diff --git a/docs/api-reference/core/resources/query-set.md b/docs/api-reference/core/resources/query-set.md
new file mode 100644
index 0000000000..67b6a3cb42
--- /dev/null
+++ b/docs/api-reference/core/resources/query-set.md
@@ -0,0 +1,81 @@
+# QuerySet
+
+:::caution
+This page is incomplete.
+:::
+
+A `QuerySet` object provides an API for using asynchronous GPU queries of the following types
+- timer queries.
+- 'Occlusion' 
+- 'Transform Feedback'
+
+A `QuerySet` holds a number of 64 bit values. 
+
+Timer queries are available if the `timestamp-query` extension is available. (On WebGL 2 this is equivalent to the 
+[`EXT_disjoint_timer_query_webgl2`](https://www.khronos.org/registry/webgl/extensions/EXT_disjoint_timer_query_webgl2/)
+being supported on the current browser.
+
+Note that even when supported, timer queries can fail whenever a change in the GPU occurs that will make the values returned by this extension unusable for performance metrics, for example if the GPU is throttled mid-frame. 
+
+## Usage
+
+Create a timestamp query set:
+
+```typescript
+import {QuerySet} from '@luma.gl/core';
+...
+const timestampQuery = device.createQuerySet({type: 'timestamp'}});
+```
+
+## Types
+
+### `QueryProps`
+
+- `type`: `occlusion` | `timestamp` - type of timer
+- `count`: `number` - number of query results held by this `QuerySet`
+
+| Query Type                               | Usage                                        | Description                                                                |
+| ---------------------------------------- | -------------------------------------------- | -------------------------------------------------------------------------- |
+| `timestamp` (`RenderPass begin/end`)     | `beginRenderPass({timestampQuery: ...})`     | Time taken by GPU to execute RenderPass commands                           |
+| `timestamp`  (`ComputePass begin/end`)   | `beginComputePass({timestampQuery: ...})`    | Time taken by GPU to execute ComputePass commands                          |
+| `occlusion`                              | `beginOcclusionQuery({conservative: false})` | Occlusion query how many fragment samples pass tests (depth, stencil, ...) |
+| `occlusion`                              | `beginOcclusionQuery({conservative: true})`  | Same as above above, but less accurate and faster                          | 
+| `transform-feedback` (Not yet supported) | `beginTransformFeedbackQuery()`              | Number of primitives that are written to transform feedback buffers.       |
+
+In addition to above queries, Query object also provides `getTimeStamp` which returns GPU time stamp at the time this query is executed by GPU. Two sets of these methods can be used to calculate time taken by GPU for a set of GL commands.
+
+## DeviceFeatures
+
+`timestamp-query`: Whether `QuerySet` can be created with type `timestamp`.
+
+## Methods
+
+
+
+### `constructor(device: Device, props: Object)`
+
+`new Query(gl, {})`
+- options.timers=false Object - If true, checks if 'TIME_ELAPSED_EXT' queries are supported
+
+### `destroy()`
+
+Destroys the WebGL object. Rejects any pending query.
+
+- return Query - returns itself, to enable chaining of calls.
+
+
+### beginTimeElapsedQuery()
+
+Shortcut for timer query (dependent on extension in both WebGL 1 and 2)
+
+## RenderPass
+
+
+### `RenderPass.beginOcclusionQuery({conservative = false})`
+
+Shortcut for occlusion query (dependent on WebGL 2)
+
+### `RenderPass.beginTransformFeedbackQuery()`
+
+WebGL 2 only. not yet implemented.
+
diff --git a/docs/developer-guide/profiling.md b/docs/developer-guide/profiling.md
index 3f72996749..3b911175c0 100644
--- a/docs/developer-guide/profiling.md
+++ b/docs/developer-guide/profiling.md
@@ -1,7 +1,7 @@
 # Profiling
 
 GPU programming is all about performance, so having tools to systematically
-measure the performance impact of code changes are critical. luma.gl offers
+measure the performance impact of code changes is critical. luma.gl offers
 several built-in facilities.
 
 ## probe.gl Stats
@@ -31,4 +31,12 @@ however tracking allocations can help spot resource leaks or unnecessary work be
 
 ## Performance Profiling
 
-Queries are supported if available.
+`device.createQuerySet()` can be used to create GPU queries that 
+
+- Occlusion Queries always supported.
+- Timestamp Queries are supported if the `timestamp-query` feature is available, check with `device.features.has('timestamp-query')`.
+
+`QuerySet` instances can be supplied when creating `RenderPass` and `ComputePass` instances.
+
+Results are available through
+`commandEncoder.resolveQuerySet()`
diff --git a/docs/sidebar.json b/docs/sidebar.json
index 8627be713e..4fcc73b3bc 100644
--- a/docs/sidebar.json
+++ b/docs/sidebar.json
@@ -94,7 +94,8 @@
           "api-reference/core/resources/shader",
           "api-reference/core/shader-logs",
           "api-reference/core/resources/texture",
-          "api-reference/core/resources/vertex-array"
+          "api-reference/core/resources/vertex-array",
+          "api-reference/core/resources/query-set"
         ]
       },
       {
diff --git a/modules/core-tests/test/adapter/resources/query-set.spec.ts b/modules/core-tests/test/adapter/resources/query-set.spec.ts
index 6b1e96a7fd..3012597f66 100644
--- a/modules/core-tests/test/adapter/resources/query-set.spec.ts
+++ b/modules/core-tests/test/adapter/resources/query-set.spec.ts
@@ -6,8 +6,7 @@ import test from 'tape-promise/tape';
 import {getTestDevices} from '@luma.gl/test-utils';
 import {QuerySet} from '@luma.gl/core';
 
-
-test('QuerySet construct/delete', async (t) => {
+test('QuerySet construct/delete', async t => {
   for (const device of await getTestDevices()) {
     const querySet = device.createQuerySet({type: 'occlusion', count: 1});
     t.ok(querySet instanceof QuerySet, 'QuerySet construction successful');