Which is better, DDR4 at 3000MHz CL15 or 3200MHz CL16

[Analog]

I finally got my DDR4 quad-channel kit running correctly, but after some more tinkering I managed to bring it down from it's default 3000MHz CL16 rating to a CL15. I also found ot that if I keep the CL16 I can do 3200MHz no problem. In both cases I keep the voltage at stock - 1.35V. 

 

So, as the title says, which would be better:

 

- 3000MHz with CL15 (15-15-15-35 -1T) 

 or

- 3200MHz with CL16 (16-16-16-36 -1T)

 

I tried to find some info on this, but couldn't get a definitive answer on whether there is actually any difference in terms of performance. 

 

[spartaman64]

probably the higher mhz

[ZetZet]

yeah pretty much no difference, imo. get more MEGAhz just cuz MEGAhz.

[Glenwing]

There won't be any difference in performance, but theoretically the 3200MHz will be better. The latency is exactly the same in either case (5ns).

[ProKoN]

I finally got my DDR4 quad-channel kit running correctly, but after some more tinkering I managed to bring it down from it's default 3000MHz CL16 rating to a CL15. I also found ot that if I keep the CL16 I can do 3200MHz no problem. In both cases I keep the voltage at stock - 1.35V. 

 

So, as the title says, which would be better:

 

- 3000MHz with CL15 (15-15-15-35 -1T) 

 or

- 3200MHz with CL16 (16-16-16-36 -1T)

 

I tried to find some info on this, but couldn't get a definitive answer on whether there is actually any difference in terms of performance. 

 

in the real world your not gonna notice any difference. if you use winrar  lots go with the faster 3200mhz speed (who knows maybe your the dude that rar's up my newsgroup DL's)

 

if they both require the same voltage then take your pick

 

run some aida64 cache\ memory benchies to see how they perform at different speeds\timings or maxxmem

[TriceraFLOPS]

I finally got my DDR4 quad-channel kit running correctly, but after some more tinkering I managed to bring it down from it's default 3000MHz CL16 rating to a CL15. I also found ot that if I keep the CL16 I can do 3200MHz no problem. In both cases I keep the voltage at stock - 1.35V. 

 

So, as the title says, which would be better:

 

- 3000MHz with CL15 (15-15-15-35 -1T) 

 or

- 3200MHz with CL16 (16-16-16-36 -1T)

 

I tried to find some info on this, but couldn't get a definitive answer on whether there is actually any difference in terms of performance. 

I don't know what you're testing with, but when you find the speed you like, I'd do a few runs of memtest86 to verify stability.

[Analog]

run some aida64 cache\ memory benchies to see how they perform at different speeds\timings or maxxmem

 

This is actually a very good idea. I will do that! Thanks!

[ProKoN]

This is actually a very good idea. I will do that! Thanks!

would you be kind enough  to post your results?

 

I did some tests of my own out of curiosity and for the sake of science. im curious to see how much ddr3 differs from ddr4 in this benchmark.

 

I tried to keep cpu as close to 4.7 as possible with cache at 1:1 ratio

 

im using this kit here      http://www.newegg.ca/Product/Product.aspx?Item=N82E16820104347&cm_re=2666_predator-_-20-104-347-_-Product

 

ran 3 tests

 

2800 cl 11

 

3000 cl 12

 

3200 cl 13 (well almost 3200)

 

 

 

 

 

 

[Analog]

would you be kind enough  to post your results?

 

I did some tests of my own out of curiosity and for the sake of science. im curious to see how much ddr3 differs from ddr4 in this benchmark.

 

I tried to keep cpu as close to 4.7 as possible with cache at 1:1 ratio

 

im using this kit here      http://www.newegg.ca/Product/Product.aspx?Item=N82E16820104347&cm_re=2666_predator-_-20-104-347-_-Product

 

ran 3 tests

 

2800 cl 11

 

3000 cl 12

 

3200 cl 13 (well almost 3200)

 

 

Yeah, sure, I will post some results, but I didn't save the ones I made yesterday so I will have to run the benchmarks again when I get home from work.

