Strange Bench Results From Gskill. Higher CAS, better Bench?

[GenieGOR]

This isnt an issue. I was just curious as I have 2 sets of memory here. Both are G.Skill TridentZ. Just wanted to see what you guys thought kind of cool since its $30 cheaper too.

1. 3200 16-16-16-36 Part # F4-3200C14D-16GTZ

2. 3200 14-14-14-34 Part # F4-3200C16D-16GTZ


I ran the aida64 memory bench and got slightly faster results on the 16D memory. Of course the latency is better on the 14D. However when overclocking memory to 3600mhz its probably never going to see a cas latency of 14. So both memory sets are perfectly stable @ 16-18-18-38.

I was just curious as to why the higher latency memory has more bandwidth when overclocked. First picture is the 14D, Second picture is the 16D.

 

[Lochekey]

I see that the cpu speed was different between the 2 tests was there any other variance in the system as well. If you want Ann apples for Apple's comparison I would 're run the tests and double check your settings.

 

[GenieGOR]

I uploaded the right photo. I did the test 5 times each set. I only saved the test 2 times. CPU clock and all paramaters were Identical. Unless the mobo changed something itself since a lot of settings are on auto.

 

[Kenrou]

Its doing the same on my rig, atm have the cache/ring clock on auto 4.1ghz but every now and then it starts up at 4.5ghz or 4.6ghz. Mine also has slightly different results even though its at 3200 14-14-14-34 1T, but it's more or less on the range.

 

[Woomack]

@ GenieGOR, Cache clock is much higher in 2nd case. There you have the difference.

3200 CL14 kit can probably work at 3600 16-16-16 1.35V. I have 3000 C14 and 3600 C16 on the same IC. Both can work @4000 17-17-17 or something like 4133 18-18-18 ~1.35V. Some results here -> http://www.overclockers.com/forums/...dent-Z-2x8GB-DDR4-3000-CL14-F4-3000C14D-16GTZ

Also 3200 CL14 is single rank kit while 3200 C16 is dual rank kit. There are differences in internal transfers what make dual rank slightly faster at the same settings ( not much but sometimes it's visible in tests like AIDA64 ).

 

[EarthDog]

Cache clock is the difference, yep!

 

[GenieGOR]

Cache clock is the difference, yep!

So these two different memory modules are identical then at the same timings and speed right? I wonder why the CAS14 is more expensive than the CAS16 if they are both going to be timed to 16-16-36 anyway as I imagine 14-14-34 would be impossible @ 3466/3600 would it not?

 

[Woomack]

I guess it depends from delivery date or store. 3600 should cost more than 3200 or 3000. There is also new 3733 17-17-17 on the same IC and some higher kits but I'm not sure if there is exactly the same IC or something else.

My 3000 CL14 and 3600 CL16 are working at 3600 15-15-15 1.35-1.40V. At higher voltage you can set them to 14-14-14 or 13-13-13. 13-13-13 and 12-12-12 may cause stability issues what is related more to platform than memory itself. Simply you may see error window that Windows couldn't address memory. Still can use these tigher settings for benchmarks.
3200 should be stable at 13-13-13 and maybe even at 12-12-12 but higher frequencies may cause issues as I mentioned.

On X99 I was able to stabilize both kits @3466 14-14-14 and for benchmarks at 12-12-12/13-13-13.

 

[GenieGOR]

I didn't know that 14-14-14-34 was possible @ 3200+. I have that ram here and also 16-16-16-36 and it defaults to 16 @ 3200+ xmp. Should I be running manual? I also thought 1.35v was the max for ddr4 ram?

 

[Kenrou]

My ram stock - http://www.gskill.com/en/product/f4-3200c14d-16gvk

 

[Mattr89]

My ram stock - http://www.gskill.com/en/product/f4-3200c14d-16gvk

View attachment 177303View attachment 177304

Nice. Less latency. How can I tweak my tridentZ to achieve 3600 @ 15/16?