diff --git a/src/gallium/drivers/radeonsi/radeon_vcn_dec.c b/src/gallium/drivers/radeonsi/radeon_vcn_dec.c
index 92b256d3b51..d8a04436afa 100644
--- a/src/gallium/drivers/radeonsi/radeon_vcn_dec.c
+++ b/src/gallium/drivers/radeonsi/radeon_vcn_dec.c
@@ -2414,6 +2414,8 @@ static void radeon_dec_destroy(struct pipe_video_codec *decoder)
 
    dec->ws->fence_reference(dec->ws, &dec->prev_fence, NULL);
    dec->ws->cs_destroy(&dec->cs);
+   if (dec->ectx)
+      dec->ectx->destroy(dec->ectx);
 
    if (dec->stream_type == RDECODE_CODEC_JPEG) {
       for (i = 0; i < dec->njctx; i++) {
@@ -2762,8 +2764,14 @@ struct pipe_video_codec *radeon_create_decoder(struct pipe_context *context,
    if (!dec)
       return NULL;
 
+   if (sctx->vcn_has_ctx) {
+      dec->ectx = pipe_create_multimedia_context(context->screen);
+      if (!dec->ectx)
+         sctx->vcn_has_ctx = false;
+   }
+
    dec->base = *templ;
-   dec->base.context = context;
+   dec->base.context = (sctx->vcn_has_ctx) ? dec->ectx : context;
    dec->base.width = width;
    dec->base.height = height;
    dec->max_width = width;
@@ -2792,7 +2800,9 @@ struct pipe_video_codec *radeon_create_decoder(struct pipe_context *context,
    dec->sq.ib_total_size_in_dw = NULL;
    dec->sq.ib_checksum = NULL;
 
-   if (!ws->cs_create(&dec->cs, sctx->ctx, ring, NULL, NULL)) {
+   if (!ws->cs_create(&dec->cs,
+                      (sctx->vcn_has_ctx) ? ((struct si_context *)dec->ectx)->ctx : sctx->ctx,
+                      ring, NULL, NULL)) {
       RVID_ERR("Can't get command submission context.\n");
       goto error;
    }
@@ -3056,6 +3066,8 @@ struct pipe_video_codec *radeon_create_decoder(struct pipe_context *context,
 
 error:
    dec->ws->cs_destroy(&dec->cs);
+   if (dec->ectx)
+      dec->ectx->destroy(dec->ectx);
 
    if (dec->stream_type == RDECODE_CODEC_JPEG) {
       for (i = 0; i < dec->njctx; i++) {
diff --git a/src/gallium/drivers/radeonsi/radeon_vcn_dec.h b/src/gallium/drivers/radeonsi/radeon_vcn_dec.h
index f14df3d8189..769039ed952 100644
--- a/src/gallium/drivers/radeonsi/radeon_vcn_dec.h
+++ b/src/gallium/drivers/radeonsi/radeon_vcn_dec.h
@@ -140,6 +140,8 @@ struct radeon_decoder {
    struct pipe_fence_handle *destroy_fence;
    bool dpb_use_surf;
    uint64_t dpb_modifier;
+
+   struct pipe_context *ectx;
 };
 
 void send_cmd_dec(struct radeon_decoder *dec, struct pipe_video_buffer *target,
diff --git a/src/gallium/drivers/radeonsi/si_uvd.c b/src/gallium/drivers/radeonsi/si_uvd.c
index 3f01b6f5591..52491151d4b 100644
--- a/src/gallium/drivers/radeonsi/si_uvd.c
+++ b/src/gallium/drivers/radeonsi/si_uvd.c
@@ -97,6 +97,20 @@ static void si_vce_get_buffer(struct pipe_resource *resource, struct pb_buffer_l
       *surface = &res->surface;
 }
 
+static bool si_vcn_need_context(struct si_context *ctx)
+{
+   /* Kernel does VCN instance scheduling per context, so when we have
+    * multiple instances we should use new context to be able to utilize
+    * all of them.
+    * Another issue is with AV1, VCN 3 and VCN 4 only support AV1 on
+    * first instance. Kernel parses IBs and switches to first instance when
+    * it detects AV1, but this only works for first submitted IB in context.
+    * The CS would be rejected if we first decode/encode other codecs, kernel
+    * schedules on second instance (default) and then we try to decode/encode AV1.
+    */
+   return ctx->screen->info.ip[AMD_IP_VCN_ENC].num_instances > 1;
+}
+
 /**
  * creates an UVD compatible decoder
  */
@@ -105,10 +119,13 @@ struct pipe_video_codec *si_uvd_create_decoder(struct pipe_context *context,
 {
    struct si_context *ctx = (struct si_context *)context;
    bool vcn = ctx->vcn_ip_ver >= VCN_1_0_0;
+   struct pipe_video_codec *codec = NULL;
 
    if (templ->entrypoint == PIPE_VIDEO_ENTRYPOINT_ENCODE) {
       if (vcn) {
-         return radeon_create_encoder(context, templ, ctx->ws, si_vce_get_buffer);
+         codec = radeon_create_encoder(context, templ, ctx->ws, si_vce_get_buffer);
+         ctx->vcn_has_ctx = si_vcn_need_context(ctx);
+         return codec;
       } else {
          if (u_reduce_video_profile(templ->profile) == PIPE_VIDEO_FORMAT_HEVC)
             return radeon_uvd_create_encoder(context, templ, ctx->ws, si_vce_get_buffer);
@@ -119,9 +136,10 @@ struct pipe_video_codec *si_uvd_create_decoder(struct pipe_context *context,
               templ->entrypoint == PIPE_VIDEO_ENTRYPOINT_PROCESSING)
       return si_vpe_create_processor(context, templ);
 
-   if (ctx->vcn_ip_ver == VCN_4_0_0)
-      ctx->vcn_has_ctx = true;
-
-   return (vcn) ? radeon_create_decoder(context, templ)
-                : si_common_uvd_create_decoder(context, templ, si_uvd_set_dtb);
+   if (vcn) {
+      codec = radeon_create_decoder(context, templ);
+      ctx->vcn_has_ctx = si_vcn_need_context(ctx);
+      return codec;
+   }
+   return si_common_uvd_create_decoder(context, templ, si_uvd_set_dtb);
 }