diff --git a/tests/unit/test_fmha_hd64_smem_p.cu b/tests/unit/test_fmha_hd64_smem_p.cu
index 6ac8fa07..735c2179 100644
--- a/tests/unit/test_fmha_hd64_smem_p.cu
+++ b/tests/unit/test_fmha_hd64_smem_p.cu
@@ -189,37 +189,6 @@ test_fmha_hd64_smem_p(const bf16_t* q, const bf16_t* k, const bf16_t* v,
         }
     }
 
-    // ===== Register-math PV for verification =====
-    float* s_ref_o = s_p_vals;  // Reuse s_p_vals (already consumed by PV)
-    if (tid == 0) {
-        float s_ref[SK];
-        for (int j=0;j<SK;j++) s_ref[j] = s_p_vals[j];
-        for (int d=0;d<HD;d++) {
-            float ov = 0.0f;
-            for (int j=0;j<SK;j++) ov += s_ref[j] * bf16_to_f32(v[d*SK+j]);
-            s_ref_o[d] = ov;
-        }
-    }
-    __syncthreads();
-    if (tid == 0) { printf("RegPV[0..7]: "); for(int d=0;d<8;d++) printf("%.6f ", s_ref_o[d]); printf("\n"); }
-    __syncthreads();
-
-    // Verify V SMEM values for kt=0
-    if (tid == 0) {
-        bf16_t* sv0 = sV;  // kt=0, nt=0
-        // Read a few positions and compare with v[]
-        for (int d = 0; d < 4; d++) {
-            for (int r = 0; r < 4; r++) {
-                int g_mn = d / 8, g_k = r / 8;
-                int llr = d % 8, lc = r % 8;
-                int idx = g_k * 8 * 64 + g_mn * 64 + llr * 8 + lc;
-                printf("V_smem[%d,%d]=%.4f vs V_gmem=%.4f\n", d, r,
-                    bf16_to_f32(sv0[idx]), bf16_to_f32(v[d * SK + r]));
-            }
-        }
-    }
-    __syncthreads();
-
     if (wid == 0) tmem_dealloc(tb, 128);
 }