I have been thinking about my hardware future and the idea of SLI/crossfire and asking myself if it's worth it.
I'm an AMD guy but let's take the GTX 1080 as a reference point because my target is the high-end. And let's look at two GTX 1080s in SLI.
The question is, how long would this setup last and still be competitive with future high-end GPUs?
The yearly improvements of GPU power have slowed down. Nvidia released their first 28nm GPU, the GTX 680 in March of 2012 (and the GTX 780 in the next year) but it took 2.5 years until they released their last 28nm GPU with the GTX 980 in September 2014. The GTX 980 is about 75% faster than the GTX 680. And it took nearly another two years until their first 16nm GPU, the GTX 1080 came out in May 2016, 4 years and 2 months in total.
So it's not unreasonable to expect that it could take, very roughly and pure speculation, a similar amount of time of 2.5 years, for a future 16nm high-end GPU to come out that is roughly 75% faster than a GTX 1080. Let's hypothesize this. We could call this GPU the GTX 1280.
And since two GTX 1080s in SLI are 50-65% faster than a single GTX 1080, we can conclude that they would be almost on par with the hypothesized GTX 1280 (a 75% increase versus 50-65%). The maths give us that a single GTX 1280 then would be only 6-17% faster than two GTX 1080's in SLI!
Which means that it would be almost pointless to upgrade even in November 2018 (2.5 years after the release of the GTX 1080 in May 2016), which in turn means that I wouldn't have to upgrade until beyond November 2018, perhaps not until the GTX 1380 comes out!
I know the problems with SLI, microstuttering, uneven scaling, some games not even supporting SLI, increased power consumption etc. And you're welcome to take these matters into consideration.
Thoughts?
