When to upgrade your PC, a golden rule?

After basically thinking aloud writing some comments on this Youtube video from BPS Customs, I thought I could elaborate further here on my own blog.

If you have a cell phone, a gaming PC, or if you are using or have used any computer for that matter, very likely you have experienced the harsh realities of tech obsolescence. Science and technology improve on a daily basis, often quite drastically. Give it just enough time, and you are left with a gadget that is only a few years old, yet newer gadgets are way more powerful, or support newer standards and protocols and connectors that yours does not, even though yours was probably not exactly "cheap" when you got it not that long ago. Sounds familiar?

There's nothing we can do about this, except to upgrade to some newer, more powerful gadget at some point in time, whenever we decide to do so. But when is it a good time to do so? Not always an easy decision.

Isn't your gadget/computer still powerful enough? Do you really need already that newer one? Is the newer one really that much different/better/faster/more capable? Can your current "old" rig not serve you well enough for some more time, before you drop all that cash for the newer stuff, which will for sure go through the same aging process quite inevitably anyway?

Tough call. Tough call.

Everyone can approach such upgrade decisions their own way. Each has his/her own interests, priorities, and most importantly, pockets. If you have money to burn, simply get the latest/best equipment you want whenever it becomes available or whenever you want, and done with it. No choice paralysis whatsoever :) But plausibly many people do not have deep enough pockets to adopt such an approach. Some others may want to use their resources in an efficient, sustainable manner. In any case, when to upgrade? Is there a golden rule?

Here is a rule I have sort of internalized to guide my own decision making, trying to optimize the utilization of my money, the usability of my existing rig, and yet staying with a current and well performing system.

Technology obsolescence aside, my golden rule is the following:

Consider upgrading only when you can get 2x the performance for the same price you paid last time.

Notice, such rule is not at all the same as suggesting to upgrade when you get similar performance for half the price. Those are two completely different things, and here's a concrete example why: right now an AMD Radeon RX 5700 XT graphics card (currently a best value in the mid to upper-range GPU category) offers about the same performance the "old" Nvidia GTX 1080 Ti did three years ago, at somewhat near half the price (considering european prices.) But there is simply no existing option right now that would offer you 2x the performance of a 1080 Ti for its original price. GPUs have not evolved that quickly. So whenever you get the same performance for half the price, you don't necessarily get double that performance for the same price.

Updates / corrections early Feb. 2020:

With respect to CPUs, the AMD Ryzen 3700X currently does provide about twice (in fact 2.1x) the multi-threaded performance of the now 4+ years old Intel i7-6700K, for about the same price, or even slightly less. Such an upgrade would perfectly exemplify the application of this golden rule --if you mostly cared about multi-threaded performance, that is.

Notice that the rule can be applied even if planning to jump up to a much higher performance class of equipment.

The AMD Ryzen 9 3900X CPU offers about 3x (more exactly 2.9x, according to PassMark's CPU Mark) the multi-threaded performance of the i7-6700K, but at a higher price, more exactly, 1.5x the price in the US market. So as of early February 2020, their relative performance/cost ratio is really 1.9x, close to but not 2x quite yet. (Cost of the 3900X would need to get to about or below $450 to match that 2x ratio.) Still a clearly beefy upgrade.

For the AMD Ryzen 9 3950X, the relative performance/cost ratio gets worse at 1.34x - 1.5x because of its much higher cost (2.2x) over the older Intel, while its multi-threaded performance is only slightly higher than the 3900X's (3.2x  vs.  2.9x). The higher a performance jump you aim at, the louder the law of diminishing returns will scream at you. That ~12% extra performance offered by the 3950X over the 3900X costs however ~60% more.

If your system had some different old parts, or if you are eyeing different new parts, the situation might be different. Also if you have some urgency to upgrade (e.g. you need support for some new standard, you want some new feature, or you want a better gaming experience just because,) then you could make that 2x smaller and to your taste, let's say 1.75x, 1.5x, or even lower? Up to you and your needs. This golden rule at least gives you a good framework to keep in check how much you would be spending, vs. how much or how little extra performance you would be paying for yet again.

Technology improves quickly enough, so I think it's not worth it to use inflation-adjusted costs when applying this golden rule. But as they say, your mileage may vary, so don't quote me on that ;)