FEATURED G.Skill Trident Z 2x8GB DDR4-3000 CL14 - F4-3000C14D-16GTZ

[Woomack]

Lucky you. Most kits will throw errors at 12-12-12 above ~3700-3800 regardless of timings or voltages ( still will pass benchmarks ). Maybe it's because your kits are 4133 and are better binned. I couldn't make any of by "B" ( 3000-3600 kits ) run below CL14 @4000+ and below CL13 @3700+ stable on ambient temps. I could run 3866 12-12-12 in XTU, Spi32 and 4000 in some other benchmarks but hard to call it stable.
However I hardly doubt it's 'rock stable' at 4000 C12 as 64b Windows is crashing when you address more memory at higher frequency and tight timings. All are using 32b OS with limited memory size to run benchmarks without issues, not to mention full stability. Whole top of the XTU ranking is like copy/paste settings on almost the same memory kits where some kits can make 3700 and some will pass 4000.

What you can make in benchmarks doesn't mean is fully stable and I know you understand it.
Your results are exception which is not happening often and most users simply can't count on the same settings. Btw. nice results in maxxmem, I saw it couple of days ago.

I'm also not sure what is inaccurate in what I said. This thread is dated at Feb 2016 so some info can be outdated already. If you mean the info about 3700+ kits on Samsung B IC then when I started this thread then max clock available in stores was 3600 CL16/17 ( 16 just appeared back then ).

Of course feel free to make a thread about your memory kits. I doubt I will get one of the 4133 kits while some forum members can find it interesting. The only kits which I have now for tests are 3000 Kingston/Hynix and 3000 Crucial/Micron so I won't show anything special. Maybe will add a thread about 3200 CL14 TridentZ but I just sold Z170 platform so it has to wait some time till I get something new.

 

[Fasttrack]

First of all, I did not notice that the initial post was back 02/2016.
Rock solid at 4000 is a fact if maxmem is set at 4000 on Windows X 64.
Hours and hours of benching is possible without a slightest problem.

I am not suggesting that these kind of super extreme settings are for everyday use.
Neither appropriate for all benchmarks.

After week 37 all trident Z memory modules are single sided B-die.

I must also note that these kind of settings are achievable only under specific conditions :

1) Two motherboards
a) AsRock M-OCF
b) Asus Impact VIII

2) A processor with very strong IMC, preferably under cold ( especially above 1.95Vram )

Thank you

 

[Woomack]

For me anything 4000+ was possible only on ASRock OCF or MSI Z170I Gaming Pro AC while MSI has some limits in voltages so it's not the best for high clock/tight timings at the same time. I still reached ~4500 memory clock on that board and 4200 at more relaxed timings was fully stable ( there is screenshot in Trident Z 3600 thread ). I had no chance to test Impact ( I was testing 3 other ROG boards ) but MSI was 50% cheaper and is easier for memory overclocking ( at least on ambient temps ) so I got it as 2nd board.

ASRock M OCF seems the best for memory overclocking but since I had full ATX OCF already then didn't want to spend money on one more motherboard. Right now I don't even have Z170 as I had a chance to sell both motherboards and CPUs so I just did it and now I wait for Z270. I hope I won't have to wait long.

DIMM voltage is separated and has nothing to do with IMC. However cache and some other things are related to memory stability and higher memory bandwidth/lower access time has to be stabilized with higher voltages ( cache etc ) or lower temps. Personally I haven't seen any special differences with CPU on cold but I don't remember if I was trying to set max clock and tightest possible timings when I was benching on cold.
In most benchmarks memory performance is barely affecting total score and because I'm limited to ss/dice then I don't really push for the best time/score in benchmarks like Spi32. I have really average luck to CPUs and I simply never get good results. Well, I have good results for ss/dice but it's far from what guys on LN2 can do.

 

[Kilgur]

Got same memory F4-3000C14D-16GTZ and got problem Why in aida64 test my read/copy is so low on 4000+ im using Woomack settings from post 2.

 

[Johan45]

Check tRDRD SCL and tWRWR SCL if these are too high on auto then it can affect those scores in AIDA. Try a setting around 2 or 3 if it's stable

 

[Alaric]

Got same memory F4-3000C14D-16GTZ and got problem Why in aida64 test my read/copy is so low on 4000+ im using Woomack settings from post 2. View attachment 198515 View attachment 198516

Every RAM set/stick and motherboard has their own "personality". Woomack's results are a good place to start fine tuning your RAM, but no guarantee you'll get the exact same results. You'll have to work out the best combination of speed/timings for your rig.

 

[Woomack]

Hard to compare that since my results were on Z170 and I don't even remember additional settings. Some of my results are with auto sub timings, some are with tweaked ( depends on what I want to show or how much time I have ).
I will be testing G.Skill TridentZ 4266 CL19 and Team Xtreem 4133 CL18 on MSI Z370 soon. Both are on the same Samsung B, maybe higher binned, will see when I find time to test them. Right now I'm busy with some other hardware.

 

[Kilgur]

So cuz my last OC exp was 15 yrs ago, and im bit lazy, i used google to find any info for OC memory on asrock z370 I found a movie on YT by Kapteeni Kuolio:

Around 31:30 he talking about this write/copy drop, and how fix it Anyway, I copied some of the 2nd and 3rd timings, and got good results
In spoilers there is my time and result in aida64. ASRock Timing Configurator is rly nice. Anyway before i go forward with OC, I have to read the links under this video, cuz im rusty in OC

Spoiler

 

[Woomack]

I'm glad you figured it out.
I had 5 mins to run one quick test. Here is XMP@4266 on MSI board and TridentZ RGB kit.

 

[Kilgur]

I do not know why I have lower test results in MaxxMEM2 v1.99. Maybe because my system is windows 7

 

[Woomack]

Don't even use maxxmem. It's showing wrong results since it was released. Can run test multiple times and have +/-20% difference in bandwidth. Also on new motherboards it almost always have problems with latency what can be related to that bandwidth differences.