diff --git a/wgpu-core/src/binding_model.rs b/wgpu-core/src/binding_model.rs
index a64fe78465..3443bcf782 100644
--- a/wgpu-core/src/binding_model.rs
+++ b/wgpu-core/src/binding_model.rs
@@ -892,7 +892,6 @@ impl<A: HalApi> BindGroup<A> {
         &self,
         bind_group_index: u32,
         offsets: &[wgt::DynamicOffset],
-        limits: &wgt::Limits,
     ) -> Result<(), BindError> {
         if self.dynamic_binding_info.len() != offsets.len() {
             return Err(BindError::MismatchedDynamicOffsetCount {
@@ -908,7 +907,8 @@ impl<A: HalApi> BindGroup<A> {
             .zip(offsets.iter())
             .enumerate()
         {
-            let (alignment, limit_name) = buffer_binding_type_alignment(limits, info.binding_type);
+            let (alignment, limit_name) =
+                buffer_binding_type_alignment(&self.device.limits, info.binding_type);
             if offset as wgt::BufferAddress % alignment as u64 != 0 {
                 return Err(BindError::UnalignedDynamicBinding {
                     group: bind_group_index,
diff --git a/wgpu-core/src/command/compute.rs b/wgpu-core/src/command/compute.rs
index 4c6e005d46..533383c633 100644
--- a/wgpu-core/src/command/compute.rs
+++ b/wgpu-core/src/command/compute.rs
@@ -592,7 +592,7 @@ impl Global {
 
                     bind_group.same_device_as(cmd_buf).map_pass_err(scope)?;
 
-                    let max_bind_groups = cmd_buf.limits.max_bind_groups;
+                    let max_bind_groups = state.device.limits.max_bind_groups;
                     if index >= max_bind_groups {
                         return Err(ComputePassErrorInner::BindGroupIndexOutOfRange {
                             index,
@@ -610,7 +610,7 @@ impl Global {
 
                     let bind_group = state.tracker.bind_groups.insert_single(bind_group);
                     bind_group
-                        .validate_dynamic_bindings(index, &state.temp_offsets, &cmd_buf.limits)
+                        .validate_dynamic_bindings(index, &state.temp_offsets)
                         .map_pass_err(scope)?;
 
                     state.buffer_memory_init_actions.extend(
@@ -766,7 +766,8 @@ impl Global {
 
                     state.flush_states(raw, None).map_pass_err(scope)?;
 
-                    let groups_size_limit = cmd_buf.limits.max_compute_workgroups_per_dimension;
+                    let groups_size_limit =
+                        state.device.limits.max_compute_workgroups_per_dimension;
 
                     if groups[0] > groups_size_limit
                         || groups[1] > groups_size_limit
diff --git a/wgpu-core/src/command/mod.rs b/wgpu-core/src/command/mod.rs
index b237a375a6..b4669f8e7a 100644
--- a/wgpu-core/src/command/mod.rs
+++ b/wgpu-core/src/command/mod.rs
@@ -305,7 +305,6 @@ impl<A: HalApi> CommandBufferMutable<A> {
 ///   whose contents eventually become the property of the submission queue.
 pub struct CommandBuffer<A: HalApi> {
     pub(crate) device: Arc<Device<A>>,
-    limits: wgt::Limits,
     support_clear_texture: bool,
     pub(crate) info: ResourceInfo<CommandBuffer<A>>,
 
@@ -344,7 +343,6 @@ impl<A: HalApi> CommandBuffer<A> {
     ) -> Self {
         CommandBuffer {
             device: device.clone(),
-            limits: device.limits.clone(),
             support_clear_texture: device.features.contains(wgt::Features::CLEAR_TEXTURE),
             info: ResourceInfo::new(label, None),
             data: Mutex::new(
diff --git a/wgpu-core/src/command/render.rs b/wgpu-core/src/command/render.rs
index 2508ae26f7..e2721e71d1 100644
--- a/wgpu-core/src/command/render.rs
+++ b/wgpu-core/src/command/render.rs
@@ -1522,7 +1522,7 @@ impl Global {
                             .map_pass_err(scope)?;
 
                         bind_group
-                            .validate_dynamic_bindings(index, &temp_offsets, &cmd_buf.limits)
+                            .validate_dynamic_bindings(index, &temp_offsets)
                             .map_pass_err(scope)?;
 
                         // merge the resource tracker in