diff --git a/include/libtransistor/display/fence.h b/include/libtransistor/display/fence.h
index c1d0cd27..10cf228c 100644
--- a/include/libtransistor/display/fence.h
+++ b/include/libtransistor/display/fence.h
@@ -12,11 +12,12 @@ extern "C" {
 #include<libtransistor/types.h>
 
 /**
-* @struct fence_t
-* @brief Description here...
-*/
+ * @struct fence_t
+ * @brief Represents conditions for the completion of an asynchronous graphics operation
+ */
 typedef struct {
-	uint32_t unknown[11];
+	uint32_t is_valid;
+	gpu_fence_t sync[4];
 } fence_t;
 
 #ifdef __cplusplus
diff --git a/include/libtransistor/display/graphic_buffer_queue.h b/include/libtransistor/display/graphic_buffer_queue.h
index ebfcae5b..e9eb328a 100644
--- a/include/libtransistor/display/graphic_buffer_queue.h
+++ b/include/libtransistor/display/graphic_buffer_queue.h
@@ -81,14 +81,17 @@ typedef enum {
 * https://source.android.com/reference/hidl/android/hardware/graphics/bufferqueue/1.0/IGraphicBufferProducer#queuebufferinput
 */
 typedef struct {
+	uint32_t size;
+	uint32_t num_fds;
 	int64_t timestamp;
-	bool is_auto_timestamp;
+	int32_t is_auto_timestamp;
 	rect_t crop;
 	int32_t scaling_mode;
 	uint32_t transform;
+	uint32_t sticky_transform;
+	uint32_t unknown[2];
 	fence_t fence;
-	// and a few more unknown fields
-} queue_buffer_input_t;
+} __attribute__((packed)) queue_buffer_input_t;
 
 /**
 * @struct queue_buffer_output_t
diff --git a/include/libtransistor/display/surface.h b/include/libtransistor/display/surface.h
index 86a592cb..2ed08ab1 100644
--- a/include/libtransistor/display/surface.h
+++ b/include/libtransistor/display/surface.h
@@ -33,12 +33,14 @@ typedef struct {
 	surface_state_t state;
 	bool has_requested[2];
 	uint32_t current_slot;
-
+	
 	gpu_buffer_t gpu_buffer;
 	uint32_t *gpu_buffer_memory;
 	uint32_t *gpu_buffer_memory_alloc;
 
 	graphic_buffer_t graphic_buffers[3];
+
+	fence_t current_fence;
 } surface_t;
 
 /**
@@ -59,9 +61,15 @@ result_t surface_create(surface_t *surface, uint64_t layer_id, igbp_t igbp);
 * Using \ref gfx_slow_swizzling_blit is recommended. Call \ref surface_queue_buffer
 * when you're done rendering to submit it to be displayed.
 */
-// for now, we ignore fences
 result_t surface_dequeue_buffer(surface_t *surface, uint32_t **image);
 
+/**
+ * @brief Wait for any asynchronous operations on the current buffer to complete
+ *
+ * @param surface Surface
+ */
+result_t surface_wait_buffer(surface_t *surface);
+
 /**
 * @brief Submit the current buffer to be displayed
 *
diff --git a/include/libtransistor/err.h b/include/libtransistor/err.h
index c7201598..0e523137 100644
--- a/include/libtransistor/err.h
+++ b/include/libtransistor/err.h
@@ -78,7 +78,8 @@ extern "C" {
 #define LIBTRANSISTOR_ERR_SURFACE_DEQUEUE_BUFFER_FAILED LIBTRANSISTOR_RESULT(5007)
 #define LIBTRANSISTOR_ERR_SURFACE_QUEUE_BUFFER_FAILED LIBTRANSISTOR_RESULT(5008)
 #define LIBTRANSISTOR_ERR_SURFACE_INVALID_STATE LIBTRANSISTOR_RESULT(5009)
-	
+#define LIBTRANSISTOR_ERR_DISPLAY_INVALID_FENCE LIBTRANSISTOR_RESULT(5010)
+
 // GPU
 #define LIBTRANSISTOR_ERR_GPU_BUFFER_UNALIGNED LIBTRANSISTOR_RESULT(6001)
 
diff --git a/include/libtransistor/gpu/gpu.h b/include/libtransistor/gpu/gpu.h
index 48db4efd..1b4cc7f7 100644
--- a/include/libtransistor/gpu/gpu.h
+++ b/include/libtransistor/gpu/gpu.h
@@ -22,6 +22,11 @@ typedef struct {
 	uint8_t kind;
 } gpu_buffer_t;
 
