Cookie Monster
Diamond Member
- May 7, 2005
- 5,161
- 32
- 86
And what would be the benefit of having two power hungry dies? MCM for core 2 quads makes sense since the chips themselves that make up the "quad core" doesn't have an IMC, nor a PCI-e conctroller and other display handling output logic units. For a GPU they come with all sorts of things that would go wasted if you go MCM. Not to mention the inevitable question. How do you communicate GPU1 to GPU2? This isnt as easy as some people are led to believe. Current technology still lacks the ability to take advantage of multiple GPUs (limitation of scaling, limitation of resource sharing etc etc).
That being said, one GPU itself is made up of LOTS of small cores anyway. These cores many not be powerful like ones found in dual core CPUs but the number of these small cores make up for this.
Its interesting to note that both IHVs decided to skip 45nm all together and go for the 40nm half node instead.
That being said, one GPU itself is made up of LOTS of small cores anyway. These cores many not be powerful like ones found in dual core CPUs but the number of these small cores make up for this.
Its interesting to note that both IHVs decided to skip 45nm all together and go for the 40nm half node instead.
Facebook
Twitter