• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

F4-3600C18D-32GTRS (i7 8700k) overclocking help sought

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

vulcan78

New Member
Joined
Feb 26, 2020
Hi all, new member here, simply googling my memory and the query term "overclocking" has resulted in no hits other than Woomack's OC review of the 16GB variant of this memory (I presume, same timings).
https://www.overclockers.com/forums...Royal-2x8GB-DDR4-3600-CL18-F4-3600C18D-16GTRS

Anyhow, it's stable under XMP, timings are a bit higher than I'd like, any recommendations as to where to start to try to OC these bad boys a little?

F4-3600C18D-32GTRS

Trident Z Royal
DDR4-3600MHz CL18-22-22-42 1.35V
32GB (2x16GB)

I love the way they look but I primarily got them after much research and coming to the conclusion that pushing for 4000Mhz+ with my motherboard is extremely difficult as it has single phase DRAM VRM (Gigabyte Z370 Aorus Gaming 7). The web is littered with posts of people complaining about not being able to attain 4000 MHz+ even at 1.45v+ with this particular board. So knowing that, and seeing a few comparison videos showing the gains to be had simply going from 16 to 32GB of memory with the same freq and timings I decided to upgrade from my previous memory: F4-3200C16D-16GTZR (3200MHz CL16-18-18-38 1.35V) based on a 3200 MHz CL16 vs 3600 MHz CL18 video.

I figured there was some performance to be gained here and this would be about as good as it gets without spending a few hundred dollars more upgrading the motherboard to something with better DRAM VRM.

16GB vs 32GB

3200 MHz CL16 vs 3600 MHz CL18

I just ran Firestrike and although I wasn't expecting a gain in per-se in the Physics test I was surprised to see the CL18 3600 MHz memory run just slightly slower here, although the difference is too close and could be attributed to how variable this benchmark can be (I can run the same benchmark back to back with same temps and have Physics score vary by 50-100 points or more).

https://www.3dmark.com/compare/fs/21916785/fs/21849156

I know that this is not the benchmark to gauge memory performance, it's just that the last time I tried upgrading the former memory (without success, it wasn't stable ultimately) from 3200 to 3466 MHz I did actually pick up a few hundred points in Firestrike Physics test.

I intend to run Cinebinch R15 and Aida64 later.

I can't find anything pertaining to overclocking this memory, I would be extremely grateful for any help with this.

Thanks in advance!

Post script:

I forgot to add that this memory is on the G.Skill's updated QVL for this motherboard.
 
Last edited:
As I mentioned in another thread, check something like 16-19-19 or 16-20-20 at 3600. Voltage 1.35-1.40V should be enough. Since you OC it on Intel then if CL16 won't work then check CL17-19-19 or 17-20-20. I'm not sure how 16GB Hynix modules are acting at 3600+. Once you drop timings then check how high it will go at CL18. 8GB modules from new series usually make 4000-4200 18-22-22 or 18-24-24. Voltages also 1.35-1.40V... maybe up to 1.45V.
 
As I mentioned in another thread, check something like 16-19-19 or 16-20-20 at 3600. Voltage 1.35-1.40V should be enough. Since you OC it on Intel then if CL16 won't work then check CL17-19-19 or 17-20-20. I'm not sure how 16GB Hynix modules are acting at 3600+. Once you drop timings then check how high it will go at CL18. 8GB modules from new series usually make 4000-4200 18-22-22 or 18-24-24. Voltages also 1.35-1.40V... maybe up to 1.45V.

Hynix is totally different but results on Intel and AMD should be similar up to 3600 and next at 4000+. How high can you OC depends on many variables so I can't help you directly, especially that I had no new Hynix memory kit with 16GB modules.
However, new Hynix in general OC like 16-19-19 at 3466, 3600, 3733, 3800, 3866 <-- how high clock it can make depends on your luck and IC quality and the motherboard.

Thanks for the quick reply with this! Sorry for multiple threads, anyhow they are brand new Hynix apparently, dated to Feb 2020 on the modules. You say that newer Hynix 8GB variant series overclock better than old? Thanks for the suggested starting values but what about cycle time?

I gather that Cycle Time value is some kind of sum of the first two numbers but the math doesn't seem to work out exactly, i.e, current values: CL18-22-22-42, and that of the modules these replaced,CL16-18-18-38. Not sure what to put here for say, 16-19-19, maybe 40 or 36? I'm at 1.35v out of the box so should I just bump it up to 1.40v, or something in between maybe? Is that ok for 24/7 use?

And what is the best, safest method of checking for stability? MemTest86? What about the Aida64 Extreme? Should I make a system image before hand or is MemTest86 somewhat safe as it's a pre-windows environment?

I'm also somewhat worried to even mess with this as the last time I attempted to overclock my previous modules my motherboard refused to boot and then, because it has dual BIOS it switched to an older BIOS on it's own, it was very problematic. I'm also kind of worried about somehow affecting the addressable RGB on these modules, G.Skill Trident-Z software stopped working with the last modules, I don't believe from overclocking per-se as the problem developed before I tried to push them to 3466 MHz. Those modules wouldn't take ANY kind of overclock. I couldn't even tighten the timings, even at 1.45v and after increasing VCCIO and VCCSA.

Thanks so much for the help!!!!!!
 
Update:

CL17-19-19-40 was not stable, there were only 2 errors in MemTest86 that occurred in Test 9 of 13 but then Aida64 Memory and Cache stability test sussed that out really quickly with a failure at the 3 min mark. So back into BIOS I went, this time I wanted to rule out possibly having bad modules so I re-ran MemTest86 with default timings and voltage, that passed, then the aforementioned Aida64 stability testing where at 15 minutes I presumed that it was at least 99% stable.

