stop killing me, I'm running benchmarks!

Benchmarking Battlegrounds is a completely different beast from singleplayer games, thanks to the randomized starting locations. Trying to get a repeatable benchmark sequence with multiple people in the vicinity just isn't going to happen—if anyone is nearby, I'm likely to wind up dead long before I can finish testing. Plus multiple players in the same area tends to drop framerates, so if I ended up with too many people, I'd just quit/suicide and restart.

The urban settings in Battlegrounds have far more polygons and other objects to render, resulting in much lower framerates—out in the grassy countryside, performance can be up to 50 percent higher, but that's usually a shortcut to getting killed. Cities and buildings are where the action takes place, and I eventually settled on a benchmark location in Yasnaya Polyana. It's not quite as popular with the locals as some of the other cities, but it's still reasonably accessible.

Assuming I can get to my starting point without dying, I then run laps around a building in the northeast section of the city, logging framerates at the various settings and resolutions. I tested each setting twice, to control for variables like other players, using the best result (and sometimes running one or two more tests if the first two results were wildly different).

As noted earlier, I tested at medium and ultra quality. I suspect a lot of competitive players are running at the minimum quality settings (except for view distance) to gain a competitive advantage, but the degraded visuals really hurt the overall experience in my book. If you drop from medium to very low quality, that will usually boost framerates by 30-50 percent, about double the performance of the ultra setting—at least if you have a sufficiently potent CPU.

Fine tuning performance

The following numbers are not from running complete benchmarks on a bunch of different cards, but were gathered using a single GTX 1080 GPU running at 1440p, and noting the average framerate. I specifically selected a location where performance was lower, then started checking each setting.

The global preset is the easiest place to start tuning performance. Ultra quality gave a baseline score of 65 fps. Dropping to high improved performance to 81 fps, medium was 92 fps, low was 105 fps, and very low was 107 fps. CPU performance is still a factor even on the very low setting, so most systems won't be able to get much above the 100-130 fps range (depending on the area of the game world, naturally).

There's currently a 144 fps framerate cap in effect, which the devs have said they plan to remove in the future (it used to be 120 fps). Getting to 144 fps in my experience is difficult, as even at very low settings Battlegrounds tends to be CPU limited to around 120-130 fps during my testing.

If you're hoping to tweak the individual settings to better tune performance, I did some quick checking of how much each setting affects framerates using the ultra preset as the baseline. I then dropped each setting down to the minimum (very low) to see how much it helped.

Screen Scale: The range is 70-120, and this represents undersampling/oversampling of the image. It's like tweaking your resolution by small amounts, but I mostly recommend leaving this at the default 100 setting.

Anti-Aliasing: Surprisingly not a major factor, but this is because Unreal Engine requires the use of post-processing techniques to do AA. If you want better AA, you could set screen scale to 120 to get a moderate form of super-sampling. Going from ultra quality AA to very low quality AA only improved performance by 3 percent.

Post-Processing: A generic label for a whole bunch of stuff that can be done after rendering is complete, this is a moderate impact on performance—going from ultra to very low improved framerates by 10 percent.

Shadows: No surprises here, this is the most demanding individual setting (outside of resolution). This setting affects ambient occlusion and other forms of shadow rendering, and going from ultra to very low improved performance by just over 20 percent.

Texture: Only a minor impact on performance, provided you have enough VRAM. Dropping from ultra to very low increased framerates by 8 percent.

Effects: This setting relates to things like explosions, among other elements. It might have a more noticeable impact on performance with explosions going off (eg, in a red zone), but for the test scene dropping to very low only improved performance by 1 percent.

Foliage: Given all the trees and grass, you might expect this to have a larger impact on performance, but I only measured a 1 percent (1 fps) difference after setting it to very low.

View Distance: This has a much greater impact on CPU performance than on graphics performance, so if your CPU is up to snuff you can safely set it to ultra. On a Core i7 system, dropping to very low only made a 1 percent difference in framerates.

Motion Blur: There's a reason this is off by default, right? Spotting enemies while moving around is more difficult with motion blur enabled. But if you like the effect, turning it on has almost no impact on performance, maybe 1-2 percent.

Thanks again to MSI for providing the hardware. All testing was done with the latest Nvidia and AMD drivers, Nvidia 382.05 and AMD 17.5.1. My recommendation for Battlegrounds players is that Nvidia GPUs are the superior solution, for what should be obvious reasons. Hopefully AMD can improve the situation with new drivers, but currently even the fastest AMD GPUs are struggling.

If you already have an AMD card, plan on dropping settings below ultra if you want to hit 60 fps. Either that, or hope that game updates and new drivers will eventually improve the situation. AMD's 2GB cards are currently unable to handle the game at all, with the 460 2GB averaging just 20 fps at minimum quality 1080p. That's part of the pre-final nature of the game.

That wraps up my current Battlegrounds testing. I plan to revisit the game once it leaves Early Access. In the meantime, if you see me running laps around a building in Yasnaya, pop on over and say hi—maybe bring me an energy drink while you're at it. Whatever you do, don't shoot me—I'm only here for the science!