But that doesn't really prove anything in his case.
Because 3800 with 1900 IF are sometimes a whopping 200MHz less IF clock than with the 4466 run.
The relevant comparison would be 3800/1900 versus 4466/1900 to make it fair in his example. It is well known that the IF clock scales in areas where bandwidth is critical. Especially with the single CCD Ryzens, which artificially limit write IN to the RAM, this is of course useful.
Benches from Aida are also only of moderate value for copying, since the multi-threaded process runs and of course brings points, especially with dual CCD Ryzens, which then may not exist in real (games??) operation, because the OS tries to find the relevant one Keeping work on a CCD - which is the case with Win10 for example.
Up to now I have always been of the opinion that IF Clock brings the added value in almost all cases. At least that's what your linked post seems to confirm. It is only logical that you can achieve a little more here and there with the same high IF clock level and significantly more bandwidth with the RAM. What I don't get - where does the view that seems to be going around here come from, that everything that isn't 1:1 isn't good? Didn't AMD even publish slides before the Zen3 release, where the scaling of the RAM and IF clock rates in terms of internal latencies was discussed? I mean yes... With enough RAM clock you logically equalize the latency disadvantage again at some point due to the async mode... You just have to look at the price/effort. 5100 kits, as you use them, are likely to be very few... For me, for example.:eat: Already at 3450 the bunch won't boot anymore. Fully equipped with 4x16GB...
Schon bei 3450 bootet der Haufen nicht mehr. Vollbestückung mit 4x16GB...