+typedef struct {
+	uint32_t syncpt_id;
+	uint32_t syncpt_value;
+} gpu_fence_t;
+
 /**
 * @brief Initialize GPU
 */
@@ -63,6 +68,14 @@ result_t gpu_buffer_get_id(gpu_buffer_t *gpu_b, uint32_t *id);
 */
 result_t gpu_buffer_initialize_from_id(gpu_buffer_t *gpu_b, uint32_t id);
 
+
+/**
+ * @brief Waits on a fence to complete
+ *
+ * @param fence Fence to wait for completion on
+ */
+result_t gpu_wait_fence(gpu_fence_t *fence, uint32_t timeout);
+
 /**
 * @brief Finalize GPU
 */
diff --git a/include/libtransistor/gpu/nv_ioc.h b/include/libtransistor/gpu/nv_ioc.h
index 4245e081..c11f3f6b 100644
--- a/include/libtransistor/gpu/nv_ioc.h
+++ b/include/libtransistor/gpu/nv_ioc.h
@@ -9,20 +9,20 @@
 extern "C" {
 #endif
 
-#define NVHOST_IOCTL_CTRL_SYNCPT_READ 0xC0080014
-#define NVHOST_IOCTL_CTRL_SYNCPT_INCR 0x40040015
-#define NVHOST_IOCTL_CTRL_SYNCPT_WAIT 0xC00C0016
-#define NVHOST_IOCTL_CTRL_MODULE_MUTEX 0x40080017
-#define NVHOST_IOCTL_CTRL_MODULE_REGRDWR 0xC0180018
-#define NVHOST_IOCTL_CTRL_SYNCPT_WAITEX 0xC0100019
-#define NVHOST_IOCTL_CTRL_SYNCPT_READ_MAX 0xC008001A
-#define NVHOST_IOCTL_CTRL_GET_CONFIG 0xC183001B
-#define NVHOST_IOCTL_CTRL_EVENT_SIGNAL 0xC004001C
-#define NVHOST_IOCTL_CTRL_EVENT_WAIT 0xC010001D
-#define NVHOST_IOCTL_CTRL_EVENT_WAIT_ASYNC 0xC010001E
-#define NVHOST_IOCTL_CTRL_EVENT_REGISTER 0xC004001F
-#define NVHOST_IOCTL_CTRL_EVENT_UNREGISTER 0xC0040020
-#define NVHOST_IOCTL_CTRL_EVENT_KILL 0x40080021
+#define NVHOST_IOC_CTRL_SYNCPT_READ 0xC0080014
+#define NVHOST_IOC_CTRL_SYNCPT_INCR 0x40040015
+#define NVHOST_IOC_CTRL_SYNCPT_WAIT 0xC00C0016
+#define NVHOST_IOC_CTRL_MODULE_MUTEX 0x40080017
+#define NVHOST_IOC_CTRL_MODULE_REGRDWR 0xC0180018
+#define NVHOST_IOC_CTRL_SYNCPT_WAITEX 0xC0100019
+#define NVHOST_IOC_CTRL_SYNCPT_READ_MAX 0xC008001A
+#define NVHOST_IOC_CTRL_GET_CONFIG 0xC183001B
+#define NVHOST_IOC_CTRL_EVENT_SIGNAL 0xC004001C
+#define NVHOST_IOC_CTRL_EVENT_WAIT 0xC010001D
+#define NVHOST_IOC_CTRL_EVENT_WAIT_ASYNC 0xC010001E
+#define NVHOST_IOC_CTRL_EVENT_REGISTER 0xC004001F
+#define NVHOST_IOC_CTRL_EVENT_UNREGISTER 0xC0040020
+#define NVHOST_IOC_CTRL_EVENT_KILL 0x40080021
 
 #define NVMAP_IOC_CREATE 0xC0080101
 #define NVMAP_IOC_FROM_ID 0xC0080103
@@ -32,15 +32,25 @@ extern "C" {
 #define NVMAP_IOC_GET_ID 0xC008010E
 
 /**
- * @struct nvhost_ioctl_ctrl_event_wait_args
+ * @struct nvhost_ioctl_ctrl_syncpt_wait_args
  * @brief Arguments to wait on a syncpt
  */
+typedef struct {
+	uint32_t syncpt_id; ///< In
+	uint32_t threshold; ///< In
+	int32_t timeout; ///< In
+} nvhost_ioc_ctrl_syncpt_wait_args;
+
+/**
+ * @struct nvhost_ioctl_ctrl_event_wait_args
+ * @brief Arguments to wait on a syncpt event
+ */
 typedef struct {
 	uint32_t syncpt_id; ///< In
 	uint32_t threshold; ///< In
 	int32_t timeout; ///< In
 	uint32_t value; ///< Inout
-} nvhost_ioctl_ctrl_event_wait_args;
+} nvhost_ioc_ctrl_event_wait_args;
 
