Strictly speaking, there is no such thing as a "present family", nor is there a "graphics family". There are just different queue families which support different types of operations, like presentation or graphics.
How many different queue families are supported and which capabilities they have depends on your specific GPU model. To get a good overview of all this information, I can recommend the tool Hardware Capability Viewer by Sascha Willems. (It is also used to populate the database behind the awesome gpuinfo.org if you choose to upload your data.)
On an NVIDIA RTX 2080, for example, I get the following different queue families with the following capabilities:
- Queue family #0 supports transfer, graphics, compute, and presentation
- Queue family #1 supports only transfer, and nothing else
- Queue family #2 supports transfer, compute, and presentation
As you can see, I could use queues from queue families #0 or #2 to send images to the presentation engine, but not queues from queue family #1.
I find the capabilities quite interesting. It means that for certain use cases, using one of the specialized queue families (i.e. #1 or #2) can lead to more optimal performance than using queues from family #0 which are able to perform any operation. Using different queues can enable your application to parallelize work better, but it will generally also require some sort of work-synchronization between the different queues.
Queues from family #2 are often referred to as "async compute queues" (I think, this terminology came mostly from the DirectX world), and can be enabled in games' graphics settings for quite a while now (if supported). What I have spotted recently is the option to enable "present from compute" (Doom Eternal offers this setting) and again, this would refer to queues from family #2. I would guess that this does not automatically lead to increased performance (which is why it can be enabled/disabled) but on some GPUs it definitely will.
Answering your specific question: A queue family does not have to support graphics capabilities in order to support presentation. There are queue families (e.g. on an RTX 2080) which support compute and presentation, but not graphics. All of this depends on the specific GPU model. I don't know if there are any GPUs that offer transfer-only queue families with presentation support --- maybe that doesn't make too much sense, so I guess rather not.