 

I remember, thought, that actually 3000MHz at CL15 was a a bit faster than 3200MHz at CL16. There was like 0.5-0.6ns difference in favor of the former. 

[Analog]

@ProKoN

 

I did some more testing and benchmarking and I found some interesting stuff. I wasn't able to replicate yesterdays result with the 3200MHz overclock, but I did some other stuff. Keep in mind this is not a motherboard-certified memory kit. For my board there are only two kits of and over 3000MHz that are certified to work. 

 

CPU: 4.4GHz

Cache: 3GHz (Stock)

RAM: 3000MHz 15-15-15-35

 

 

CPU: 4.4GHz

Cache: 3GHz (Stock)

RAM: 3000MHz 16-16-16-36

 

 

CPU: 4.4GHz

Cache: 3.5GHz

RAM: 3000MHz 16-16-16-36

 

Interestingly, I cannot get the memory to run at CL15 with the cache overclocked. I played around with the system voltage, but it didn't really add to the stability at all. That was also something that I noticed yesterday when I managed to boot at 3200MHz speeds, even though I kept the RAM at CL16. 

 

 

 

The last preset feels quite a bit more snappy in Windows compared to the others. Despite that the differentiation in the latency is only marginal. 

 

PS: Since I am bit new to memory overclocking, could you explain what exactly is the Command Rate and how it impacts performance. I am interested, because my motherboard has it pre-configured to 2T, but it is impossible to get the memory to run at 3000MHz without setting it at 1T. Also, at 1T Command Rate all programs seem to load considerably faster. Also the Windows turns on and off more rapidly. 

[Frankz]

15-15-15-35 -1T are pretty nice memory settings at 3000mhz, what voltage are you using for it?

[Analog]

15-15-15-35 -1T are pretty nice memory settings at 3000mhz, what voltage are you using for it?

 

I am using 1.35V

[mariohead123]

yeah pretty much no difference, imo. get more MEGAhz just cuz MEGAhz.

Ouch!

[ProKoN]

@ProKoN

 

The last preset feels quite a bit more snappy in Windows compared to the others. Despite that the differentiation in the latency is only marginal. 

 

PS: Since I am bit new to memory overclocking, could you explain what exactly is the Command Rate and how it impacts performance. I am interested, because my motherboard has it pre-configured to 2T, but it is impossible to get the memory to run at 3000MHz without setting it at 1T. Also, at 1T Command Rate all programs seem to load considerably faster. Also the Windows turns on and off more rapidly. 

 

that bandwidth is pretty epic. The quad channel is really shining in the benchmarks. funny you say you can notice a difference with the different latencies. others report the same thing. I personally cant tell the difference between ddr 1600 cl9 and ddr3 2400 cl9.

 

im no expert in memory overclocking.  command rate is basically a latency setting. as with all latency settings the lower the value the "quicker" your memory will perform. 1T is the most desirable command rate value so keep it there.it was more of a thing back in the ddr2 days where the difference in command rate may equate to marginal performance gains\losses. with ddr3 its not even within margin of error for the most part aside from synthetic memory benchmarks which are kool and neat but really i dont think are indicative of real world performance for the most part.

 

really good read. thank you for posting your results. if your ever bored take out a couple mem chips and run some dual channel results

 

your 20mb cache + ddr4+ quad channel is hauling ass

[Analog]

@ProKoN

 

I did some further testing just for the fun of it. I had to back the CPU overclock down to 3.6GHz, the BLCK to 120 and to put the CR back to the default setting of 2T, but I did manage to get to 3200MHz, while still running the RAM at 1.35V. 

 

CPU: 3.6GHz

Cache: 3GHz (Stock)

RAM: 3200MHz 16-16-16-36-2T