I have a lot of confidence in them dethroning intel this time around.
I don't. We've seen certain patterns in Ryzen and its successors. The specifications that are rumored here only confirm my theories on what to expect from a Ryzen 4000 series. As we know, despite Zen2's radical departures from Zen and Zen+, it's still iterative. Thus, it retains a lot of the drawbacks of its progenitors despite overcoming some other key flaws. I have every reason to believe this trend will continue. I don't see Zen3 getting such a massive IPC increase that it can overshadow Intel's slowly growing clock speed lead in single threaded applications. I also don't really think Ryzen 4000 will behave better than Ryzen 3000 in terms of boost clocking behavior.
That is to say, I think 4.8GHz is likely wishful thinking and you won't see it very often. The Ryzen architectures simply haven't clocked all that well and while generally efficient, they lose this efficiency quickly when the clock speeds ramp up. The small size of the die also gives them poor heat dissipation for what they are which is a key reason why they can't be brute forced into higher clock speed ranges the way you can on the Intel side. That and Ryzen binning seems to be at the edge of what the silicon is capable of anyway.
Also keep in mind that sticking with backwards compatibility on the platform means that certain design decisions with the CPU have to be made to maintain that compatibility. AMD can't simply change the relationship between its Infinity Fabric clocks and the memory clocks so easily. Aspects of the CPU's internal layout will be predetermined due to socket compatibility which can effect how the silicon performs. An example of this is the relationship between the Infinity Fabric and the CCD's and thus, CCX complexes within them. While they all have virtually equal access to the memory controller and Infinity Fabric, there is a lot of latency baked into that arrangement. You don't have the same cross CCX boundary issues or performance penalties associated with earlier NUMA architectures that we saw in previous Ryzen CPU's, but penalties still exist. AMD chose to mitigate this with a large amount of L3 cache, but as we've seen this only goes so far.
I do think we will see marked improvements in gaming performance, but I doubt it will reach parity with Intel in that regard. There are a number of factors behind my stance on this. I think Ryzen's basic design will always carry latency issues within them that, while mitigated largely, will never allow parity with Intel. Combine with that the fact that Intel can scale its clock speeds higher and that the core count advantage AMD enjoys doesn't come into play with gaming, and Intel will still have a slight advantage in that area. Intel has other advantages that people seem to forget about or underestimate. Intel has SDK's and compilers that are entrenched in the software industry and due to their presence in the corporate PC world, most games end up optimized for Intel CPU's specifically. Even some applications still run faster on Intel hardware or at the very least, remain strangely competitive despite having a core count deficit compared to AMD.
We can see examples of this with some of the games I routinely use in my CPU and motherboard articles. Destiny 2 and Ghost Recon Breakpoint. The former was never tested on AMD hardware and suffered horribly for it at the launch of the Ryzen 3000 series CPU's. Destiny 2 flat out wouldn't run on them and until a proper fix was issued via AGESA code updates, the performance using the work around was outright terrible. Intel still often enjoys better minimum frame rates here, although this advantage has waned as more and more security exploit mitigation ends up being baked into Windows updates. Ghost Recon on the other hand was developed on AMD hardware and as a result, it shows no advantage towards Intel and all of the top CPU's show virtually identical results across the board. We see this at both 1920x1080 low and 3840x2160 at high settings disproving any notion that 4K is simply GPU limited.
That said, we don't know the details of Zen3's architecture, so it could surprise us all. However, giving us the same kind of increases we saw between Zen+ and Zen2 again would be an impressive feat in itself. One that seems unlikely, but I suppose it is possible. One final thing to keep in mind is that AMD doesn't really need to match Intel on the gaming front. AMD's already winning in the price vs. performance arena and in every area besides gaming. Intel has to have both a clock speed advantage and a core count advantage to match AMD right now. This is something that costs a lot more for Intel to achieve. Thus, while some of Intel's 8 core CPU's are often better than AMD's at many things, they cost a lot more as well. AMD also wins in the total core count category where Intel hasn't a prayer of matching AMD.
Outside of a few niche areas like 240Hz and 4K gaming, AMD is just as good. Most people would argue that AMD's gaming performance is more than fast enough at 1080P and the data backs this idea up. At 240Hz, AMD simply isn't fast enough today, but it's a niche market and Intel can't always achieve the desired frame rates either. Lastly, at 4K, you are often still more GPU than CPU limited, making it a moot point. Where it isn't, Intel still wins by a good margin from what I've seen, making Intel the way to go. But again, how common is 4K gaming compared to 1080P / 144Hz or 1440/120Hz? It's a niche at best.
