Let me quote myself once again:
I have never said timings are not important. In fact they are more important than the pure bandwidth for single CCD SKUs due to IF limitation. Second if you read through his article paying some attention you will see that:
So this 20 % AI Benchmark is not thanks to memory alone. The results shown are 9700X PBO + MEMORY EXPO vs 9700 Stock at 4800MHz RAM.
I get you really want to prove your point Igor, but now it turns into pure spam of hastily thrown links that are supposed to validate what you say. I will drop the subject now in order not to deteoriate the thread further.
65W TDP... pffft! my Asrock X670E board runs my 9700X with auto power limits for PBO well beyond that limit - 105W for example when running Mem Test Pro & this with a negative Vcore offset as well, my PBO has no tuning, all that on AGESA 1.2.0.0a which is the only
current AGESA Asrock implement for my board atm. MSI & Asus have released updated AGESA at the time of writing this so that promotes 105W TDP for Zen 5 but Asrock are yet to catch up to that besides Asrock are pushing the power limits of Zen 5 as it is. If I engage motherboard limits, the power limits are pushed even further, but so is performance however that is beyond the thermal solution I use atm.
Fabric clock sweet spot is 2000MHz, CCD to IOD link is 32B/c -> max bandwidth 64GB/s. If you are lucky you get 70,4GB/s with 2200MHz.
For DDR: 6400MT/s with 128b(16B) bus is 102GB/s. With current sweetspot of 6000MT/s you are at 96GB/s.
In other words, the new sweet spot will be meaningless to 1 CCD SKUs. Only 2 CCD SKUs will be able to benefit provided you can engage both CCD dies. When it comes to bandwidth, at least.
If 6000 MT/s is the so called "sweet spot" then why do Tech tubers like Tech YES City conduct
testing between 9700X, 7700X & 7800X3D with 6200 MT/s? are they cheating or something?
FCLK at 2000 is playing it safe for poor quality silicon chips thanks to the silicon lottery still a thing, this why AMD recommends it. The higher the FCLK can go with stability shows increases in memory bandwidth & reduced latency. Gaming benchmarks have proven fps increases with faster Memory up to a given point & that is IF the chip can do 6600MT/S 1:1 even on Zen 4. Your technical theories are wrong in real world applications like gaming.
Using a chip with 2 CCDs brings in the dreaded latency between the CCD problem & when running in 2:1 that is even worse as the memory controller is only operating at half speed. 1:1 is where its at for pure unadulterated performance.