cancel
Showing results for 
Search instead for 
Did you mean: 

Confusion in the selection of mobo and memory for Kaby Lake

wildapiospi
Level 7
Hello. I need mobo and RAM for i7 7700k but I have a problem with the choice.

Which motherboard for overclock? What do I need to consider in choosing? I think about Maximus CODE for 5 GHZ Kaby Lake. Good choice?


The hardest choice 😕 - RAM. G.Skill Trident Z DDR4 2x16GB 3466MHz, CL16 (F4-3466C16D-32GTZ) vs G.Skill Trident Z DDR4 4x8GB 4000MHz CL18 (F4-4000C18Q-32GTZ), what is better? dual channel or quad for kaby lake?



P.S. M.2 SSD is worth for gaming? I wonder about the 960 EVO 256 GB. Will it work with 2 GPU's in SLI, two normal SSD and one HDD?

Please help, Im a noob.
5,565 Views
10 REPLIES 10

Arne_Saknussemm
Level 40
Yeah good choice...code is same as Formula basically without watercooling block. So, easily capable of 5.0 I would think...this depends more on your cooling...with an AIO water cooler then you would be set I reckon.

I believe two sticks of RAM are easier on the memory controller of the CPU so 2x16, despite being higher density modules, should be easier to run than 4x8. Of course 2x16 or 4x8 will both be dual channel...even if one set is rated quad...

Worth it is hard to answer...certainly 960 Pro is a great drive and the fact you can hide it away with no cables etc is great. Will work fine with SLI and there are 6 SATA ports that will work with it too. It won't make a huge difference though for gaming.

Raja
Level 13
On the T-Topology boards, four DIMMs can sometimes do better than two.

Arne_Saknussemm
Level 40
Right...don't get better info than that...

That depend on anything easily explainable Raja?

Arne Saknussemm wrote:
Right...don't get better info than that...

That depend on anything easily explainable Raja?


Not really easy to explain, as one would need to look at the signal analysis to see what happens. From my limited knowledge of signaling, it's likely to do with inductance and capacitance.

JustinThyme
Level 13
Code would be a decent choice. We could help you more if you were a bit more specific. Exactly what are you intending on using the machine for. If its just competitive overclocking then hold out until the APEX is readily available. That board is set up specifically for competitive overclocking.

What do you intend yo use for cooling?
Cooler=better OC potential 100% of the time.

Ram is a delicate balance. 10Ghz wont do you squat if you have to use latency of 100 to get it. Try to keep the latency at 16 or less as its not about clock speed of the ram its about throughput and the faster the clocks, the higher the latency. Think of it this way. Getting your new MOBO shipped same day by courier doesnt account for squat if the courier waits 3 days before he leaves. Same with ram, having 4000Mhz doesnt do you any good if you have to wait 25 clock cycles to get it.

In days of old ram and overclocking was a much harder choice. Ram multipliers were locked and you had to find the golden sticks and often a single or two sticks at most to get a decent OC. Today its more about if specific sticks properties allows them to play nicely. You can find a list of sticks that have been verified by ASUS in the qualified vendors list. This doesnt mean others wont work, simply that those listed have been tested by the R&D dept to work on the selected platform. Its very important to buy them by the kit and not one here and one there even if they are the same part number. Kits are binned and tested to perform well as a kit.

I still see folks posting a lot from old technology principles suggesting low density packs and only one or two sticks.
Im running 4 high density sticks a little past their rated speed. In a M9F andf previously an M8E a 64GB 3400Mhz trident kit 16,16,16,36 runs perfectly from 2133 and past 3400 at 1.35Vram. Ive been able to tighten the latency a bit at lower frequencies.

Basic math.
4000Mhz Ram at CL 19 is 19/4000 or 0.00475 seconds to respond to a command.
3400Mhz Ram at CL 16 is 16/3400 or 0.00470 seconds so about the same performance, just a miniscule edge that is hardly noteworthy. What is noteworthy is those huge frequencies amount to much of nothing when you tag a higher latency with it.

Its also noteworthy that not everyone gets the same results. All the best MOBO and best ram for someone else on the chip they sourced will not always do anything for you. The most important part of the equation is a shot in the dark being the CPU. While the Kaby Lake chips seem to be doing fairly well its a 50/50 chance the chip you get wont make it to 5Ghz no matter what you plug it into. This is an improvement as the process gets better as not long ago it was 50% not making it past 4.8Ghz. Just be prepared for that simple fact and either be willing to accept you wont make 5Ghz, pay extra for a binned chip, or source your chip from a place that has a liberal return policy and order two or more, keep the best and return the rest. I went with the last option with sourcing just two. The one that went back wouldnt pass 4.8 Ghz for anything. The one I kept went to 5Ghz with ease. delidded and liquid ultra treatment and easilty to 5.1GHz on a cardboard box with Corsair H115I AIO and 64GB 3400Mhz Trident ram that everyone said would hold me back. finishing up getting it in the case with a decent cooling solution today. Not much to go now. Been leak testing for two days now just because work took over my life for a few days and the fact that my kids gotta eat comes before adult play time. Im still perplexed at the number of people that dont get the last part. Saw a guy a couple weeks back leaving a microcenter having just dumped $2K in one visit had a 5 YO little girl on his shirt tail dressed in rags and left in a 30 year old toyota with bald tires blowing smoke out the back. That crap really bothers me. I would eat mud pies, go naked and walk everywhere just so my kids could eat a decent meal and have decent clothes to wear.



“Two things are infinite: the universe and human stupidity, I'm not sure about the former” ~ Albert Einstein

JustinThyme wrote:

Basic math.
4000Mhz Ram at CL 19 is 19/4000 or 0.00475 seconds to respond to a command.
3400Mhz Ram at CL 16 is 16/3400 or 0.00470 seconds so about the same performance, just a miniscule edge that is hardly noteworthy. What is noteworthy is those huge frequencies amount to much of nothing when you tag a higher latency with it.



The read turnaround for an already latched row (when only taking CAS and the resulting burst into the equation):

DDR4-4000 @ CAS 19 = 11.5ns
DDR4-3400 @ CAS 16 = 11.76ns

The former is faster by a small amount per read, but one would need to check all other timings to make a true assessment of which would be faster on average. What is certain is the higher frequency would require more SA/IO voltage and a reasonable IMC.

Arne_Saknussemm
Level 40
Righto...understood...i was thinking of rule-of-thumb I could take to the store when RAM shopping heh 😮
https://www.reddit.com/r/Jokes/comments/27uv5x/a_sexual_encounter_between_a_capacitor_and_an/


Have you seen this guide wildapiospi? https://rog.asus.com/forum/showthread.php?90109-ASUS-Z270-buyers-guide-Q-A-thread

Good for highlighting the range from ASUS...which IMO is very well thought out!

MrAgapiGC
Level 14
interesting thread. I still have my old trindent z modules. It is correct regarding using 2 instead of 4 modules. It is easy to the MOBO to make the calculations. Question. the same calculations apply using a kit of 4 modules that are tested, and a kid of 2? i have a 32 kit (4) model F4-3200C116Q-32GTZB. These are QVS modules. I only see the 16 variant on the list. Can anyone confirm that these modules are ASUS ready modules. I am ordering a z270 ROG with 7700K, If so good, if not, i have to hurry. Prices of rams are starting to climb the hill on prices!
Learn, Play Enjoy! We help and collaborate, NOT complain!

Arne_Saknussemm
Level 40
I would imagine that they work...they don't have time to qualify every kit...