I was told to do a presentation on GPU and was told to focus on the major types of graphics solutions, disadvantages and advantages of the graphics solutions and the different features on each one. I'm a programmer and don't know much about hardware so I'm a bit confused here. From my research, a GPU (graphics processing unit) is different from a graphics card (a GPU is part of the graphics card - this is referred to as integrated? and some GPU's are part of the motherboard - this is known as dedicated?).
When I searched wiki, I found this article: https://en.wikipedia.org/wiki/Graphics_processing_unit#GPU_forms
The article talks about GPU forms and mentions "Dedicated graphics card", "Integrated graphics solutions", "Hybrid solutions", "Stream Processing and General Purpose GPUs (GPGPU)" and "External GPU". Are these all different types of GPU's? (I'm asking because the first form is called a 'graphics card', the next two are called 'graphics solutions' and the last two are called 'GPUs'). Are they all different types of graphics solutions?
Answer
What your asking almost belongs on an english language site :-) Is there a proper technical term for a solution? Or did this just get made up from two common words , and would not be any different than saying "connection answer" for a ethernet wire, it just sounds good.
The dictionary definition of the word "solution" that applies here, is "The method or process of solving a problem" Every displaying device built-in or carded or soldered in or even running off the side with a USB connection, would be a solution for displaying grafics.
The dictionary definition for grafics , that applies here is "The process by which a computer displays data pictorally."
Conclusion: They are all "Grafics Solutions". Manufactures may have created , creative and accepted terminology words and word combos, that get adopted and will eventually sound technically more correct, that does not make anyone wrong for calling a shovel "shovel" :-)
Integrated or On-Die - Usually referring to GPUs that exist on the same pcb, more often in the same die (silicon chip) as the CPU. Most of thier ram use will be the same ram modules as the CPU itself is using.
Dedicated (card) - Referring to a carded GPU that sticks in some slot in the motherboard that is a GPU and video memory and frame buffer and video ports, all capable of doing the work somewhat seperated , with fast I/O lines to get the data to them.
Onboard - Integrated as a chip onto the motherboard, usually weak processing capability , intended for displaying more than for huge processing and rendering 3d stuff. Does the job of being able to see.
Dedicated (laptop) - But onboard :-) In laptops the dedicated GPU unit may very well be socketed or soldered direct to the motherboard, or a daughterboard , is usually called dedicated. Even if a laptop has a weak "onboard" or "On-Die" gpu capability for simple displaying, a "dedicated" GPU will be capable of fast processing, and has its own video memory. The final picture (frame buffer) will be sent out to the onboard or on-die gpu for display. The seperate GPU chip that kicks in to do the really hard GPU work is called dedicated in the latptops.
notes: a few laptops have seperate video ports and fully seperated gpu capability.
Hybrid - Combo of having a seperated GPU chip to do processing, but still utalising the normal motherboard ram for the video ram.
External - well anything that can be stuffed on the outside, which includes low end to high end gpus that get thier data i/O from various ports like USB and even PCI/PCI-E ports that are accessable ports on most laptops External could possibly even refer to a dedicated (card) that is slotted into a "docking" type hardware that is external to a laptop providing much of the same capability as a desktop.
Processing unit - stream processing, whatever they want to call it, similar to CPU cards that would drop into a slot and be capable of processing what was sent to them, the GPU processing unit becomming more capable of being used for other stuff other than render and display. To process general code that can be easily parellel processed, code that would previously be cpu processed, that the GPUs parellel many processing units would work well with.
No comments:
Post a Comment