diff --git a/ggml/src/ggml-cuda.cu b/ggml/src/ggml-cuda.cu
index 5272a38bf..3e151ac7a 100644
--- a/ggml/src/ggml-cuda.cu
+++ b/ggml/src/ggml-cuda.cu
@@ -4119,6 +4119,9 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
                 if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_Q8_0) {
                     return true;
                 }
+                if (src0_type == GGML_TYPE_Q8_0 && src1_type == GGML_TYPE_F32) {
+                    return true;
+                }
                 if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_Q4_0) {
                     return true;
                 }
diff --git a/src/llama.cpp b/src/llama.cpp
index 4860cb0ea..0d8fdf421 100644
--- a/src/llama.cpp
+++ b/src/llama.cpp
@@ -8367,17 +8367,36 @@ struct llm_build_context {
             const int64_t n_head_kv = hparams.n_head_kv(il);
             const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(il);
             struct ggml_tensor * rope_factors = build_rope_factors(il);
-            struct ggml_tensor * tmp =
+            struct ggml_tensor * k =
+                ggml_view_3d(ctx0, kv_self.k_l[il],
+                    n_embd_head_k, n_head_kv, n_ctx,
+                    ggml_row_size(kv_self.k_l[il]->type, n_embd_head_k),
+                    ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa),
+                    0);
+
+            struct ggml_tensor * tmp;
+            if (ggml_is_quantized(k->type)) {
+                // dequantize to f32 -> RoPE -> quantize back
+                tmp = ggml_cast(ctx0, k, GGML_TYPE_F32);
+                cb(tmp, "K_f32", il);
+                for (auto * backend : lctx.backends) {
+                    // Figure out which backend KV cache belongs to
+                    if (ggml_backend_supports_buft(backend, lctx.model.buft_layer[il].buft)) {
+                        ggml_backend_sched_set_tensor_backend(lctx.sched, tmp, backend);
+                        break;
+                    }
+                }
+                tmp = ggml_rope_ext_inplace(ctx0, tmp,
+                        lctx.inp_K_shift, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow);
+                cb(tmp, "K_shifted_f32", il);
+                tmp = ggml_cpy(ctx0, tmp, k);
+            } else {
                 // we rotate only the first n_rot dimensions
-                ggml_rope_ext_inplace(ctx0,
-                        ggml_view_3d(ctx0, kv_self.k_l[il],
-                            n_embd_head_k, n_head_kv, n_ctx,
-                            ggml_row_size(kv_self.k_l[il]->type, n_embd_head_k),
-                            ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa),
-                            0),
+                tmp = ggml_rope_ext_inplace(ctx0, k,
                         lctx.inp_K_shift, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
                         ext_factor, attn_factor, beta_fast, beta_slow);
-
+            }
             cb(tmp, "K_shifted", il);
             ggml_build_forward_expand(gf, tmp);
         }