My GPU is probably as good with my resolution as your GPU is with your 4k resolution. My fps can sometimes stay around 30-40 in cities but it rarely goes under 30.
I can run Star Citizen at a playable fps but bigger cities can be tough. Also, when it comes to single core performance threadrippers aren't the best. Even though the game does support multi-core very well, it's far from optimized and high core/thread numbers can cause issues. You CPU, while very powerful, isn't optimal for this game at the moment. Honestly I believe that your fps count is pretty much what it should be if you don't leave your starting location.
Big cities are easily the worst places when it comes to fps and these are the places where people usually start playing (Lorville, Area 18 and New Babbage). These two things mean that it's extremely heavy on your hardware.Ĥk resolution is unplayable for most. Try to remember that this game is still in alpha and it's graphically ahead of every other game that is out today. The general penalty for cross talk is less on Intel, but on AMD if it's contained it should be faster. Intel has a similar issue, they use a mesh fabric where all their cpu's talk to eachother, so the more they talk the slower everything gets.
#Process lasso and threadripper windows
Manually assigning afinity and removing cores is a hack to work around windows inability to correctly schedule programs on CPU's.
#Process lasso and threadripper how to
Good OSes (linux) generally understand the underlying CPU architecture and will correctly assign things so they dont spill over, Windows is shit as always and technically it should have some understanding on how to correctly schedule programs to run fast. So long as the program is contained to one chiplet it will be very fast, but when the program needs to fetch data from another chiplet there will be a big performance hit. ie a group of cores on its own (currently they use 8 cores per chiplet). Hopefully this will pick up several more frames per second, until we get a better optimized engine that takes better advantage of the system resources.ĪMD uses chiplets. Memory Allocation Policy 0x00000002 WKS_MEMORY_ALLOCATION_POLICY_AGGRESSIVE_PRE_ALLOCATION Threaded Optimization: Auto ( you would think this would make a differnce. Multi-display/Mixed-GPU acceleration: Single display performance mode Power managment Mode: Prefer maximum performance (or test the other 2 Perferred_Pstate modes to see iff it works better) If your streaming CUDA force P2 State: (this thread may be worth your while ) Maximum pre rendered frames: 2 (this seemed to give me a boost)Īmbient occlusion setting: off (should be by default this is a frame killer) ambient occlusion usage: Disabled Then use nvidiaProfileInspector.exe and tweak your squadron42 profile ( ) my choices are none, channel, die, socket, and auto, out of all those the setting 'none' worked the best. (hpet is big barrel of confusion so dont listen to others recommendation, you will need to test and find what combination of settings give you better frames)Īlso the numa setting, for me i set whats called "memory interleave" options in the bios to none and this also helped with frames. me i found i had to disable hpet and it performed better. Take time to investigate what hpet settings work for you. (highly recommend process lasso as this gives you more options to tweak and no need to run both programs from above, ) use this to set the cpu affinity to best performing numa node 1 or 2for starcitizen.exe. Yeah for some reason SC does not like to work well on the first gen threadrippers here is what i have found out over the years with mine
0 kommentar(er)
