diff --git a/src/freedreno/vulkan/tu_acceleration_structure.cc b/src/freedreno/vulkan/tu_acceleration_structure.cc
index 6ab0c81a90b..c1355d2b26c 100644
--- a/src/freedreno/vulkan/tu_acceleration_structure.cc
+++ b/src/freedreno/vulkan/tu_acceleration_structure.cc
@@ -234,11 +234,6 @@ encode(VkCommandBuffer commandBuffer,
    tu_dispatch_unaligned_indirect(commandBuffer,
                                   intermediate_header_addr +
                                   offsetof(struct vk_ir_header, ir_internal_node_count));
-
-   *(VkDeviceSize *)
-      util_sparse_array_get(&device->accel_struct_ranges,
-                            vk_acceleration_structure_get_va(dst)) = dst->size;
-
 }
 
 static VkResult
diff --git a/src/freedreno/vulkan/tu_descriptor_set.cc b/src/freedreno/vulkan/tu_descriptor_set.cc
index a79c6d950a7..ff4e286cc30 100644
--- a/src/freedreno/vulkan/tu_descriptor_set.cc
+++ b/src/freedreno/vulkan/tu_descriptor_set.cc
@@ -1160,15 +1160,16 @@ write_sampler_descriptor(uint32_t *dst, VkSampler _sampler)
 }
 
 static void
-write_accel_struct(uint32_t *dst, uint64_t va, uint64_t size)
+write_accel_struct(uint32_t *dst, uint64_t va)
 {
    dst[0] = A6XX_TEX_CONST_0_TILE_MODE(TILE6_LINEAR) | A6XX_TEX_CONST_0_FMT(FMT6_32_UINT);
 
-   /* The overall range of the entire AS may be more than the max range, but
-    * the SSBO is only used to access the instance descriptors and header.
-    * Make sure that we don't specify a too-large range.
+   /* We don't actually use the bounds checking in the shader, since the
+    * instance array is accessed entirely with a driver-controlled offset.
+    * Therefore just always specify the maximum possible size to avoid having
+    * to keep track of the size.
     */
-   dst[1] = MAX2(DIV_ROUND_UP(size, AS_RECORD_SIZE), MAX_TEXEL_ELEMENTS);
+   dst[1] = MAX_TEXEL_ELEMENTS;
    dst[2] =
       A6XX_TEX_CONST_2_STRUCTSIZETEXELS(AS_RECORD_SIZE / 4) |
       A6XX_TEX_CONST_2_STARTOFFSETTEXELS(0) |
@@ -1229,13 +1230,9 @@ tu_GetDescriptorEXT(
       break;
    case VK_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR: {
       if (pDescriptorInfo->data.accelerationStructure == 0) {
-         write_accel_struct(dest, device->null_accel_struct_bo->iova,
-                            device->null_accel_struct_bo->size);
+         write_accel_struct(dest, device->null_accel_struct_bo->iova);
       } else {
-         VkDeviceSize size = *(VkDeviceSize *)
-            util_sparse_array_get(&device->accel_struct_ranges,
-                                  pDescriptorInfo->data.accelerationStructure);
-         write_accel_struct(dest, pDescriptorInfo->data.accelerationStructure, size);
+         write_accel_struct(dest, pDescriptorInfo->data.accelerationStructure);
       }
       break;
    }
@@ -1361,11 +1358,9 @@ tu_update_descriptor_sets(const struct tu_device *device,
             VK_FROM_HANDLE(vk_acceleration_structure, accel_struct, accel_structs->pAccelerationStructures[j]);
             if (accel_struct) {
                write_accel_struct(ptr,
-                                  vk_acceleration_structure_get_va(accel_struct),
-                                  accel_struct->size);
+                                  vk_acceleration_structure_get_va(accel_struct));
             } else {
-               write_accel_struct(ptr, device->null_accel_struct_bo->iova,
-                                  device->null_accel_struct_bo->size);
+               write_accel_struct(ptr, device->null_accel_struct_bo->iova);
             }
             break;
          }
@@ -1703,11 +1698,9 @@ tu_update_descriptor_set_with_template(
             VK_FROM_HANDLE(vk_acceleration_structure, accel_struct, *(const VkAccelerationStructureKHR *)src);
             if (accel_struct) {
                write_accel_struct(ptr,
-                                  vk_acceleration_structure_get_va(accel_struct),
-                                  accel_struct->size);
+                                  vk_acceleration_structure_get_va(accel_struct));
             } else {
-               write_accel_struct(ptr, device->null_accel_struct_bo->iova,
-                                  device->null_accel_struct_bo->size);
+               write_accel_struct(ptr, device->null_accel_struct_bo->iova);
             }
             break;
          }
diff --git a/src/freedreno/vulkan/tu_device.cc b/src/freedreno/vulkan/tu_device.cc
index 8207695cc4e..43c627b5229 100644
--- a/src/freedreno/vulkan/tu_device.cc
+++ b/src/freedreno/vulkan/tu_device.cc
@@ -2658,8 +2658,6 @@ tu_CreateDevice(VkPhysicalDevice physicalDevice,
    if (result != VK_SUCCESS)
       goto fail_queues;
 
-   util_sparse_array_init(&device->accel_struct_ranges, sizeof(VkDeviceSize), 256);
-
    mtx_init(&device->radix_sort_mutex, mtx_plain);
 
    {
@@ -2948,7 +2946,6 @@ fail_free_zombie_vma:
    u_vector_finish(&device->zombie_vmas);
    ir3_compiler_destroy(device->compiler);
 fail_compiler:
-   util_sparse_array_finish(&device->accel_struct_ranges);
    vk_meta_device_finish(&device->vk, &device->meta);
 fail_queues:
    for (unsigned i = 0; i < TU_MAX_QUEUE_FAMILIES; i++) {
@@ -3001,8 +2998,6 @@ tu_DestroyDevice(VkDevice _device, const VkAllocationCallbacks *pAllocator)
 
    vk_meta_device_finish(&device->vk, &device->meta);
 
-   util_sparse_array_finish(&device->accel_struct_ranges);
-
    ir3_compiler_destroy(device->compiler);
 
    vk_pipeline_cache_destroy(device->mem_cache, &device->vk.alloc);
diff --git a/src/freedreno/vulkan/tu_device.h b/src/freedreno/vulkan/tu_device.h
index 8a476f0f45e..18bebb6c76c 100644
--- a/src/freedreno/vulkan/tu_device.h
+++ b/src/freedreno/vulkan/tu_device.h
@@ -300,8 +300,6 @@ struct tu_device
    radix_sort_vk_t *radix_sort;
    mtx_t radix_sort_mutex;
 
-   struct util_sparse_array accel_struct_ranges;
-
 #define MIN_SCRATCH_BO_SIZE_LOG2 12 /* A page */
 
    /* Currently the kernel driver uses a 32-bit GPU address space, but it