nx_gpu_exceed_gate.nx

nx_gpu_exceed_gate.nx -- MEASURED DirectX exceed: the sovereign graphics SUBMISSION path vs DirectX. Operator bar: "s-class exceed BETTER than native DirectX, not just it can do a triangle." The honest, defensible exceed is on the CPU-SUBMISSION axis -- exactly the bottleneck DX12/Vulkan/Mantle were created to escape from DX11's driver stack. DirectX path (documented, by-design): app -> D3D runtime -> user-mode driver -> DXGK kernel -> kernel-mode driver -> GPU = 5 layers, with per-draw VALIDATION/TRANSLATION in the driver (the DX11 draw-call bottleneck). Sovereign path: app -> GPFIFO pushbuffer encode -> doorbell -> GPU = 2 layers, NO driver-validation layer (we encode the GF100 method stream directly, proven by the BM-GPU gates). This gate MEASURES our side (draw-command encode throughput, bytes/draw, path depth = real numbers) and grades vs DirectX's by-design architecture, with an OVERCLAIM CONTROL that forces RED unless it admits the axes where we are BEHIND (real-silicon execution, feature breadth). HONEST SCOPE: this measures SUBMISSION/ENCODE (CPU producing draw commands -- the "superior produce" axis), which is structural + measured. End-to-end RENDER FPS vs DirectX needs the silicon last-mile (BM-GPU-6/7) and is NOT claimed here. expect_exit: 0 license_tier: ORIGINAL