Dual Channel @Z170A Chipset above 2400mhz

Hi everyone,

i write this post cause i am really running out of ideas and hope to hear about some similar problems someone solved. Duh.

A bit about me: i am working on computers since about 20y, overclocked my first 486 dx2 66mhz without knowing what i did (at this time), then later started again with "modern hardware" (Athlon Classic 750mhz to 1,1Ghz with Goldfinger Device), killed a few Athlon XP's, etc. besides this i work in IT since 15y. long story short: i believe i know a thing or two about hardware.

Anyway, it all started with the idea to build a new system, nothing overpowered, but since i was still working with my 6y old i5-2500K system i thought it is about time.

i spent a while with choosing some proper components and came up with this list:

MSI Z170A Gaming M7

Intel i7-6700K

Alpenföhn Brocken 2 (CPU FAN)

G.Skill F4-3000C15Q-32GVR DDR4-3000 15-15-15-35 1,35V - 4 Sticks, 32GB (4x8) - i know, you will ask yourself why i bought G.Skill and write in the Corsair Forum, i will get to that later

OCZ OCZ700MXSP (still had it from the old system)

MSI Geforce 1070 GTX Aero 8G OC

i mostly need stability in my systems, so overclocking the hell out of it was never the plan, but therefor i choose a bit higher default clock rates for the memory. The MSI Z170A Gaming M7 is (on paper) able to run the memory up to 3200mhz without any need to overclock the CPU.

So i started to build the new system, everything went well, system posts at first try, everything went really well. i read about problems with XMP Profiles, so setting the memory to those and get a post at 3000mhz right away made me certain - i like this system. few days later i thought - well, i should bench a bit and run the proper stability tests - before i even started with memtest86 and prime95 i got some strange errors on some of the benchmarks, pointing out the memory might be bad. to make sure i startet memtest86 and prime95. Prime95 had troubles right away, after a few minutes one process after another crashed. memtest86 also took only test #6 and hat thousands of errors in a "full run" (4 times). so i started to clock manually down (still at XMP Settings, but clock down to 2133mhz). went fine. i tweaked around a bit but nothing more then 2400mhz was stable (rest at XMP settings). I started to put more juice to the memory (since 1,35v gave actually only 1,31v), heard about the VCSA and VCIO tweaks that should help - but it didnt. i got desperate so i turned all relevant voltages close the recommended max. no changes, not stable at 3000mhz, but always posting. so i tested every memory stick in every memory slot - and was supprised. not only all the memory was good at JEDEC settings, they all where fine at XMP settings too - as long as i only tried one at the time. just to be sure i ordered another set of memory (Corsair CMK32GX4M2B3000C15 DDR4-3000 15-17-17-35 1,35V - 2 Sticks, 32GB (2x16)). Put them in, set the memory to XMP aaaaand nothing. Didnt even post. Again same tests - both memory sticks in any slot running fine at XMP. further i found out that the problem is attached to use of dual-channel. as soon as any configuration is set up that would activate dual-channel, the memory would fail at speeds above 2400mhz.

well, then it must be the board cause in the meanwhile i changed EVERY other component EXCEPT the CPU. till now i never ever had a defective CPU, so i assumed it could not be the problem. So i blamed MSI, still friendly and patient i opened a support ticket at MSI. 2 weeks later and no changes. i flashed a few bioses, cleared CMOS for my friends at MSI about 50 times on 5 different ways - no changes. i triple checked the pins at the cpu socket, nothing bend. the cpu cooler was not mounted to strong, when i checked first time i saw that my artic cooling 2 from the older days was not even spread over the cpu, and i really really was carefull cause i heard a lot about problems resulting from use of too much power at the wrong place. In the end MSI told me they cannot help me unless i change the CPU - what a bulls*** i thought to myself, went to the next store, bought a Asus Z170A Gaming Pro Aura, came home, changed the board, double checked the pins at the socket, build everything up, put in the corsair and fired up the system. again, first time post, everything fine, system boots, everything stable at JEDEC. then came the moment of truth - set memory to XMP, the Asus Board asked another question regarding some higher voltages for stability reasons - i said yes, cause i wanted to see it work. well. it didnt. same msg like the MSI board threw at me. too much OC, blahblah, unstable, blahblah, reset to default settings, blahblah.

ok... i thought. there must be a explaination. talked to my old friend aunt google, searched for "dual channel xmp no post" and found a few things, mostly things i checked already - "socket pins", "all memory slots good?", "all memory sticks good?", etc. nothing that really helped me to solve the problem. i bought 3000mhz memory for a board that says it can take it but it will not run stable like this. then i found a post from a guy who had a similar problem, cant find it again but it was pretty recent. one of the answers was that with the skylake Arch the memory controller went from the board to the CPU, i didnt knew that but that sounds like a lead. in the intel ARK i found a spec for the i7-6700K saying the memory controller supports up to 2133mhz in dual-channel mode. no f***** way - really?! come on, rly?!

well, and that is the end of my journey i had for the last few weeks. My question to you brave guys (i mean, if you still reading this my kudos go to you) who are still here is: is this the answer? is it luck (like chances to oc a cpu) how high the clock can go in dual-channel? since the memory is advertised with clock rates up to 4000mhz in special made memory kits for the skylake platform (Ripjaw 5, Corair LPX, etc.), with mainboards, advertised to manage memory up to 4000mhz - i just find that hard to believe. Do i have to be happy to get 2400mhz in dual-channel stable? i bought all of that hardware at amazon, i still can send back the CPU - but i really cant tell if the CPU is the problem. In Kaby Lake the Dual Channel clock is spec with 2400mhz btw.

i still have all the components here (except a spare 1151 CPU) and actually i dont care so much about the 1% performance difference i would expect in compare between 2400mhz and 3000mhz, but it is a matter of just finding the goddamn bug in here.

i am happy for any feedback. thank you.

update: changed the CPU to i7-7700k, gives me stable 2800mhz in dual channel. also changed the GTX 1070 Aero to a MSI GTX 1070 Gaming X.

maybe i ask the question other way around: anyone here got any kind of memory running with XMP enabled at 3000mhz+ in dual channel on a 6th or 7th gen i5/i7?

i had now 2 boards, 2 cpu's and 2 sets of memory, still cant reach memory specs (JEDEC ist fine, just XMP). do i miss something obvious here?

Besides that i changed the PSU already earlier in the process.

Back in the early days of Skylake i had no issues with a kit of Corsair Vengeance LPX CMK16GX3M4A2800C16 at their second XMP for DDR4-3000 and could even overclock them to DDR4-3200 and DDR4-3333 keeping the XMP timings with just a little bit of tinkering (manually adjusting voltages). Since then I had several other kits including some 2x8GB and 2x16GB models and tested most of them up to DDR4-4000.

In terms of memory frequency some board need manual adjustments earlier than others. While some of the better memclockers like the ASRock Z170(M) OC Formula should have no issues with XMP-Auto settings even in the DDR4-3866+ range.

Edit, my test setup: Intel Core i7-6700K, ASUS ROG Maximus VIII Gene