If you game on an entry level or low budget PC you are probably familiar with the anxiety of having to buy a game without knowing for sure if your PC will be able to run it in an acceptable manner on the lowest settings.
Pre-release demos seem to be getting rarer with each passing year, but Final Fantasy XV has decided to make the difference by releasing a tool that allows you to benchmark the game on a number of graphical presets, providing a score that you can use to make an informed purchase.
The benchmark can only initially run in one of the presets, but there is a way to expose individual settings and tweak the game under what is usually allowed, enabling much higher benchmark scores.
This can serve as a glimpse at what the full game will allow and provide just some plain bizarre results.
To achieve something like this you will need the ffxvBenchCustom tool by DrDaxxy. Installing it as simple as extracting its contest into the install folder of the Benchmark Tool (Usually in Program Files/Square Enix/Final Fantasy XV Benchmark).
You can create a custom config file by using one of the ini files as a base and load it into the "game" using launch commands.
To add said launch commands you will need to create a shortcut for ffxv.exe (Not ffxvbench.exe), go into the properties and add some text to the end of the "target" field.
--graphicsIni "C:\Program Files\SquareEnix\FINAL FANTASY XV BENCHMARK\ffxv_custom.ini"
This is the most important bit, and it must point to the correct location of your custom configuration file.
This will make the tool run in English.
This will make the tool run in Fullscreen
With that out of the way, let's get to changing that configuration file.
Starting for the settings in the "lite" pre-set there is still a fair amount to play around with.
Display Resolution is, as you would expect, the main resolution control of the game, but render resolution controls the internal resolution of the 3D elements. This means that you can drop it a fair bit without ever affecting UI elements or critical text, and since you are almost supersampling it back into the external resolution it will look a bit better than if you had run the whole game in that low resolution.
A fair amount of modern games are coming with internal resolution scalers, and they can be a godsend for low-end and integrated GPUs. Do not be afraid of using it.
There are two variables that control shadows.
The first is ShadowDistanceScaling which seems to control how far the shadows are rendered from the player.
Dropping it to 1 means no shadows close to the player, and a giant infinite shadow that covers everything after a certain distance.
An alternative is ShadowResolution which, you guessed it, controls the resolution of shadows. Dropping it to 0 crashes the game, but a value of 1 practically removes all shadows from the game.
Any of these two variables (or both!) can provide a substantial increase in performance.
Model Level of Detail
Modern games save resources by automatically reducing the number of details in models that are far away from the player, loading better models as the distance decreases. This is usually referred to as Level of Detail system (LOD).
Final Fantasy XV seems to limit the number of polygons on game models depending on distance for this very purpose, but the ModelLODScaling variable provides some control over this system. Setting it 1 severely limits the number of polygons that can be rendered close to the camera and this has some dramatic effects on the graphical quality of the game.
Thankfully, you do not have to go to these extremes. Just a value of 10 will cause a significant drop in quality without making the making unusable.
Let's test this completely and see how much of a difference it makes.
For testing, I combined an old, cheap and reliable first generation i5 750 quadcore at 2.66 GHz with a GT 1030, Nvidia's current entry-level GPU, and 8 GB of DDR3 ram.
I first tested on 720p and the regular "lite" settings. To my amazement, the game managed to maintain over 30 fps while moving through the world and during combat. There were some short freezes due to CPU bottlenecks while the game was loading new areas, but overall the game seems perfectly playable. Very impressive for such a large title.
The final score was 3742.
Now, let's disable shadows and reduce LOD to 10 and the game immediately climbs to somewhere between 40 and 50 with an overall score of 4633. Not bad!
Given these results, I decided to try to test on a weaker GPU. I own a Xiaomi Pro laptop with an i5-8250u CPU, Nvidia MX 150 GPU and 8 GB of RAM. I decided to temporarily disable the MX 150 and try to run the benchmark on the i5's IntelUHD 620 with all settings to the lowest, shadows disabled, LOD at 1 and internal resolution set to 640x360.
The IntelUHD is almost able to keep 20 FPS under these conditions, which is not quite playable but impressive for one of the biggest games of this year.
Only time will tell if the full game will allow this level of granularity when modifying settings in the configuration file, but these initial results are highly encouraging.