Further resources for queues from AMD and NVIDIA

An application submits work to a VkQueue, normally in the form of VkCommandBuffer objects or sparse bindings.

Command buffers submitted to a VkQueue start in order, but are allowed to proceed independently after that and complete out of order.

Command buffers submitted to different queues are unordered relative to each other unless you explicitly synchronize them with a VkSemaphore.

You can only submit work to a VkQueue from one thread at a time, but different threads can submit work to a different VkQueue simultaneously.

How a VkQueue is mapped to the underlying hardware is implementation-defined. Some implementations will have multiple hardware queues and submitting work to multiple VkQueue​s will proceed independently and concurrently. Some implementations will do scheduling at a kernel driver level before submitting work to the hardware. There is no current way in Vulkan to expose the exact details how each VkQueue is mapped.

Not all applications will require or benefit from multiple queues. It is reasonable for an application to have a single “universal” graphics supported queue to submit all the work to the GPU.

Queue Family

There are various types of operations a VkQueue can support. A “Queue Family” just describes a set of VkQueue​s that have common properties and support the same functionality, as advertised in VkQueueFamilyProperties.

The following are the queue operations found in VkQueueFlagBits:

  • VK_QUEUE_GRAPHICS_BIT used for vkCmdDraw* and graphic pipeline commands.

  • VK_QUEUE_COMPUTE_BIT used for vkCmdDispatch* and vkCmdTraceRays* and compute pipeline related commands.

  • VK_QUEUE_TRANSFER_BIT used for all transfer commands.

    • VK_PIPELINE_STAGE_TRANSFER_BIT in the Spec describes “transfer commands”.

    • Queue Families with only VK_QUEUE_TRANSFER_BIT are usually for using DMA to asynchronously transfer data between host and device memory on discrete GPUs, so transfers can be done concurrently with independent graphics/compute operations.

    • VK_QUEUE_GRAPHICS_BIT and VK_QUEUE_COMPUTE_BIT can always implicitly accept VK_QUEUE_TRANSFER_BIT commands.

  • VK_QUEUE_SPARSE_BINDING_BIT used for binding sparse resources to memory with vkQueueBindSparse.

  • VK_QUEUE_PROTECTED_BIT used for protected memory.


Knowing which Queue Family is needed

Each operation in the Vulkan Spec has a “Supported Queue Types” section generated from the vk.xml file. The following is 3 different examples of what it looks like in the Spec:


Querying for Queue Family

The following is the simplest logic needed if an application only wants a single graphics VkQueue

uint32_t count = 0;
vkGetPhysicalDeviceQueueFamilyProperties(physicalDevice, &count, nullptr);
std::vector<VkQueueFamilyProperties> properties(count);
vkGetPhysicalDeviceQueueFamilyProperties(physicalDevice, &count, properties.data());

// Vulkan requires an implementation to expose at least 1 queue family with graphics
uint32_t graphicsQueueFamilyIndex;

for (uint32_t i = 0; i < count; i++) {
    if ((properties[i].queueFlags & VK_QUEUE_GRAPHICS_BIT) != 0) {
        // This Queue Family support graphics
        graphicsQueueFamilyIndex = i;

Creating and getting a Queue

Unlike other handles such as VkDevice, VkBuffer, VkDeviceMemory, there is no vkCreateQueue or vkAllocateQueue. Instead, the driver is in charge of creating and destroying the VkQueue handles during vkCreateDevice/vkDestroyDevice time.

The following examples will use the hypothetical implementation which support 3 VkQueue​s from 2 Queue Families:


The following is an example how to create all 3 VkQueue​s with the logical device:

VkDeviceQueueCreateInfo queueCreateInfo[2];
queueCreateInfo[0].queueFamilyIndex = 0; // Transfer
queueCreateInfo[0].queueCount = 1;
queueCreateInfo[1].queueFamilyIndex = 1; // Graphics
queueCreateInfo[1].queueCount = 2;

VkDeviceCreateInfo deviceCreateInfo   = {};
deviceCreateInfo.pQueueCreateInfos    = queueCreateInfo;
deviceCreateInfo.queueCreateInfoCount = 2;

vkCreateDevice(physicalDevice, &deviceCreateInfo, nullptr, &device);

After creating the VkDevice the application can use vkGetDeviceQueue to get the VkQueue handles

VkQueue graphicsQueue0 = VK_NULL_HANDLE;
VkQueue graphicsQueue1 = VK_NULL_HANDLE;
VkQueue transferQueue0 = VK_NULL_HANDLE;

// Can be obtained in any order
vkGetDeviceQueue(device, 0, 0, &transferQueue0); // family 0 - queue 0
vkGetDeviceQueue(device, 1, 1, &graphicsQueue1); // family 1 - queue 1
vkGetDeviceQueue(device, 1, 0, &graphicsQueue0); // family 1 - queue 0