‘Final Fantasy VII Remake’ Running Faster On Linux Than Windows

WILL IT LINUX? Oh yea! Here’s an early holiday gift for the Linux gaming enthusiasts out there. The newly-released PC version of Final Fantasy VII Remake seems to runs faster on Linux than it does on Windows 10at least on my test system.

I’m always eager to test new PC games on both Windows and Linux to evaluate differences in performance. Normally, my goal is to help Linux graphics developers and the people working on Proton (the compatibility layer allowing Windows games to run on Linux — also the Steam Deck’s secret weapon) target areas for improvement on a per-game basis.

This recently yielded positive results with Forza Horizon 5, where we identified some stuttering in certain hardware configurations. Thanks the community’s help, that issue is being worked on right now!

But I always have a secondary goal when running these benchmarks. I’m looking for that magical moment when a game performs better on Linux than it does on its native OS, Windows.

The moment just happened, and I’m absolutely thrilled to report on it!

Let’s Get Heroic

Final Fantasy VII Remake is available exclusively for PC on the Epic Games Store. (And let’s avoid talking about that price tag so we can keep things upbeat, yea?) So that meant getting some hands-on time with the Heroic launcher, an open-source alternative to the Epic Games Store.

(By the way, Heroic is also available for Windows if you’d rather avoid the official client.)

It was surprisingly easy to get Final Fantasy VII Remake up and running on this client. No command line voodoo, no crazy tweaking required. In fact, since I already had the game installed on Windows, I didn’t even need to download the massive 100GB package again. Instead, Heroic let me import the existing game files that live on that separate Windows drive. By doing so, the game client basically cataloged that installation without needing to copy the files over to my Fedora 35 partition.

The small victories continued when Heroic automatically found my Proton-GE folder and presented those customized versions as an option for running the game. Very cool!

The Testing Playground

Before we get into the actual results, let’s talk about the playground.

I do all my serious game testing on a custom-built Falcon Northwest Talon (even if I do all my serious playing on an Xbox Series X now).

Here are the crucial specs:

  • ASUS ROG Hero VII Motherboard
  • Liquid-cooled AMD Ryzen 3900X
  • AMD Radeon RX 6800 XT (Reference model)
  • 2x Samsung Evo 970 1TB NVMe SSDs
  • 32GB G.Skill DDR4 RAM at 3200MHz

Since I’m helping out the MESA graphics driver gurus when I can, I have a “leading-edge” installation of Fedora 35 on one drive, and a fully updated Windows 10 (with the newest Radeon driver) on the other.

I’ve upgraded my Fedora 35 installation with the fsync kernel, which rolls in several gaming performance optimizations. And it’s quite easy to install if you’re on Fedora! So for benchmarking Final Fantasy VII Remake, I’m running on kernel 5.15.7 with the open-source MESA graphics driver version 21.3.1.

On the Windows side of things, I’m running Windows 10 Pro 21H1 with Radeon driver version 21.12.1 (which released December 1, 2021).

Final Fantasy VII Remake ❤️ Fedora 35

Down to business, then!

To test the framerate performance on Final Fantasy VII Remake, I switched in-game resolution to 3840×2160, with texture and shadow resolution set to High. I set the framerate cap to 120FPS (there is no “unlimited” option) and left HDR off.

Then I fought Cloud’s way through the first 3 battles during the opening scene of the game, while using MangoHUD to log performance on Fedora 35, and AMD’s built-in performance logging on Windows 10.

Crucially, I logged the second run through this section, in order to warm up the GPU and ensure that shaders would be properly processed.

It’s also worth mentioning that no power profiles were changed on either OS. Nor was anything overclocked or underclocked.

Here’s what it looked like when both benchmarks crossed the finish line:

In spite of any expected performance hit incurred by using the Proton compatibility layer on Linux, Fedora 35 turned in an average framerate 20% higher than Windows 10.

What is most impressive about these results is that Final Fantasy VII Remake uses DirectX 12 on Windows. That’s VKD3D on the Linux side of things, which is less mature than its DirectX 11 equivalent.

There have also been significant improvements to the open-source Radeon drivers lately (in part, driven by the upcoming Steam Deck launch) but it’s still an awesome thing to witness Linux outperforming Windows with a non-native version of the game!

But There’s A Catch…

In terms of raw performance, Linux wins, and it’s a scenario we’ll start seeing more frequently. But to many gamers, how smooth a game plays is equally important.

In the frametime chart above, you can see that although it’s turning in a lower overall framerate, Windows 10 (red) churns out smoother performance. Meanwhile, Fedora 35 (blue) exhibits a few harsh drops in performance that manifest as noticeable stutters in-game. They’re not game-breaking or severely distracting, but they’re there.

On the bright side, the same thing happened initially with Forza Horizon 5 on Linux (also DX12/VKD3D) but diligent testing by the community led to a proposed solution. I have the highest confidence we’ll see the same thing happen very soon with Final Fantasy VII Remake.

Closing Thoughts

I have very nostalgic feelings about Final Fantasy VII and the long-anticipated arrival of Remake on PC. Do I wish it would have landed on Steam? Yep. Not just because I prefer Steam (especially for Linux gaming), but I was initially worried that getting it to run on Linux would be a chore.

I’m just used to the ease of Steam Proton, and had very little experience with the likes of Lutris, Heroic, etc. So this little test has been eye-opening and fruitful, not just because of the performance win, but because I finally got to see how terrific Heroic has become.

Seriously, it’s intuitive to use and presents so many intelligent options. Want to enable AMD FSR on a per-game basis? There’s a switch for that. Esync? Fsync? Check! Resizable BAR support for Nvidia RTX owners? You got it.

I enjoy using it more than Epic Games, but your mileage may vary.

Obviously a lot of testing remains to be done in order to be conclusive about this victory. And it would be great to have a “Sanity Check” from someone with a similar hardware setup.

Some of the answers we need as a community:

  • How well does Final Fantasy VII Remake run on Arch-based distros?
  • At 1080p and 1440p?
  • With different CPU and GPU combinations?
  • With AMD FSR on and off?
  • On Nvidia GPUs?
  • In larger open city sections of the game?

If you’re doing any testing of the game, let me know your results here in the comments section, or by talking to me on Twitter.

UPDATE, Friday Dec 17 @ 10:42am Eastern: Switching the Windows 10 power plan to “Ryzen High Performance” and the RX 6800 XT to auto-overclock results in average framerate of 115 FPS. That puts it neck-and-neck with Fedora 35. Still mighty impressive.

RESOURCES:

Share This

3 thoughts on “‘Final Fantasy VII Remake’ Running Faster On Linux Than Windows

  1. I have 120 FPS stable on my 5950X and 6800XT in 1440p Max settings. I haven’t tested in Windows but my laptop also pumps out 120 FPS stable in 1080p Max Settings with an i7-10870H and 3080 80W. This is absolutely spectacular news! 20% difference is nothing to scuff at. A proper reason for gamers to use Linux.

  2. The stuttering issue appears to be with the port itself, Windows users are reporting that too.
    It’s a good thing because this means that Wine/Proton compatibility is getting closer and closer to native Windows.

  3. So glad they remade this game God I had so many memories with the original was the first RPG turn based game I had ever experienced I just remember feeling like damn this game takes forever but I can’t stop playing it I have to Know How It Ends. Thanks for guidance now i am running game again on Linux

Join the discussion!