 /**
 * @struct nvmap_ioc_create_args
diff --git a/lib/display/graphic_buffer_queue.c b/lib/display/graphic_buffer_queue.c
index ca991cd1..6818bf88 100644
--- a/lib/display/graphic_buffer_queue.c
+++ b/lib/display/graphic_buffer_queue.c
@@ -41,26 +41,7 @@ static result_t queue_buffer_output_unflatten(parcel_t *parcel, queue_buffer_out
 }
 
 static result_t queue_buffer_input_flatten(parcel_t *parcel, queue_buffer_input_t *qbi) {
-	static uint32_t ts32 = 0x588bbba9;
-	uint32_t template[] = {
-		0x54, 0, // unknown, but always these values
-		ts32, 0x0, //u64 timestamp
-		1, 0, 0, // unknown, but always these values
-		0, 0, // sometimes zero
-		0,
-		0, // also seen 2
-
-		0, 0,
-
-		// fence?
-		1, 1, 0xa3, 0x0,
-		-1, 0, -1, 0, -1, 0
-	};
-
-	ts32++;
-	//ts32+= 0x2e45f00;
-
-	memcpy(parcel_write_inplace(parcel, sizeof(template)), &template, sizeof(template));
+	memcpy(parcel_write_inplace(parcel, sizeof(*qbi)), qbi, sizeof(*qbi));
   
 	return RESULT_OK;
 }
@@ -127,6 +108,20 @@ static result_t graphic_buffer_flatten(parcel_t *parcel, graphic_buffer_t *gb) {
 }
 
 static result_t fence_unflatten(parcel_t *parcel, fence_t *fence) {
+	struct {
+		uint32_t size;
+		uint32_t num_fds;
+	} *header;
+	header = parcel_read_inplace(parcel, sizeof(*header));
+	if(header->size != sizeof(fence_t)) {
+		return LIBTRANSISTOR_ERR_DISPLAY_INVALID_FENCE;
+	}
+	if(header->num_fds != 0) {
+		return LIBTRANSISTOR_ERR_DISPLAY_FENCE_TOO_MANY_FDS;
+	}
+	for(size_t i = 0; i < ARRAY_LENGTH(fence->sync); i++) {
+		fence->sync[i].syncpt_id = 0xffffffff;
+	}
 	memcpy(fence, parcel_read_inplace(parcel, sizeof(*fence)), sizeof(*fence));
 	return RESULT_OK;
 }
@@ -182,11 +177,17 @@ result_t igbp_dequeue_buffer(igbp_t *igbp, uint32_t width, uint32_t height, pixe
 	}
 
 	*slot = parcel_read_u32(&response);
-	if((r = fence_unflatten(&response, fence)) != RESULT_OK) {
-		return r;
+
+	int32_t has_fence = parcel_read_u32(&response);
+	if(has_fence) {
+		if((r = fence_unflatten(&response, fence)) != RESULT_OK) {
+			return r;
+		}
+	} else {
+		fence->is_valid = false;
 	}
 	*status = parcel_read_u32(&response);
-
+	
 	return RESULT_OK;
 }
 
diff --git a/lib/display/surface.c b/lib/display/surface.c
index 56b251c8..10b3e69f 100644
--- a/lib/display/surface.c
+++ b/lib/display/surface.c
@@ -104,12 +104,11 @@ result_t surface_dequeue_buffer(surface_t *surface, uint32_t **image) {
 	}
 
 	int status;
-	fence_t fence;
 	if((r = igbp_dequeue_buffer(
 		    &surface->igbp,
 		    1280, 720, 1, 0xb00,
 		    false, &status, &surface->current_slot,
-		    &fence, NULL)) != RESULT_OK) {
+		    &surface->current_fence, NULL)) != RESULT_OK) {
 		return r;
 	}
 
@@ -144,11 +143,31 @@ result_t surface_queue_buffer(surface_t *surface) {
 	if(surface->state != SURFACE_STATE_DEQUEUED) {
 		return LIBTRANSISTOR_ERR_SURFACE_INVALID_STATE;
 	}
+
+	queue_buffer_input_t qbi;
+	qbi.size = sizeof(qbi) - 0x8;
+	qbi.num_fds = 0;
+	qbi.timestamp = 0;
+	qbi.is_auto_timestamp = 0;
+	qbi.crop.left = 0;
+	qbi.crop.top = 0;
+	qbi.crop.right = 0;
+	qbi.crop.bottom = 0;
+	qbi.scaling_mode = 0;
+	qbi.transform = 0;
+	qbi.sticky_transform = 0;
+	qbi.unknown[0] = 0;
+	qbi.unknown[1] = 1;
+	qbi.fence.is_valid = 1;
+	for(size_t i = 0; i < ARRAY_LENGTH(qbi.fence.sync); i++) {
+		qbi.fence.sync[i].syncpt_id = 0xffffffff;
+		qbi.fence.sync[i].syncpt_value = 0;
+	}
 	
