PCIe only benefits NVMe type stuff at the moment, even a 2080Ti doesn't saturate PCIe 3 yet. It's a buying point, but like RTX its practical use is near zero at present.
Intel has nothing to release on desktop except a rumoured 10 core Comet Lake 14nm refresh. Their binning is topped out so they can't get better clocks on 14nm which leaves adding cores; while losing performance to security mitigations. 10nm? Do not hold your breath (rather long winded explanation as to why below...)
They've been promising 10nm since 2015 (and volume since 2016). The 10nm SKUs for Ice Lake and the single Cannon Lake SKU are laptop only as they have the double whammy of poor clocks and bad yield- and that on their lowest density 10nm setting as well, which is actually less dense than Samsung/ TSMC existing 7nm bulk process already is (Intel has three 'density' sets for 10nm, 2 are better than others' 7nm but only the worse one is working). Poor clocks don't matter on laptops as the chips are undervolted/ clocked anyway to save power and even the Ryzen models have few cores, but they're deadly to desktop performance where thermal constraints are a lot less. To get viable desktop chips they need better yields/ more working cores, higher clocks and the higher density 10nm to be working as otherwise they will be worse than the chips they already have. Even if they have 18% better IPC that is performance per clock; a 30% drop in clockspeed completely removes that advantage and then some. The realistic estimate for desktop 10nm is 2021, and perhaps the super realistic estimate is no 10nm desktop chips at all or a very limited release like the 5775C was for Broadlake, with a kludge like the 5775's expensive on chip EDRAM to boost performance. And, rumour is that Intel's 7nm process was built on 10nm, so they may not even have anything to skip to without fixing 10nm first. Only good thing from Intel's perspective is that at least desktop is a small part of the overall market, but without good yield they can't get 10nm into the extremely high margin server market either as they need lots of working cores for that even with low clockspeed. And that is why they took the rather extraordinary step of firing their CEO publicly with no benefits; not because he was boffing his secretary as stated.
Ultimately, Intel royally screwed the pooch by tying architecture improvements in their CPUs to specific manufacturing processes; so they simply cannot make Sunny Cove based cores on the 14nm process as might be a sensible stop gap as they will only work with the 10nm process, and their 14nm chips are also stuck with awful integrated graphics 2 gens behind what they could produce on 10nm, if it were working.