This file has a few corrections to some of the improper CVars set in Comanglia's config. These corrections won't be maintained very often, nor are they complete. It's recommended to use mastercomfig instead.
// Do not use dxlevel 80 or 81, it is outdated and buggy and dx9 uses your GPU more effectively, is more optimized and has many performance improvements done that were not ported to dx8 //fps_max 0 // fps caps on PCs that consistently meet that cap causes horrible input lag when attempting to turn/aim fps_max 240 // Input lag does not happen with an FPS cap as far as I can tell, and Source is better off with a frame cap of 300 due to some systems assuming FPS will not be higher than 300 //cl_cmdrate 67 // The tricks to get 1.4ms lerp are placebo you're still forced to .0152 cl_cmdrate 66 // Yes, this is entirely placebo. It is a trick that doesn't work because cmdrate is limited to 66 //cl_interp .033 // change to whatever you like for certain classes projectiles/popping uber/flames = .0152 hitscan = .031 on bad connections = .062 cl_interp 0 // By setting this higher than cl_interp_ratio / cl_updaterate, you are slightly increasing your interp for no reason. Also, interp is mostly preference at lower values and depends on how stable your connection is. //cl_smooth 0 cl_smooth 1 // This only smooths your view to prevent it jumping around, it is personal preference, but most people will like it to be on //cl_smoothtime 0.01 cl_smoothtime 0.08 // No reason to change this when cl_smooth is 0, and it is best at 0.08 when enabled //cl_updaterate 67 // The tricks to get 1.4ms lerp are placebo you're still forced to .0152 cl_updaterate 66 // Yes, this is entirely placebo. It is a trick that doesn't work because updaterate is limited to 66 //rate 60000 // anything above 62,000 is completely useless, anything above 48,000ish is useless in HL rate 84480 // Rate is not useless above 62,000 (in fact, the default is 80,000), it affects how many packets you can send out and increasing it will decrease artificial choke caused by rate limiting within the Source engine //r_lod 2 r_lod -1 // Setting r_lod 2 will cause all models to be at their second LOD level and will not let them go down to their third or lower quality LOD levels when they are small enough to do so. -1 will use r_rootlod to determine how much LOD should be offset by, so that models can properly decrease quality even more. //g_ragdoll_fadespeed 0 g_ragdoll_fadespeed 10000 // This is the rate at which ragdolls fade, setting it to 0 means that it will fade by 0 every frame and thus causes a memory leak //g_ragdoll_lvfadespeed 0 g_ragdoll_lvfadespeed 10000 // Same thing as g_ragdoll_fadespeed, but this is when low violence mode is enabled //violence_agibs 0 violence_agibs 1 // Do not set any of the violence CVars to 0, or else low violence mode will be enabled! Low violence mode reduces FPS //violence_hgibs 0 violence_hgibs 1 // Do not set any of the violence CVars to 0, or else low violence mode will be enabled! Low violence mode reduces FPS //lod_transitiondist 0 lod_transitiondist -1 // 0 causes fade LODs to fade across 0 units, which causes an error. It is best to set this to -1, which lowers the distance fade LODs can be seen and also fades them across 1 unit //mat_mipmaptextures 0 // *** mat_mipmaptextures 1 // Don't disable mipmapping on textures, it is a performance decrease for all textures in a scene, especially ones further away //mat_use_compressed_hdr_textures 0 mat_use_compressed_hdr_textures 1 // Don't disable compressed HDR textures, they help you when HDR is on and don't do anything if it's off //mod_forcedata 1 mod_forcedata 0 // Enabling this causes models to not be loaded async //mat_max_worldmesh_vertices 512 mat_max_worldmesh_vertices 1024 // The lowest Source will do is 1024 and it is only useful to reduce mesh complexity on weak GPUs, since it also increases the number of meshes used for the world, which is a bad trade off on stronger cards. //r_decal_cullsize 15 r_decal_cullsize 20 // You can increase the cullsize even more //r_staticprop_lod 4 r_staticprop_lod 63 // You can decrease the LOD even more //r_lightcache_zbuffercache 0 r_lightcache_zbuffercache 1 // Use zcache from map data //r_PhysPropStaticLighting 0 r_PhysPropStaticLighting 1 // Use static lighting for props instead of baking lighting data for them on the fly //rope_averagelight 0 rope_averagelight 1 // When 0, this will do an extra calculation on top of the already done lighting computations to get the max intensity of the lighting //r_worldlightmin 0.0001 r_worldlightmin 0.0002 // Decreasing this value makes you render more insignificant lighting! 0.0002 is the default, why lower it to 0.0001 to do more lighting work? //in_usekeyboardsampletime 0 in_usekeyboardsampletime 1 // This does not delay keyboard input at all, only makes keyboard look input more reliable //r_fastzreject -1 r_fastzreject 1 // Auto detect is broken, just use 1. Also, the benefits of this are not clear on modern GPUs. //cl_threaded_client_leaf_system 0 // DO NOT ENABLE cause crashes to desktop within 4-5 minutes of gameplay 4/18 2014 cl_threaded_client_leaf_system 1 // It's fine to enable this
mat_shadowstate 1 // Command does not exist anymore mat_forcemanagedtextureintohardware 0 // There's no need to set this in the config. TF2 will determine if it is appropriate to use it or not. mem_max_heapsize 2048 // The heap isn't TF2's entire memory pool! Don't starve other memory sections just for a huge heap for no reason. The Source Engine will determine the ideal heap size. sys_minidumpspewlines 500 // You're setting it to the default value of 500. It doesn't affect FPS or FPS variance anyway, it does not become a factor in normal play as it only is used when TF2 crashes. mat_parallaxmap 0 // Command does not exist anymore sv_forcepreload 1 // A similar command, cl_forcepreload was removed in http://www.teamfortress.com/post.php?id=19733 "Fixed two most common forms of framerate stuttering / hitching when using popular FPS configs" Need I say more?