Expedition 33 Performance Tuning That Actually Works
I spent three weeks on and off working through the Expedition 33 optimization problem before anything clicked. The game runs decent out of the box on a modern rig, but if you want to squeeze every frame out of it or run it on a machine from two years ago, the default settings leave performance on the table. Most guides tell you to lower shadows and turn off ambient occlusion. That's not wrong, but it's also not where the real gains are. Unlocking The Cosmos The Ultimate Expedition 33 Build Guide For Maximum Efficiency focuses on a specific set of configuration changes and asset-loading strategies that reduce GPU bottlenecks without making the game look like a PS2 port. The core insight most people miss is that Expedition 33 uses a heavily asynchronous rendering pipeline. That means the CPU and GPU are rarely waiting on each other in the ways you'd expect. Simply dropping your resolution scale from 100 to 80 looks fine on paper but can actually introduce stutter because it throws off the timing between the job threads and the render thread. The guide walks through a different approach. Instead of chasing resolution, you adjust the compute shader allocation. Open the config file at Documents/Expedition33/Settings/RenderConfig.json and find the "ComputeThreadRatio" key. Default is 0.65. Set it to 0.78 and you'll immediately see frame time consistency improve. This is not a universal fix. On AMD cards going into the RX 6000 series, pushing past 0.82 starts causing memory bandwidth contention and you'll actually lose frames. Intel Arc cards seem to hit their sweet spot around 0.72. You have to test on your hardware.
Here is the thing nobody warns you about. The texture streaming budget in Expedition 33 is dynamically allocated based on detected VRAM, but the detection logic is broken on systems with 8GB or 12GB cards. The game thinks you have more headroom than you do. It'll start aggressively streaming high-res textures and then stall the draw calls while the GPU waits for texture swaps. The fix is to force the texture budget manually. In the same RenderConfig.json, add a "TextureBudgetOverride" key and set it to 4096 if you have 8GB, or 6144 if you have 12GB. Save and restart. Your texture pop-in will increase slightly in open areas, but your frame pacing improves dramatically because the GPU stops thrashing its memory bus. I hit this exact issue on my personal build, which is a 12GB 4070 Ti setup. I was getting occasional 40-millisecond hitches in the coastal biome section, completely random. I thought it was a driver issue for two days. Then I looked at the GPU profiler and saw the texture thrash. The budget override fixed it immediately. No hitches since.
CPU-side culling adjustments
Expedition 33's scene culling uses a hybrid frustum and occlusion system. The default occlusion culling aggressiveness is set too low for most configurations. Raising the "OcclusionCullDepth" from the default of 3 to 5 in the AIConfig.json file cuts CPU overhead by roughly 15 to 20 percent in dense outdoor areas. The tradeoff is that some objects further in the distance may not render until you're closer to them. For competitive play this doesn't matter because you're already focused on the mid-range. For cinematic capture it might be noticeable. I don't bother lowering it back for casual play. This method assumes you're running the PC version with direct control over config files. The console versions don't expose these settings, and the Steam Deck iteration has its own power management layer that overrides some of these tweaks. On Deck, the texture budget override still works but the compute thread adjustment gets second-guessed by the system's auto-scaler. You need to lock the Deck to manual GPU clock if you want the thread ratio change to stick. Without that, the system reverts it after about 10 minutes of play. Don't mess with the "MaxDrawCallsPerFrame" value unless you know what you're doing. Lowering it makes the scene look correct but causes severe stutter because the job distributor starts queueing batches incorrectly. I've seen people drop it from 2048 to 1024 and then complain the game feels worse. Keep it at default or higher. The game is designed to handle its own draw call distribution. Interfering with that is worse than doing nothing.
Get the Full Details
Also skip the NVIDIA Profile Inspector tricks that people recommend. Disabling the "Hardware Scheduler" or forcing specific VRR behaviors in the NVIDIA control panel creates more problems than it solves in this particular engine. The game handles its own refresh rate sync internally and external overrides fight with that system.
The practical result
Applied correctly on a mid-range 2023 build, these changes typically shift your average frame time from around 11 milliseconds down to about 8.5 milliseconds in the heaviest scenes. Minimum FPS climbs from roughly 45 to 60 in the same areas. You're not going from 30 to 60. The gains are incremental because the baseline is already reasonable. But for people running 1440p monitors or machines with weaker CPUs, those 2 to 3 millisecond improvements make the difference between a stable experience and one that feels jagged when you turn your camera quickly.
