Maybe Infinity Fabric is better glue than whatever Intel does when more than one die is present on the same substrate? Just speculating on this one. Infinity Fabric is baked in to the core Zen design... even on single CCX chips, it still gets used for the cores to communicate with the separate chipset portion of the processor. Intel puts everything on a single piece of silicon so you don't need an interconnect system... until you make a chip with multiple dies. Their system for this, I'm assuming, is largely inferior to AMD's implementation because until very recently nobody was buying processors with heaps of cores unless they were in the server space, where latency and per thread performance wasn't as important as having a ton of cores. Thanks to AMD, the core wars have now truly begun, as well as looking for the most performance per core we can get... at least outside of the server space. AMD started this war, so they have a leg up at the moment. It's going to be interesting to see what Intel comes up with once they get out of the lake, and it will be equally as interesting to see what AMD has at that time.
Intel is really in a sad state at the moment. Their 10nm issues are truly unfortunate, I'll give them that, but it seems clear to me they got too comfortable with their lead over AMD when I think about how many iterations of Skylake we've seen at this point. The Zen architecture, and the growing list of *lake security flaws have been around for a while now... and so far we've been shown nothing more than more *lake chips. I don't think they had any plans to move on from *lake any time soon...
None of this is so simple as intel is on a downtrend and amd is movin' on up. It has been a good year for amd and intel continues to make strong revenue. Getting into the platforms and products each company offers we are looking at a highly competitive time. That means that it is unclear to most consumers that either amd or intel is best and purchase decisions are made based on the way the consumer expects to use the hardware.
Nobody can predict the future and a new computer often delights in unexpected ways. Hence, most of us are now using our desktop PCs for things we never did on our past PCs. For this reason, I believe any new PC should be purchased to meet the excepted needs and with an open mind about how the computer might succeed in generalized usage, now and out perhaps 3 years into what we imagine to be the future.
One might argue that a new Ryzen has the edge in future usage scenarios because we all expect the future to better utilize high cpu core counts than hardware does today. But is this the future of the PC? If you've been watching Optane develop, then you are aware that future computers may do away with the notion of memory and storage being separate hardware. The future of computing could be that all files are kept in non-volatile memory, always available with no need to ever boot the computer or open a file. That could be a lot better than a huge core count for most users.
There are steps that can be taken to make intel cpus perform better and amd cpus are generally running flat out from the factory. So if you like overclocking, intel might seem more attractive.
Most people do not do massively multi threaded avx instructions in any of their workloads. Video conversion in Handbrake would be an example of a task like that. For these tasks, amd does hold an advantage with higher core counts. But even here, a thorough overclock of cpu and ram on intel can make up a lot of the difference ... i9-9900k really perks up in the hands of an experienced overclocker. Like 40% faster in Handbrake, for example.
Stop fearing there are bad choices out there. This is a time of innovation in desktop computers. Most any new system is going to blow you away with fast transfers between storage devices and snappy response while you do all those tasks that used to take longer on your old PC, Ryzen or Core either one is sweet.
I've an i9-9900k at 5GHz all cores, avx instructions (prime95 small ffts) with low latency ram (c16, 1T command rate) running @ 3866MHz. It is water cooled and all the storage devices are M.2 NVME ssds. It does a lot of video conversion and overclocking reduced the time to convert 20GB of 1080p MKVs to MP4s from 89 minutes to 53 mintues. It's clear evidence that you can increase multi-thread performance on the i9-9900k by 40%. But buying a Ryzen 3950x would be just as good, cost less and take less time.
I like overclocking and would never be happy with a Ryzen. That's me.
