Rise of the Tomb Raider video comparison & benchmark on Linux & Windows
95Please, please, pretty please read the description for details not present in the video. Press SHOW MORE.
Keep in mind that the card used here is below game's requirements, but is the newest card I have now (sadly my Nvidia died peacefully in its sleep).
I expect also that Nvidia results to be somewhat better, considering that RADV still has some catch up to do.
For more benchmarks and info about the Linux version of this game check these sites:
https://store.feralinteractive.com/en/mac-linux-games/riseofthetombraider/
https://www.gamingonlinux.com
https://boilingsteam.com
https://www.phoronix.com/scan.php?page=home
Usually I make my benchmarks on SteamOS, but since SteamOS has an older Mesa version (17.2.4), I used here Ubuntu with Padoka PPA and a newer kernel version.
That being said, even with the older SteamOS's Mesa the game still runs with (roughly) just 3 FPS lower. For some newer cards though the game still needs at least Mesa 18.
My system:
- i7-4790K with Hyper-threading disabled.
- AMD HD 7970 (R9 280X) with 3 GB VRAM. Note: AMDGPU/RADV is not enabled by default on this older card, you'll have to do it manually.
- 16 GB RAM
- Windows 10 Pro, Radeon Software Adrenalin Edition 18.3.4
- Ubuntu 16.04 LTS, Kernel 4.16.2, Mesa 18.1.0
The game is Vulkan-only on Linux and DX11-DX12 on Windows.
Videos recorded with an external Blackmagic Intensity Pro 4K card and Media Express recording software (no game performance loss).
- All tests were done after several dry runs/restarts and allowing time for shaders to be compiled first.
- All tests at 1920x1080 resolution, performance mode enabled for the CPU.
- The benchmark data was collected with the Steam overlay disabled. I found few times in the past that the overlay affects Vulkan performance, and even here it drops the perf just a lower bit.
- The video comparison was recorded with the game running at Medium settings. In this case I had the overlay/FPS counter enabled.
- The test were done using the default settings as much as possible, so no extra AA settings, only FXAA, also:
- Ambient Oclusion set to ON since is the only option on Linux.
- Lens Flares removed because this ain't freaking Star Trek and J.J Abrams has no saying in my house.
- The "Very High" settings on Linux doesn't use Very High Textures (needs more than 4 GB VRAM) as the Windows version does. Nevertheless I set the Textures to Very High on Linux too, so keep that in mind when looking at the graph results (my card has only 3 GB VRAM).
- By default the game benchmark is divided in three sections with their own results, however in order to simplify the graphs I combined them together in a single section.
- Usage data is collected at one second intervals and written directly to RAM in order to cause no drive activity.
Some possible questions and answers:
- Why not all relevant (system/settings) info in the video? I used to put them in the video, but people seems to ignore them. Is also easier here to edit the info (if I make mistakes) or add some more later.
- Why choose Medium settings? Because Medium is the sweet spot for my older card.
- Why not Min/Max bars/values? Because they're pretty much useless if you want to see how performance varies over time, there are some graphs for that instead, or check the "Percentiles" results.
-Why redundant info in the graphs? Why two kinds of graphs for similar kind of data? Some people prefer to deal only with basic info such as average FPS or average frametimes, while others prefer more advanced info such as percentiles or want to see graphs over time, choose the ones you're comfortable with and ignore the others.
- You say "Total RAM" usage, what does it mean? It means that this is the total system RAM usage, game+everything else. I used to collect only the game's itself RAM usage, but I often found that some games are adding extra processes with their own RAM usage, and this causing an actual much higher RAM usage when running the respective game. I'm not satisfied with any of these two approaches though, if you have any ideas let me know.
- About the GPU usage graph: It may look like the data was not collected in the same manner or at the same time intervals, but they are actually identical. If you look closely you'll see that the gaps in usage between the three benchmark sections are coinciding. Is just that the GPU in the Linux version is already at full throttle when benchmark starts, while in Windows is more relaxed. Is not a one time occurrence, is always like that.
- I occasionally see some weird colors in the darker areas in the video. Yes, my recording software (Media Express) does that sometimes, while OBS doesn't. I prefer though Media Express since it records at higher quality. Is a trade off.
- You forgot the specify this and that. Possibly, what?
- You made a mistake/typo/etc...damn it, where?
I also uploaded this video on: https://archive.org/details/RiseOfTheTombRaiderBenchmarkOnLinuxWindows_20180421
Other Videos By airspeedmph
Other Statistics
Rise of the Tomb Raider Statistics For airspeedmph
airspeedmph presently has 7,380 views for Rise of the Tomb Raider across 2 videos, with his channel publishing less than an hour of Rise of the Tomb Raider content. This makes up 6.66% of the content that airspeedmph has uploaded to YouTube.