IL-2 looks for certain GL extension to determine the GPU. My guess is the new driver dropped support for some older extensions. What might work is faking the presence of these extensions. IL2GE could do this, but there might be tools that already do this
Here are the extensions present in the
21.4.1 driver which are not in the
23.4.2 driver. Considering the age of IL2 I think a probable candidate extension used for checking if Perfect mode is possible on ATI/AMD hardware would be
GL_ATI_fragment_shaderGL_AMD_conservative_depth
GL_AMD_debug_output
GL_AMD_draw_buffers_blend
GL_AMD_framebuffer_sample_positions
GL_AMD_gcn_shader
GL_AMD_interleaved_elements
GL_AMD_multi_draw_indirect
GL_AMD_name_gen_delete
GL_AMD_occlusion_query_event
GL_AMD_performance_monitor
GL_AMD_query_buffer_object
GL_AMD_seamless_cubemap_per_texture
GL_AMD_shader_atomic_counter_ops
GL_AMD_shader_stencil_export
GL_AMD_shader_stencil_value_export
GL_AMD_shader_trace
GL_AMD_sparse_texture
GL_AMD_sparse_texture_pool
GL_AMD_stencil_operation_extended
GL_AMD_texture_cube_map_array
GL_AMD_texture_texture4
GL_AMD_transform_feedback3_lines_triangles
GL_AMD_transform_feedback4
GL_AMD_vertex_shader_layer
GL_AMD_vertex_shader_viewport_index
GL_AMDX_debug_output
GL_ARB_compressed_texture_pixel_storage
GL_ARB_ES3_1_compatibility
GL_ARB_imaging
GL_ARB_map_buffer_alignment
GL_ARB_provoking_vertex
GL_ARB_robust_buffer_access_behavior
GL_ARB_seamless_cubemap_per_texture
GL_ARB_shader_atomic_counter_ops
GL_ARB_shading_language_include
GL_ARB_shading_language_packing
GL_ARB_texture_buffer_object_rgb32
GL_ARB_texture_env_crossbar
GL_ARB_texture_mirror_clamp_to_edge
GL_ARB_texture_mirrored_repeat
GL_ARB_texture_query_levels
GL_ARB_texture_query_lod
GL_ARB_texture_snorm
GL_ARB_vertex_array_bgra
GL_ARB_vertex_type_10f_11f_11f_rev
GL_ARB_vertex_type_2_10_10_10_rev
GL_ATI_envmap_bumpmap
GL_ATI_fragment_shader
GL_ATI_texture_compression_3dc
GL_ATI_texture_float
GL_ATI_texture_mirror_once
GL_EXT_bindable_uniform
GL_EXT_copy_texture
GL_EXT_draw_instanced
GL_EXT_histogram
GL_EXT_packed_float
GL_EXT_rescale_normal
GL_EXT_separate_specular_color
GL_EXT_subtexture
GL_EXT_texgen_reflection
GL_EXT_texture_cube_map
GL_EXT_texture_mirror_clamp
GL_EXT_texture_swizzle
GL_EXT_vertex_array_bgra
GL_EXT_vertex_attrib_64bit
GL_IBM_texture_mirrored_repeat
GL_KHR_context_flush_control
GL_KTX_buffer_region
GL_NV_alpha_to_coverage_dither_control
GL_NV_conditional_render
GL_NV_copy_depth_to_color
GL_NV_explicit_multisample
GL_NV_half_float
GL_NV_texgen_reflection
GL_SGIS_generate_mipmap
GL_SGIS_texture_edge_clamp
GL_SGIS_texture_lod
GL_SUN_multi_draw_arrays
GL_WIN_swap_hint
It might be worth trying to fake the presence of that extension with IL2GE to see if it gets Perfect mode working again on AMD hardware with drivers newer than 22.6.1?
EDIT: Having done some more reading - while faking the presence of the right extension(s) to get Perfect mode activated on newer AMD drivers may work, if the driver doesn't re-route calls made to the removed extensions to newer ones then it will probably cause a game crash anyway. Given that one of the issues with IL2's now 20+ year old engine is that it is using direct calls to wgl methods during context creation instead of routing through the more compatible GDI wrappers then there's a fair chance that some of the old ATI shader methods are being called directly too. All these calls would then have to be intercepted and re-routed and I don't know how large an amount of work that would become.