What I object to is the deliberate use of 2 different scales on the X-axis, which is
designed to mislead the perception of the casual viewer.
What they've done in this case is to make, for example, the 57% improvement "look" like 570%, and the 21% "look" like 210%, etc...
You often see politicians/economists blatantly "truncate" one of the axes in order to exaggerate the appearance of a trend, and it p**ses me off whenever I see this sort of trickery.
A 57% improvement is still a good improvement, not to be sniffed at, and they would do better to represent it accurately.
IMO, attempting to mislead customers by misrepresenting the data is not good business, and may have a backlash in terms of customer belief/trust in other claims made by the company.
For example, I am currently considering a new laptop, and my choice of graphics manufacturer is to a large extent determined by how much I "trust" their product (because it's such a vitally important component of the computer).
Anything that undermines my trust in the company (eg. marketing tricks) may also undermine my trust in their product.
I would like to think that most people these days are savvy enough to not be fooled by this sort of trickery, but sadly, the fact that marketers continue to employ these tricks suggests that they still work on a lot of people.
C'mon Nvidia, we are not stupid, and we expect better than this from you!