so i used AI to get proper values for fps boost config [Deadlock FPS Tweak Config]

hatsunemiekuah

New member
Not sure if this is the right place to post here. but worth noting considering people have complained to have fps drops in OG New Blood update

so i found a config from gamebanana but it's somewhat outdated. i grab the codes and put it to grok, telling it to adjust the config values based on my laptop (intel core i5, 12 GB RAM, Nvidia Geforce 940MX 2GB VRAM). after several prompts and testings in street brawl and bot matches, i found the most ideal config values that i put in gameinfo.gi below Convars. here is the example that i used. with this config, my teamfight experience in street brawl and normal match has less FPS drops, improving from 20 to 40 fps average. 30 fps at worst

Thanks for coming by at my post. if this post somehow violates the rules, i will delete it when i notice your warning.

Code:
    ConVars
    {   
"gpu_level" "0"                     // Minimum Shader Details Level
"cpu_level" "0"                     // Minimum Effect Details Level
"mat_set_shader_quality" "0"
"fps_max" "60"                      // Cap for stability

// ================ FOV ================
"r_aspectratio" "2.15"              // Keep your 90fov vibe

// ================ LIGHTING & SHADOWS ================
"r_directlighting" "1"
"r_ssao" "0"
"r_shadows" "0"
"r_rendersun" "1"                   // Reverted - no more breaks
"r_cheapshadows" "1"                // Added for cheap shadows
"lb_enable_shadow_casting" "0"
"lb_csm_draw_alpha_tested" "0"
"lb_csm_draw_translucent" "0"
"lb_barnlight_shadowmap_scale" "0.4"  // Tuned milder
"lb_csm_cascade_size_override" "1"
"lb_dynamic_shadow_resolution_quantization" "48"  // Balanced
"lb_csm_override_staticgeo_cascades_value" "0"
"lb_csm_receiver_plane_depth_bias" "0.00002"
"lb_csm_receiver_plane_depth_bias_transmissive_backface" "0.0002"
"lb_sun_csm_size_cull_threshold_texels" "25"     // Tuned
"lb_dynamic_shadow_resolution_base" "192"        // Balanced low-res shadows
"lb_enable_fog_mixed_shadows" "0"
"r_citadel_sun_shadow_slope_scale_depth_bias" "1.5"  // Milder bias
"r_lightmap_size_directional_irradiance" "3"     // Low but stable
"r_size_cull_threshold_shadow" "150"             // Aggressive but safe
"csm_max_shadow_dist_override" "0"
"sc_cache_envmap_lpv_lookup" "1"
"cl_retire_low_priority_lights" "1"
"r_flashlightshadows" "0"           // Added safe shadow disable
"r_screen_space_shadows" "0"        // Added

// ================ FOG & ATMOSPHERE ================
"r_enable_volume_fog" "0"
"r_enable_gradient_fog" "0"
"r_enable_cubemap_fog" "0"
"volume_fog_intermediate_textures_hdr" "0"
"r_citadel_fog_quality" "0"         // Added

// ================ SKY & ENVIRONMENT ================
"r_draw3dskybox" "0"
"r_drawskybox" "1"                  // Reverted - fixes sky/HUD weirdness
"r_monitor_3dskybox" "0"
"r_world_wind_strength" "0"

// ================ SSAO ================
"r_citadel_ssao_bent_normals" "false"
"r_citadel_ssao_denoise_passes" "0"
"r_citadel_ssao_quality" "0"
"r_citadel_ssao_radius" "0"
"r_citadel_ssao_thin_occluder_compensation" "0"
"r_citadel_distancefield_ao_quality" "0"  // Added

// ================ PARTICLE SYSTEM ================
"r_particle_max_detail_level" "0"
"r_particle_cables_cast_shadows" "0"
"r_RainParticleDensity" "0"
"r_physics_particle_op_spawn_scale" "0"
"r_particle_min_timestep" "0"
"r_particle_max_size_cull" "600"               // Loosened for stability
"particle_cluster_nodraw" "1"
"r_particle_mixed_resolution_viewstart" "600"  // Balanced
"cl_particle_sim_fallback_threshold_ms" "0"
"cl_parallel_readpacketentities_threshold" "3"  // Milder
"sc_layer_batch_threshold_fullsort" "60"       // Balanced batches
"r_particle_gpu_implicit_lds_cache" "1"
"r_particle_model_new8" "0"
"cl_show_splashes" "0"
"r_drawparticles" "1"                // Added - keeps essentials
"r_particle_depth_feathering" "0"    // Added

// ================ PHYSICS & CLOTH ================
"cloth_update" "1"                   // Back ON as requested
"cloth_sim_on_tick" "0"
"presettle_cloth_iterations" "2"     // Tuned down mildly
"pred_cloth_pos_max" "0.5"           // Tuned for perf without breaking
"pred_cloth_pos_multiplier" "0.15"
"pred_cloth_pos_strength" "0.05"
"pred_cloth_rot_high" "0.025"
"pred_cloth_rot_low" "0.0025"
"pred_cloth_rot_multiplier" "0.1"
"cl_phys_timescale" "1"              // Back ON
"cl_fasttempentcollision" "30"       // Balanced

// ================ MODEL & DECAL OPTIMIZATIONS ================
"r_drawmodeldecals" "0"
"r_character_decal_resolution" "1"
"r_decals" "8"                       // Milder than 0
"r_propsmaxdist" "600"               // Balanced fade
"props_break_max_pieces_perframe" "0"
"r_citadel_screenspace_particles_full_res" "0"
"r_citadel_gpu_culling_shadows" "1"
"skeleton_instance_lod_optimization" "1"
"r_size_cull_threshold" "2.0"        // Balanced aggressive cull
"r_hair_ao" "0"
"ik_final_fixup_enable" "0"
"ik_fabrik_align_chain" "0"
"cl_impacteffects" "0"
"enable_boneflex" "false"
"r_drawdetailprops" "1"              // Added

