Buzz
Ever wondered why integrated GPUs share RAM with the system? Sharing RAM with the system not only consumes a portion of the RAM allocated for the CPU but also does not optimize graphics processing. So why don't manufacturers equip integrated GPUs with dedicated GDDR memory blocks to avoid using system RAM and enhance graphics performance?
1. Cost and initial objectives
This is undoubtedly the first issue everyone notices. Providing integrated GPUs with some of their own RAM isn't difficult for giants like Intel and AMD. However, cramming too much into a CPU also means they have to increase its price, reducing its price competitiveness. Not everyone needs separate memory for integrated GPUs, but certainly if they cram more memory into the CPU and increase its price, there will be a lot of people who won't like it. If you notice a bit, Intel's recent F-series CPUs are very popular; they omit integrated GPUs altogether and are a bit cheaper than the regular version. It can be seen that users really like not having to pay extra for something they don't need.
Regarding the purpose of the appearance of integrated GPUs, it is used to help the CPU render images and process simple graphics tasks and... that's it. Its job is only that much and it doesn't significantly affect the price advantage of the CPU. It's a secondary part, and secondary parts shouldn't be prioritized over the main part. Even the G-series CPUs with powerful integrated GPUs from AMD are not equipped with dedicated memory. You might think a GPU integrated with its own RAM is good, but you'll need to rethink the price you have to pay for it. Who wants to spend money on integrated GPUs when they could use that amount for something more useful?
2. Size constraints
Although RAM chips may not be large, cramming them into a small component like the CPU poses a significant challenge. Adding more RAM means the silicon die also has to be larger, resulting in a much larger overall size. Just placing a RAM stick next to the CPU reveals the issue immediately. Just a pair of memory chips on the RAM stick is roughly the size of the CPU's silicon die. If you keep trying to cram more RAM into the CPU, a normal Core i chip might end up as big as Threadripper. Then the standard socket size on the motherboard would also need to be changed to accommodate such a large CPU.
Frankly, adding more RAM just for an integrated GPU is really not worth it.
3. Using system RAM for it... convenient
Basically, system RAM and dedicated GPU RAM are essentially the same DDR RAM, they just differ in how they are tuned for their respective purposes. System RAM is designed for the fastest response speed to smoothly coordinate and synchronize with the CPU. On the other hand, GPU VRAM is designed with the largest possible bandwidth to support the GPU when it tackles massive workloads. System RAM is still fine for temporarily storing image data, it's just not as optimized as VRAM.
Furthermore, anyway, system RAM has to work together with the CPU, so extracting a bit for graphics processing is somewhat more convenient than cramming it with a dedicated GDDR memory block, driving up its price when hardly anyone needs it.
