Shadow of Mordor: Performance Tests

topmordor

As mentioned in the previous post, Shadow of Mordor finally made it to Linux thanks to Feral. Since that game includes a proper benchmark tool, we can actually compare how well this port runs versus the Windows version. Sadly, there is still a wide gap in performance as we will see below.

The first video below tested its performance on GTX 680, a three years old card from nVidia, but one of the higher end cards of the 600 series. The configuration was as follows:

  • CPU: Intel i7-4770 at 3.4Ghz
  • GPU: NVIDIA GTX 680 2GB
  • RAM: 8GB DDR3 Corsair 1333MHz
  • OS: Ubuntu 15.04 64Bit/Windows 10 Pro 64Bit (Dual Boot)

The result leaves nothing to speculation:

  • Max FPS: 81.40 on Windows vs 50.87 on Ubuntu [ 62% of the Windows Performance ]
  • Average FPS: 55.83 on Windows vs 30.16 on Ubuntu [ 54% of the Windows Performance ]
  • Lowest FPS: 31.65 on Windows vs 6.84 on Ubuntu [ 22 % of the Windows Performance ]
  • Amplitude (Max vs Min FPS) : 49.75 on Windows vs 44.03 on Ubuntu

As you can see the game runs about half as fast as the Windows version on average. Even though we are lucky to have this port, frankly it’s hard to find such gap acceptable. If we could get within 10-20% of the Windows version framerate it would not be such an issue, but half as much makes it very hard to recommend Linux for gaming at all in this particular case.

On the same day another Linux user shared a benchmark on newer GPU card. the GTX980 – which is tbe best nVidia card available currently (if you exclude the newer Titan).

  • CPU: Intel I7 4790k at 4Ghz
  • GPU: Nvidia GTX980 4Gb
  • Nvidia drivers: 352.30 on Linux and 353.62 on Windows
  • RAM: 8Gb
  • OS: Ubuntu 15.04 64 Bit / Windows 8.1

As you can see the key difference here vs the previous configuration is the GPU card. The CPU is very much in the same class while clocked higher in this case.

This time the gap is not as bad as for the previous case:

  • Max FPS: 152.18 on Windows vs 114.33 on Ubuntu [ 75% of the Windows Performance ]
  • Average FPS: 107.99 on Windows vs 73.57 on Ubuntu [ 68% of the Windows Performance ]
  • Min FPS: 75.36 on Windows vs 39.14 on Ubuntu [ 52 % of the Windows Performance ]
  • Amplitude (Max – Min) : 76.82 on Windows vs 75.19 on Ubuntu.

It seems like a newer card is doing slightly better on Linux vs a less recent one. This is particularly showing if you compare the improvement facts between the two benchmarks:

  • On average the GTX980 gets 193% more frames per second on Windows vs the GTX680, while on Linux the increase is at 243 %.

But still, less than 70% of the Windows performance on average is hardly stellar. It would not matter that much if the Linux figures were anyway above 60 FPS, but it’s not the case: while we lack the information regarding the distribution of its performance, there’s a good likelihood that 30% of the time you will get less than 60 fps on Linux in-game. And that’s on really, really expensive hardware.

Now, this is native port, so one could expect higher performance than when you do a port with a wrapper like eON right ? Well… we actually have a comparison available, once again from the same Ubuntu User, on Bioshock Infinite with a GTX 680. Check it out:

There is no precise summary just like in Shadow of Mordor, but you can see that at Max settings the framerate on Ubuntu is about 50% of the Windows rate. So is a native port better ? Yeah, probably A LITTLE BIT. But we have been waiting for a while and if it takes that much effort just to squeeze a couple more percents for a native port vs a wrapper job, I’m wondering if that’s really the right way to go, since Feral and Aspyr are going to be constrained in terms of resources (they can’t handle probably more than a dozen titles a year each).

It will be interesting to see how well Shadow of Mordor will run on WINE when DX11 is implemented around the end of the year.

More importantly, apart from an ultra-specific case where Valve managed to get more FPS on L4D2 on the Linux version vs Windows (now one has to wonder if it was more of PR stunt than anything else), there is still no demonstrated way to get nearly as much performance on Linux vs Windows. The previous ports from Aspyr showed the same issue as well.

While the Linux games catalogue continues to grow, it will be hard for anyone to recommend the Steam Machines as console or Windows PC replacements if games perform so badly. Vulkan may improve things in the future, but until then, whether it’s native or wrapped, we seem to be condemned to get way less juice out of our hardware on Linux.


At BoilingSteam, we strongly dislike ads and that is why you won't find any during your visit. If you like what we do, please consider supporting us on Patreon. You can also sign up to our newsletter (No Spam!), or our RSS feed or even Twitter.

Related Posts

Ekianjo

12 Comments

  1. Thanks for the well done benchmark tests. This is really, really depressing to read, I would’ve expected newly native games to perform on par with Windows. Especially running on new high-end graphic cards. Do you know what the main cause of the low performance is? Is it the drivers, optimization of the game, OpenGL vs. DirectX, etc?

    • Note that I did not perform the benchmarks – I simply linked the ones that were publicly available.
      The cause of low performance? Just my guess, but I would say:
      – OpenGL drivers are not as good as DirectX drivers
      – It is known to be more difficult to manage multiple threads in OpenGL vs DirectX.
      – High Profile DirectX games are usually developed together with nVidia to get the most performance out of DirectX.
      – Finally: porters. Feral and Aspyr may be very good, that does not mean they necessarily know how to get the best out of OpenGL yet.

      • 50-60% of Direct3D version framerate is true for most (if not all) modern ports. If OpenGL shader performance is 50% lower than DX one, then nothing can help you except adding stronger GPU. That’s why gap is lower with GTX 980.

        Wrapper or not, in-house or external porter, multi-threading, all these things are not important when GPU shader performance (a problem known since S4L release) is a clear bottleneck.

        • Thanks for your comment. Do you have a source/rationale for the 50% slower shader performance figure you mention ? What is actually causing this issue ?

          • http://steamcommunity.com/app/41070/discussions/0/828927580787451324/?l=polish#c846942156137491894
            “There’s this strange situation with OpenGL drivers ATM, especially on Linux. Strictly speaking, latest NVidia driver itself is slightly faster in OpenGL/Linux, than in Windows/D3D. However, seems that shader performance when running with it is much lower. This is the 50% speed difference people are talking about. Shader performance is largely a factor of shader compiler embedded in the driver. NVidia is working on this, and if they make it as fast as on d3d, cards like yours will probably benefit the most, due to the aforementioned bandwidth issue. “

          • Thanks Kwahoo. I realized the post you linked dates back to 2013… so I am not sure if “NVidia is (still) working on this” since there were no real improvements in OpenGL vs DirectX performance in the past year.

  2. The CPU is also an important factor in framerate differences since OpenGL is single-threaded and games tend to use one core to 100% (I haven’t checked if it is the case for Mordor).

    In this case of course the faster CPU in the second benchmark cannot explain the boost from 30 to 70 FPS :).

  3. Yeh, I have no faith in seeing equal or close performances to Windows with most ports. In the long run, it will take developers targeting Vulkan to see big name next gen games on Linux.

Leave a Reply

Your email address will not be published. Required fields are marked *