diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c
index 2da06e838b9..705a2f7c354 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c
@@ -309,6 +309,8 @@ void amdgpu_bo_destroy(struct amdgpu_winsys *aws, struct pb_buffer_lean *_buf)
       aws->allocated_vram -= align64(bo->b.base.size, aws->info.gart_page_size);
    else if (bo->b.base.placement & RADEON_DOMAIN_GTT)
       aws->allocated_gtt -= align64(bo->b.base.size, aws->info.gart_page_size);
+   else if (bo->b.base.placement & RADEON_DOMAIN_OA)
+      aws->allocated_oa -= bo->b.base.size;
 
    simple_mtx_destroy(&bo->map_lock);
    FREE(bo);
@@ -716,6 +718,8 @@ static struct amdgpu_winsys_bo *amdgpu_create_bo(struct amdgpu_winsys *aws,
       aws->allocated_vram += align64(size, aws->info.gart_page_size);
    else if (initial_domain & RADEON_DOMAIN_GTT)
       aws->allocated_gtt += align64(size, aws->info.gart_page_size);
+   else if (initial_domain & RADEON_DOMAIN_OA)
+      aws->allocated_oa += size;
 
    amdgpu_add_buffer_to_global_list(aws, bo);
 
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.cpp b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.cpp
index a1b960e4792..8578d0cd9c8 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.cpp
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.cpp
@@ -1392,12 +1392,7 @@ static int amdgpu_cs_submit_ib_kernelq(struct amdgpu_cs *acs,
 
    assert(num_chunks <= 8);
 
-   /* Submit the command buffer.
-    *
-    * The kernel returns -ENOMEM with many parallel processes using GDS such as test suites
-    * quite often, but it eventually succeeds after enough attempts. This happens frequently
-    * with dEQP using NGG streamout.
-    */
+   /* Submit the command buffer. */
    int r = 0;
 
    do {
@@ -1406,7 +1401,12 @@ static int amdgpu_cs_submit_ib_kernelq(struct amdgpu_cs *acs,
          os_time_sleep(1000);
 
       r = ac_drm_cs_submit_raw2(aws->dev, acs->ctx->ctx_handle, 0, num_chunks, chunks, seq_no);
-   } while (r == -ENOMEM);
+
+      /* The kernel returns -ENOMEM with many parallel processes using GDS such as test suites
+       * quite often, but it eventually succeeds after enough attempts. This happens frequently
+       * with dEQP using NGG streamout.
+       */
+   } while (r == -ENOMEM && aws->allocated_oa);
 
    return r;
 }
@@ -2048,21 +2048,7 @@ static void amdgpu_cs_submit_ib(void *job, void *gdata, int thread_index)
       r = 0;
    } else {
       if (queue_type != USERQ) {
-         /* Submit the command buffer.
-          *
-          * The kernel returns -ENOMEM with many parallel processes using GDS such as test suites
-          * quite often, but it eventually succeeds after enough attempts. This happens frequently
-          * with dEQP using NGG streamout.
-          */
-         r = 0;
-
-         do {
-            /* Wait 1 ms and try again. */
-            if (r == -ENOMEM)
-               os_time_sleep(1000);
-
-            r = amdgpu_cs_submit_ib_kernelq(acs, num_submit_real_buffers, bo_list, &seq_no);
-         } while (r == -ENOMEM);
+         r = amdgpu_cs_submit_ib_kernelq(acs, num_submit_real_buffers, bo_list, &seq_no);
 
          if (!r) {
             /* Success. */
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h
index be14f49aca2..66922653cd2 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h
@@ -219,6 +219,7 @@ struct amdgpu_winsys {
    uint32_t next_bo_unique_id;
    uint64_t allocated_vram;
    uint64_t allocated_gtt;
+   uint32_t allocated_oa;
    uint64_t mapped_vram;
    uint64_t mapped_gtt;
    uint64_t slab_wasted_vram;