Given that PCs need a standalone GPU in order to play games, how does a small chip like the RDNA2 manage to process that much raw power, much almost similar to what a PC’s GPU can do?
In: Technology
the actual size of the silicon (ie die size) of the RDNA2 and modern PC GPU’s are similar in size.
the reason the PC graphics card is “huge” is because of all the cooling required, whereas the cooling for a console GPU is built into the console itself.
PC’s don’t need a discrete GPU to play games. Many CPU’s have onbord graphics that can play a decent amount of games. The main issues are power consumption, heat dissipation, and the expectations of the user. The steam deck for example has a 720p screen. That will mean less work for the graphics hardware how ever even with this advantage the processor has a power limit of I think 15 watts. This is because all that heat in a small area is hard to get rid of especially for a handheld.
The reason that you have a dedicated GPU in a PC is because the cores of a GPU are different from the cores of a normal CPU. They are optimized for doing some specific kinds of calculations that need to be done many thousands of times per second to get your video output correct, whereas your CPU is made for general purpose computing, which means it needs to have everything available to it.
Think of a GPU like a drag race car, designed for that short burst of speed. It isn’t going to do very well on city streets, taking turns, hauling cargo, etc. It can certainly do its one job very very well, but isn’t good for the general purpose.
Now why they’re separate in a PC is still another question. It comes down to two elements. First is consumer choice, being able to choose a separate CPU and GPU. The second is in design. Not every CPU maker wants to be a GPU maker, and not every GPU maker wants to be a CPU maker. They can each specialize in their own product.
Now applications like a console or the Steam Deck have different requirements. They are going to ship with the same CPU and same GPU for everyone, but they are going to be more constrainted on space and cooling than a traditional PC. However, they don’t have the same requirements for modularity. This means they can throw away the idea of keeping the CPU and GPU separate, and create one processor that has cores for both kinds in it. This lets them optimize the physical space on the board, as well as the cooling, etc.
Latest Answers