 	queue_buffer_output_t qbo;
 	int status;
 	if((r = igbp_queue_buffer(&surface->igbp, surface->current_slot,
-	                          NULL, &qbo, &status)) != RESULT_OK) {
+	                          &qbi, &qbo, &status)) != RESULT_OK) {
 		return r;
 	}
 
@@ -160,3 +179,26 @@ result_t surface_queue_buffer(surface_t *surface) {
 	
 	return RESULT_OK;
 }
+
+result_t surface_wait_buffer(surface_t *surface) {
+	if(surface->state != SURFACE_STATE_DEQUEUED) {
+		return LIBTRANSISTOR_ERR_SURFACE_INVALID_STATE;
+	}
+
+	if(!surface->current_fence.is_valid) {
+		return RESULT_OK;
+	}
+
+	for(size_t i = 0; i < ARRAY_LENGTH(surface->current_fence.sync); i++) {
+		result_t r;
+		gpu_fence_t *f = &surface->current_fence.sync[i];
+		if(f->syncpt_id == 0xffffffff) {
+			continue;
+		}
+		if((r = gpu_wait_fence(f, -1)) != RESULT_OK) {
+			return r;
+		}
+	}
+
+	return RESULT_OK;
+}
diff --git a/lib/gpu/gpu.c b/lib/gpu/gpu.c
index 36fdf05f..ae8f5aae 100644
--- a/lib/gpu/gpu.c
+++ b/lib/gpu/gpu.c
@@ -7,6 +7,7 @@
 
 static int nvas_fd;
 static int nvmap_fd;
+static int nvhost_ctrl_fd;
 static int gpu_initializations = 0;
 
 result_t gpu_initialize() {
@@ -29,8 +30,15 @@ result_t gpu_initialize() {
 		goto fail_nvas;
 	}
 
+	if((nvhost_ctrl_fd = nv_open("/dev/nvhost-ctrl")) < 0) {
+		r = nv_result;
+		goto fail_nvmap;
+	}
+
 	return RESULT_OK;
 
+fail_nvhost_ctrl:
+	nv_close(nvhost_ctrl_fd);
 fail_nvmap:
 	nv_close(nvmap_fd);
 fail_nvas:
@@ -140,7 +148,23 @@ result_t gpu_buffer_initialize_from_id(gpu_buffer_t *gpu_b, uint32_t id) {
 	return RESULT_OK;
 }
 
+result_t gpu_wait_fence(gpu_fence_t *fence, uint32_t timeout) {
+	INITIALIZATION_GUARD(gpu);
+
+	nvhost_ioc_ctrl_syncpt_wait_args nvh_wait;
+	nvh_wait.syncpt_id = fence->syncpt_id;
+	nvh_wait.threshold = fence->syncpt_value;
+	nvh_wait.timeout = timeout;
+
+	if(nv_ioctl(nvhost_ctrl_fd, NVHOST_IOC_CTRL_SYNCPT_WAIT, &nvh_wait, sizeof(nvh_wait)) != 0) {
+		return nv_result;
+	}
+
+	return RESULT_OK;
+}
+
 static void gpu_force_finalize() {
+	nv_close(nvhost_ctrl_fd);
 	nv_close(nvmap_fd);
 	nv_close(nvas_fd);
 	nv_finalize();
diff --git a/test/test_display.c b/test/test_display.c
index 399698ed..4d3f5a7c 100644
--- a/test/test_display.c
+++ b/test/test_display.c
@@ -38,9 +38,12 @@ int main() {
 		printf("begin frame %d\n", i);
 		uint32_t *out_buffer;
 		ASSERT_OK(fail_display_event, surface_dequeue_buffer(&surf, &out_buffer));
-
+		if(i < 300) {
+			ASSERT_OK(fail_display_event, surface_wait_buffer(&surf));
+		}
+		
 		for(size_t p = 0; p < (0x3c0000/sizeof(uint32_t)); p++) {
-			out_buffer[p] = 0xFF0000FF;
+			out_buffer[p] = (i < 300) ? 0xFFFF0000 : 0xFF0000FF;
 		}
 		int x = (cos((double) i * 6.28 / 60.0) * 300.0 + 350.0);
 		int y = (sin((double) i * 6.28 / 60.0) * 300.0 + 350.0);