Nearly all modern electronic devices include graphics capabilities of some form, and the range of features and performance can be incredibly vast.
© GETTY IMAGES/ANDREW BROOKES
The PC has gone through a lot of changes, branching out in myriad ways and names. But whatever they’re called, every computing device needs certain features: CPU, GPU, RAM, storage, the mainboard, and power. Take any one of those away, and the functionality plummets or fails. How much capability you need from each component varies by the intended use case.
Our focus for this state of the PC investigation will be on the GPU, specifically integrated graphics solutions. The GPU replaced the lesser video controllers of early PCs, which in turn replaced the punch cards of a bygone era. Whatever the name or complexity, the purpose is to output something useful.
Watches, phones, TVs, tablets, laptops, and desktops all have some form of display these days, with plenty of differentiation. For the PC, however, there are two categories of GPU: integrated graphics, which resides either in the same silicon as the CPU, or at least in the same package, and dedicated graphics. The integrated versions are usually lower power and performance, but are also basically ‘free’—they come as part of the CPU.
These days, integrated GPUs can do just about everything you’d need, as far as typical computing goes. But how do they compare with dedicated GPUs, and how much have things improved over the past few generations? Let’s dig deeper into integrated graphics.
A BRIEF HISTORY OF INTEGRATED GRAPHICS
Go back far enough into the history of PCs, and much of the functionality that is now part of the CPU used to be in a separate chip. Most desktop PCs prior to the late 90s had a dedicated graphics card—there were SVGA, VGA, EGA, and MCGA devices. But as technology advanced and Moore’s Law allowed for increasingly large numbers of transistors to be packed into a chip, various elements were consolidated.
Our modern processors are effectively a system on a chip: nearly everything needed for main functionality is part of the CPU package. Memory controllers, cache, USB controllers, and graphics are now in a single die—what else are you going to do when you can cram over 25 billion transistors into a 250 mm2 chip? For reference, the entire 486 processor consisted of less than 2 million transistors in the late ’80s and early ’90s. That’s over a thousandfold increase in about 35 years.
Before graphics processing was moved into the CPU package and die, it was first integrated into the chipset. Where today we normally have a single chipset for motherboards, we used to have two chips: the northbridge and southbridge. The northbridge housed the memory interface, which made it the ideal place to integrated graphics functionality, and that’s what Intel did with its 810 chipset in 1999—that’s officially generation one of Intel Graphics, if you’re wondering.