// ================ LOD ================
"r_lod" "3"                          // Added MAX LOD reduction
"sc_instanced_mesh_lod_bias" "7"     // Pumped but not max crazy
"sc_instanced_mesh_lod_bias_shadow" "7.5"
"sc_instanced_mesh_motion_vectors" "0"
"sc_instanced_mesh_size_cull_bias" "7"
"sc_instanced_mesh_size_cull_bias_shadow" "7.5"
"r_citadel_antialiasing" "0"         // Added
"r_citadel_motion_blur" "0"          // Added

// ================ ROPE PHYSICS ================
"rope_collide" "0"
"rope_subdiv" "0"
"rope_wind_dist" "0"
"rope_smooth_enlarge" "0"
"rope_smooth_maxalpha" "0"
"rope_smooth_maxalphawidth" "0"
"rope_smooth_minalpha" "0"
"rope_smooth_minwidth" "0"
"r_ropetranslucent" "0"
"r_drawropes" "0"

// ================ TERRAIN & FOLIAGE ================
"r_grass_quality" "0"
"r_grass_start_fade" "0"
"r_grass_end_fade" "0"

// ================ UI & HUD ================ (FULL REVERT - NO BREAKS)
"panorama_disable_box_shadow" "0"
"r_dashboard_render_quality" "1"
"closecaption" "false"
"citadel_hud_objective_health_enabled" "2"
"citadel_boss_glow_disabled" "1"
"citadel_damage_offscreen_indicator_disabled" "0"
"citadel_portrait_world_renderer_off" "false"
"panorama_use_new_occlusion_invalidation" "1"
"panorama_temp_comp_layer_min_dimension" "128"
"panorama_max_overlay_fps" "30"
"panorama_max_fps" "60"
"citadel_damage_text_lifetime" "1.5"
"citadel_damage_text_lifetime_new" "0.75"
"citadel_damage_text_lifetime_accumulated_new" "2"
"citadel_show_new_damage_feedback_numbers" "1"
"hud_free_cursor" "0"

// ================ NETWORK & PREDICTION ================
"cl_prediction_savedata_postentitypacketreceived" "1"
"r_frame_sync_enable" "0"
"cl_clock_buffer_ticks" "0"
"cl_async_usercmd_send" "true"
"cl_smooth" "0"
"cl_smoothtime" "0.01"
"cl_smooth_draw_debug" "0"

// ================ AI OPTIMIZATIONS ================
"ai_strong_optimizations_no_checkstand" "1"
"ai_expression_optimization" "1"

// ================ AUDIO ================
"audio_enable_vmix_mastering" "0"
"dsp_volume" "0"
"dsp_off" "0"                        // Reverted

// ================ MISC ================
"sc_hdr_enabled_override" "0"
"r_texturefilteringquality" "0"
"cl_ragdoll_limit" "0"
"r_pipeline_stats_present_flush" "0"
"r_pipeline_stats_command_flush" "0"
"r_max_portal_render_targets" "2"    // Reverted
"citadel_hideout_ball_show_juggle_count" "1"  // Reverted
"citadel_hideout_ball_show_juggle_fx" "1"     // Reverted
"citadel_camera_soft_collision" "0"
"cl_input_enable_raw_keyboard" "1"
"mat_colcorrection_disableentities" "0"  // Reverted
"mm_idle_show_warning_at_s" "999"
"r_effects_bloom" "0"                // Added
"r_post_bloom" "0"                   // Added
"r_arealights" "0"                   // Added
"r_citadel_distancefield_reflections" "0"  // Added
"r_citadel_distancefield_shadows" "0"     // Added
"r_shadow_distance" "0"              // Added
 
I hope you know that the Ai is just choosing values practically randomly. There's no dataset of how these values effect performance, and grok can't test anything as an LLM.
The "keep your 90 fov vibe" line is particularly worrisome.

If it works for you that's great, but it's most likely because all the values from the original config are changing the performance metrics and grok is just tweaking random numbers by small amounts.
 
I hope you know that the Ai is just choosing values practically randomly. There's no dataset of how these values effect performance, and grok can't test anything as an LLM.
The "keep your 90 fov vibe" line is particularly worrisome.

If it works for you that's great, but it's most likely because all the values from the original config are changing the performance metrics and grok is just tweaking random numbers by small amounts.
there is sometimes a mishap when the culling values went too aggressive. i ordered the ai to fix the values according to the testing results. basically, i do the testing, and grok receive the testing feedback and suggest the values accordingly.

i also use the thinks hard mode to ensure proper values adjusted based on my specs
 
there is sometimes a mishap when the culling values went too aggressive. i ordered the ai to fix the values according to the testing results.

i also use the thinks hard mode to ensure proper values adjusted based on my specs
It can think as hard as it wants (use more tokens) It doesn't know what any of the values do. It just guesses.
 
It can think as hard as it wants (use more tokens) It doesn't know what any of the values do. It just guesses.
Nobody knows what those values do, every programmer is just a monkey randomly mashing buttons, thats why you need 3000 of them to make a new Ubishit game via data inputs randomly correlating into a pattern that works.

(this might be a joke post)
 
there is sometimes a mishap when the culling values went too aggressive. i ordered the ai to fix the values according to the testing results.

i also use the thinks hard mode to ensure proper values adjusted based on my specs

Nobody knows what those values do, every programmer is just a monkey randomly mashing buttons, thats why you need 3000 of them to make a new Ubishit game via data inputs randomly correlating into a pattern that works.

(this might be a joke post)
jokes aside, successful codes are built through trials and errors
 
Back
Top