This is a simple question that doesn't have a simple answer. I'll try to make it simple and we can go to deeper details if you want.
So... every game does a lot of calculations. Part of those is for disk operations, memory management, caching, pre-fetching textures, part of those is for managing audio (or positional audio), part of those is for physics, part for game engine (much more if it is a MMO game due to more complex tasks to do), then goes AI and in the end of this list comes graphic engine.
With every generation of graphic API (be it Glide in the far past, OpenGL, DirectX, Vulcan etc.) more things from cpu part of the equation can be effectively moved from cpu and being computed on gpu, but it can hardly be all of it. Now if you imagine all those non-graphic parts beating your cpu (and some other components), we come to graphics. It eats a lot of resources even while majority of the work is done on graphics. Now imagine that for cpu there is no difference whether you play 1024x768 or 3840x2160 - there is still same scene graph that needs to be processed and rendered - same volume of triangles that need to be transferred between cpu and gpu... and here it comes - either your gpu can process that volume faster than cpu can feed it, or it can't. If your gpu is work hungry, either the graphical engine doesn't need more or your cpu has reached limits of what it can provide in the given time. There you get bottleneck on gpu or cpu.
In the past, games were strictly single threaded. Well, maybe you had audio processing running in separate thread, but nothing worth noting. This was due to neither API nor chips themselves were able to effectively (or at all) work with concurrency. With newest API and gpu chips these limits are getting gradually lifted and then you can get more work done on cpu side. But, it puts more stress on game development process since proper multithreading is not trivial if it is to be done properly (risk of deadlocks, starvation, issues with synchronization of data between threads etc. - it can do for some decent headaches).
And here we go to the last part of the equation... After all that hardware and API improvements, it comes to game developer and it purely depends on him, how he utilizes the given resource budget. It can be done well and then majority of bottlenecks with really lay on the slowest piece of silicon, but it can be more of an issues of developers not making the optimal choices... or, not making optimal choices for your particular setup (imagine "strong cpu" and "weak gpu" and a game designed to calculate as few as possible on cpu and as much as possible on gpu... won't really give you optimal results).
In the end, we get to i5 vs i7. The difference between the two in majority is a presence or absence of hyperthreading technology (in Intel's implementation this means two logical cpu sharing one physical core). With HT turned off, you are pretty much (not entirely, but close enough) to having i7 = i5. With HT turned on, you can obtain some free computing power, but those additional logical cpus have to share the physical silicon. Which can be done quite effectively (simply by using different code paths in cpu to do different things and not competing with itself for own resources), but there is still some cost resulting in slightly lower performance of single core. If your game is (almost) single threaded, your performance can even go lower compared to i5 or i7 without HT. But with every cpu generation HT works better than it used to. Further, modern OSes first utilize physical cores and only when they run out of available computing power they start to schedule on "secondary" logical cpus... Modern games still mostly use single thread "with sauce", so i5 vs i7... nothing gained. Absolutely newest games or games that need significantly more cpu computing (MMOs, we can get to reason why later if you want), can already utilize more cores, but it hardly gets up to 4. In future we can expect utilization will gradually increase, but it won't be fast - developers need to design their games so the majority of gamers can actually play it well, and these days, majority of gaming PCs are 4 core machines.
I can't imagine a common gaming scenario where your 4690k wouldn't be able to feed GTX-1080, let alone GTX-1070 so unless you do some cpu heavy computing, I'd go for i5 without worries.