diff --git a/src/amd/common/ac_nir.c b/src/amd/common/ac_nir.c
index 5814175202c..aa71b1a00d2 100644
--- a/src/amd/common/ac_nir.c
+++ b/src/amd/common/ac_nir.c
@@ -401,6 +401,9 @@ lower_intrinsic_to_arg(nir_builder *b, nir_instr *instr, void *state)
                                      nir_imul_imm(b, load_subgroup_id_lowered(s, b), s->wave_size));
       }
       break;
+   case nir_intrinsic_load_subgroup_invocation:
+      replacement = nir_mbcnt_amd(b, nir_imm_intN_t(b, ~0ull, s->wave_size), nir_imm_int(b, 0));
+      break;
    default:
       return false;
    }
diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 56d55e16a00..e4a18d24227 100644
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -8139,10 +8139,6 @@ visit_intrinsic(isel_context* ctx, nir_intrinsic_instr* instr)
       break;
    }
 
-   case nir_intrinsic_load_subgroup_invocation: {
-      emit_mbcnt(ctx, get_ssa_temp(ctx, &instr->def));
-      break;
-   }
    case nir_intrinsic_ballot_relaxed:
    case nir_intrinsic_ballot: {
       Temp src = get_ssa_temp(ctx, instr->src[0].ssa);
diff --git a/src/amd/compiler/aco_instruction_selection_setup.cpp b/src/amd/compiler/aco_instruction_selection_setup.cpp
index 14eb1448ac4..9ded11da086 100644
--- a/src/amd/compiler/aco_instruction_selection_setup.cpp
+++ b/src/amd/compiler/aco_instruction_selection_setup.cpp
@@ -550,7 +550,6 @@ init_context(isel_context* ctx, nir_shader* shader)
                case nir_intrinsic_load_per_vertex_input:
                case nir_intrinsic_load_per_vertex_output:
                case nir_intrinsic_load_interpolated_input:
-               case nir_intrinsic_load_subgroup_invocation:
                case nir_intrinsic_load_tess_coord:
                case nir_intrinsic_write_invocation_amd:
                case nir_intrinsic_mbcnt_amd:
diff --git a/src/amd/compiler/tests/test_isel.cpp b/src/amd/compiler/tests/test_isel.cpp
index f8b8361e813..c6514a01fee 100644
--- a/src/amd/compiler/tests/test_isel.cpp
+++ b/src/amd/compiler/tests/test_isel.cpp
@@ -1671,9 +1671,9 @@ BEGIN_TEST(isel.cf.empty_exec.repair_ssa)
             //>> BB6
             //! /* logical preds: BB5, / linear preds: BB5, / kind: uniform, */
             //>> s1: %sgpr0 = p_parallelcopy 42
-            //>> v1: %vgpr0 = v_mbcnt_hi_u32_b32_e64 -1, %_
+            //>> v1: %vgpr0 = v_mbcnt_hi_u32_b32_e64 %_, %_
             val_sgpr = nir_imm_int(nb, 42);
-            val_vgpr = nir_load_subgroup_invocation(nb);
+            val_vgpr = nir_mbcnt_amd(nb, nir_imm_intN_t(nb, UINT64_MAX, 64), nir_imm_int(nb, 0));
 
             //>> BB7
             //! /* logical preds: BB6, / linear preds: BB6, / kind: uniform, break, */
diff --git a/src/amd/llvm/ac_nir_to_llvm.c b/src/amd/llvm/ac_nir_to_llvm.c
index 570025121b9..e58183a1609 100644
--- a/src/amd/llvm/ac_nir_to_llvm.c
+++ b/src/amd/llvm/ac_nir_to_llvm.c
@@ -2774,9 +2774,6 @@ static bool visit_intrinsic(struct ac_nir_context *ctx, nir_intrinsic_instr *ins
    case nir_intrinsic_as_uniform:
       result = ac_build_readlane(&ctx->ac, get_src(ctx, instr->src[0]), NULL);
       break;
-   case nir_intrinsic_load_subgroup_invocation:
-      result = ac_get_thread_id(&ctx->ac);
-      break;
    case nir_intrinsic_load_workgroup_id: {
       LLVMValueRef values[3] = {ctx->ac.i32_0, ctx->ac.i32_0, ctx->ac.i32_0};