I have read about the newest AMD Ryzen CPU's and their impressive performance. But - They have only two memory channels which I find a bit peculiar. Is the total processing power/throughput not also depending on the number of memory channels? The Intel 775's had two. The 1366's had three, the 2011's four, So my question is if selecting a top performing CPU should besides clock speed and the amount of cache also be made upon the number of available memory channels. And furter, if looking for a CPU also should include support chips (bridges) that can handle registered DRAMs. Comments invited.
There are so many new things going on with this entirely new architecture--and so many variables--that a single spec like that is relatively meaningless when weighed alone. The only real assurances on performance (for a particular task) will be when benchmarks are presented comparing a Ryzen CPU to its Intel "equivalent." For example, core clock speed is not an accurate metric for comparing the two. The numer of memory channels can be offset by the type of memory supported (NVidia's entire 20-series lineup comes to mind), and in the case of this new PCIE4, by the number of lanes supported. So, again, looking at a single spec will leave you with an "out-of-context" conclusion. AMD's new CPUs might shine in some areas and simply match performance in others. So look more at how you plan to use the cpu as opposed to a single spec. Benchmarks will tell us how all architecture implementations work as a whole. If all someone is ever going to do is run Adobe Premiere then they would want Intel, as those chips have been heavily optimized for Premiere. On the other hand, for gaming, multi-core rendering and other tasks overall, the cost-to-performance ratio favors Ryzen...at least at current Intel prces. You are also likely to get more cores on Ryzen for similar Intel performance. But if comparing apples to apples (i.e., Intel to Intel), them more is better.
I am considering building a "workstation" since I do simulations that lasts for hours and that rely heavily on CPU/Cores and memory transfers. The problem with AMD is that only their Epyc processors seems to handle parity type memories. The Epyc MB's that I have seen (SuperMicro) are only certified for Windows server, not for "regular" W10. Besides, an Epyc with reasonable CPU speed is very expenceive. The Threadripper outshines Intel from what I (as a non pro) can see. Apart from that parity type memory issue I am inclined to try a Threadripper this time. (Have always liked underdogs and I am pleased that AMD is doing well as far as I have read in the magazines.)
to me the best way to test machine performance is benchmark testing playing high performance games thats when u know what u have.
Same with my 4790K. I can't speak on Intel, but I think with the new AMD offerings, AMD will reduce prices for previous generation CPUs. But what's problematic and profane is GPU prices. Pascal-generation video cards currently selling for $1,300.