I am posting this as an continuation of a conversation that started in the Z390 Motherboard main sticky - Actual start of this memory topic
 
Visit that thread in the link above to get caught up to this point
I had posted links to two videos put out by Buildzoid where he was talking about what to look for when making your DDR4 buying decisions and the second delved more into the settings when overclocking and what some of them did:
 
DDR4 Overclocking/Buying absolute basics
 

 
 
 
 
 Some RAM Bandwidth vs frequency and timings data...
 

 
 
 
 
Unfortunately for me Buildzoid put those out after I had made my memory purchase for the system I was building (shameless plug)
 
This is the memory Kit I bought - F4-3733C17Q-32GTZR
 
Going by the calculation Buildzoid proposes at about the 16min mark in the first video My kit comes out at a whopping 219
If I had known that and looked at the numbers from the kits available at the time I would have probably gotten the F4-3200C14Q-32GTZR which nets 228 instead.
 
The F4-4000C17Q-32GTZR comes out to be 235 and the F4-4133C17Q-32GTZR only comes out at 243, but still an improvement.
 
The memory I used for the above was taken from this page on the G.Skill website - G.S Link
I had skipped over the 3466 and 3600 modules because their timings were too loose and the numbers were crap
 
Edit note: Found my original calculation for the 3200 / 14 was incorrect in my notes - fixed the incorrect numbers and had to change up my statements / thoughts

Hawkeye360
It seems that RAM above 3200MHz doesn’t provide much benefit in everyday use and gaming. TPU did an article about RAM speeds and timings and concluded that 3000-3200MHz is the sweet spot.

PS: 3200/14 is 228 and not 246. 3200/13 is 246 though! :)

LOL I had found that on my own and was editing my post when you posted
 
But I do agree, that 3200 C14 kit seems to be a sweet spot for price / performance. If you left the timings alone and bumped the speed to 3400 it would be 242 on the calculation

Posting for reference
 
 

One run at all auto settings XMP profile at 3200Mhz (14,14,14,34 C2) and OC'd to 3800Mhz with a little looser timings and Re set to XMP 3200Mhz profile but forced Command rate 1 vs 2.

Adding 3800Mhz OC but with tighter timings. Could not get CAS13 to boot. And all runs where one or two runs....

icslowmo
One run at all auto settings XMP profile at 3200Mhz (14,14,14,34 C2) and OC'd to 3800Mhz with a little looser timings and Re set to XMP 3200Mhz profile but forced Command rate 1 vs 2.

Great score

i run typically 4000Mhz 16 16 16 36 on a maximums xi hero mobo. i was running 4100 but it was only partially stable.
 
Its been pretty much accepted that the 3200 Cl 14 ram is the best for some reason. You will see that all over the forums and all over youtube for that matter. Even some of the famous bloggers/youtubers use it as the "gold standard"
 
The problem is that ppl dont understand speed is more important than timings. Dont get me wrong you want both, but after seeing benchmark videos of different speeds tested i was shocked. games like shadow of the tomb raider getting 19 FPS better with 4000 19cas vs 3200 14cas. It's just really eye opening 
 
But you will continue to see everywhere that 3200 is best and that above that there is little gaming benefit. simply false

Hm... nice topic. I am currently looking for good RAM and thinking about to buy (if i can find someone who may actually sell it and ship worldwide):
 
1. Team Group Dark Pro "8Pack Edition" DDR4-3200 CL14 (it will give nice 8.75ns (1/1.6*14)),
but another option is this
2. Team Group Dark Pro "8Pack Edition" DDR4-3600 CL16 (it will give also nice 8.89ns (1/1.8*16)).
 
Delay between them no so huge (0.14ns), but frequency is 400 MHz, so i guess second set is more effective.
 
Last option is
 
3. Team Group Xtreem "OnePageBook Edition" DDR4-4133 CL18 (it will give abnormal 8.71ns (1/2.066*18)). This kit is rare and modules a bit high (47.7mm tall), so i am not 100% that they will fit in my build. Another problem is - frequency really high - i am not sure that it even run.
 
P.S. Motherboard is X299 Dark.

Oh Crucial...

“To calculate a module's true latency, multiply clock cycle duration by the total number of clock cycles.”

And they forgot to tell how to calculate that “clock cycle duration”.

Here it is: DDR4-2666 (for example)

1/1,333 = 0,75ns
1 is probably 1 second (i forgot);
1333 - is effective frequency of DDR4-2666.
1,333 because end result will be more accurate, instead of 0,00075...

ehabash1
i run typically 4000Mhz 16 16 16 36 on a maximums xi hero mobo. i was running 4100 but it was only partially stable.
 
Its been pretty much accepted that the 3200 Cl 14 ram is the best for some reason. You will see that all over the forums and all over youtube for that matter. Even some of the famous bloggers/youtubers use it as the "gold standard"
 
The problem is that ppl dont understand speed is more important than timings. Dont get me wrong you want both, but after seeing benchmark videos of different speeds tested i was shocked. games like shadow of the tomb raider getting 19 FPS better with 4000 19cas vs 3200 14cas. It's just really eye opening 
 
But you will continue to see everywhere that 3200 is best and that above that there is little gaming benefit. simply false

Good for a lot of benchmarking I think. Lower timings tend to get better results in most tests from what I recall. But as RAM speed is inextricably linked to Clock speed, you're going to find a lot of conflicting results as 1 test might prove one, while another proves the other, simply based on the particular set-up, motherboard, ratios, etc...
 
Don't mistake me, I'm not extremely knowledgeable; I've just seen a lot of different results over the years which tend to provide conflicting information if you just look at the numbers.