-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Why are GPUs not single chips?
- Thread starter Edward78
- Start date
You mean like a socketed CPU, but only suitable for parallel computation? Probably because between PCI Express cards and integrated GPUs there's no market for them.
Also there's GPU chip size (Nvidia GK110 is 550 mm2, c.f. Ivybridge 4C 160 mm2), heat output, and memory bandwidth to deal with. GPUs mostly use GDDR5, but motherboards are only manufactured for DDR3 (DDR4 is on it's way). The way around that would be to include some fast memory on-chip, but then you've just reinvented the integrated GPU!
Also there's GPU chip size (Nvidia GK110 is 550 mm2, c.f. Ivybridge 4C 160 mm2), heat output, and memory bandwidth to deal with. GPUs mostly use GDDR5, but motherboards are only manufactured for DDR3 (DDR4 is on it's way). The way around that would be to include some fast memory on-chip, but then you've just reinvented the integrated GPU!
That's just embedded graphics. It already exists in laptops which have suitable sockets that support removable/replaceable chips.
Admittedly the upgrade paths are not anywhere near as diverse as PCI-E but it's possible nonetheless.
The problem is GPUs have very specific requirements. Power requirement is a huge one for example. The amount of supporting hardware that goes onto a GPU PCB is quite a big ask of a motherboard that needs to balance price, compatibility and feasibility.
Another example is heat dissipation. How do you support a socketed single GPU core on, for example, a motherboard AND be able to cool it effectively? Just not enough room in the example you pose with sub Mini ITX form factors.
