It would speed things up slightly.
But the main reason why a dedicated graphics card has its own Gddr 5 ram on the card it`s self is to do with
Bandwidth or how wide the data bus is in bit`s between the Gpu and the memory situated on the card.
The fatter the pipe, or data bus in numbers the more data you can pass through the bus, in a shorter amount of time,one data pass, cycle or refresh, at a larger volume of data.
When you use on board system memory, for example DDR 3 @ any speed.
And for example a APU or cpu with a Gpu solution, the problem is how wide the data bus is.
Depending on the motherboard and the config of DDR 3 main system memory, amount of ram sticks and memory pairing modes.
64 bit`s in single channel mode
128 bits dual channel mode.
256 bits quad channel mode.
So in effect
Apu or gpu is limited on bus or data width due to system memory, or what people call a bottle neck of data where x amount of Data is slowed down due to a difference in bus or data processing width.
Most new graphics cards have a higher quality of dedicated ram, capacity, and the bus or bit width is often larger to cope compared to what a on board cpu/apu/gpu solution provides as it`s specs.
That greatly speeds up performance in data manipulation and processing in many ways.
Going from say single channel memory system config to dual or quad channel mode would help in some way and a higher speed rating of your DDR 3 main system memory.
Again though it is dependent on the bit or bus width of the apu or gpu data bus internal to the combined cpu/ gpu solution processor you use.
Amd or Intel based.