I then tried upping the freq. and voltage while keeping timings the same and did secure 3800 MHz, which passed Pass 1 of 4 in MemTest86 and has now gone on 10 min of Aida64 stability testing as I type this.

I did have to increase DRAM voltage to 1.4v and I'm wondering if the additional voltage is worth 200 MHz.

Before settling here, after feeling confident after 3800 MHz passed MemTest86 I tried for 3900 MHz with the same voltage and it wouldn't even POST (fast boot is disabled so that the PC can train the timings etc.).

I'm tempted to try for 4000 MHz @ 1.45v but I don't know if that is prudent for 24/7 use.

Any further guidance would be greatly appreciated.

3800 MHz @ default timings is definitely an upgrade over my former modules though that wouldn't take ANY overclock to speak of.

2x8GB 3200 MHz CL16-18-18-36 Trident Z

vs

2x16 3800 MHz CL18-22-22-42 Trident Z Royal

Memory temperature according to Thaiphoon was 50 an 55C under load after 30min of Aida64 that has subsided to 40 and 44C about 5 minutes after stability testing.

I presume that higher temp was the module to the right of the first module that is getting airflow from top rear 140mm fan that I have reversed (I have my radiators exhausting out of the case). Not sure

These modules look really good, and they were only $200 on Newegg. They will match the Nickel back-plate I'm about to put on the 2080 Ti, I'm just waiting for my EK Leak Tester to arrive because I have to tear down the monoblock and clean the CPU heat-sink due to plasticizer getting into my loop from 2.5 years of soft tubing (EK Duraclear).

TLDR follows.

I have 14/10 Acrylic clear tubes coming along with a nice distribution block (Thermaltake View 71) from Ali Express along with about $200 worth of Bykski fittings (would be $300 worth if they were EK) to avoid further plasticizer problems down the road, I will just replace the current soft tubing with like kind that I have on hand until that arrives three weeks or so from now.

I just have to do simple 90 degree angle bends and Acrylic is more UV resistant (PC is currently living room that is south facing and gets a lot of sun in the afternoons, PETG holds up a little better to direct and indirect sunlight compared to soft tubing but ultimately you will have the same discoloration and leaking plasticizer problem, it just takes longer, and the other attendant problems with PETG, permeation and failure at 55C water temp are things I don't want to deal with, I will just take harder to bend acrylic since I'm only doing a few 90 degree angle bends to and from the distro block).

Anyhow, sorry for the TLDR, pretty satisfied with the memory upgrade to be honest, 4000 MHz and above is really hard to get on this board as it has single phase memory VRM (Gigabyte Z370 Gaming 7), 3800 MHz and lousy timings will be it for me until I upgrade when upgrading from 8700k @ 5.0 GHz is actually a compelling value proposition.

Aida64 stability testing just passed 30 min, I'm fairly confident it's stable here.
 
Last edited:
Update:

I spoke too soon, I didn't realize this but setting the memory to 3800 MHz while also keeping the timings on "auto" really retarded the timings and the performance was worse than at 3600 MHz @ 18-22-22-42.

3800 MHz indeed up being at 19-23-23-45, something I only discovered while doing the Aida64 benchmark (curious as to why it was taking so long to complete):

3800 MHz 1.4v default timings.png

So back to BIOS again, this time it refused to POST @ 3700 MHz with default timings (Manual) and 1.4v.

So I tried one more time to see if I could get away with at least tightening the timings and thankfully I was able to get 3600 MHz @ 17-20-20-38 @ 1.4v:

3600 MHz 17-20-20-38 1.4v.png

This is a little better than stock, but not sure if the slight performance bump warrants the additional voltage:

3600 MHz default timings and voltage.png

It passed Pass 1 of 4 in Memtest86 and just endured a full hour of Aida64 Memory and Cache stability testing.

I will probably keep it here, I may try to dial back the voltage but I'm not sure I even want to do that as 17-19-19-38 @ 1.4v wasn't stable in MemTest86 nor Aida64.

Very interesting that a slight increase to 17-20-20-38 and now it's stable @ 1.4v.

Anyhow, 3600 MHz 17-20-20-38 is still faster than 3200 MHz 16-18-18-36 and the increase in memory will be useful for a few games / applications that will benefit from more than 16GB of physical memory.
 
This IC generally runs like that. Depends on luck a bit better or worse. G.Skill, Patriot and some other brands are selling higher Hynix kits at 3600 16-19-19 (mostly G.Skill in their Neo or Royal kits) or 17-19-19 (mostly Patriot and some others).
Re 4th timing, it's usually like 2x the previous one +/-1 on these new kits so something like 16-19-19-39, 17-19-19-39, 16-18-18-38 or 39. This is how memory manufacturers are setting XMP. Depends on used IC it can be slightly lower or a lot lower and will be still stable but it doesn't really affect performance. Your settings look good so as long as all is stable and you are satisfied with the results then all is fine :)
 
This IC generally runs like that. Depends on luck a bit better or worse. G.Skill, Patriot and some other brands are selling higher Hynix kits at 3600 16-19-19 (mostly G.Skill in their Neo or Royal kits) or 17-19-19 (mostly Patriot and some others).
Re 4th timing, it's usually like 2x the previous one +/-1 on these new kits so something like 16-19-19-39, 17-19-19-39, 16-18-18-38 or 39. This is how memory manufacturers are setting XMP. Depends on used IC it can be slightly lower or a lot lower and will be still stable but it doesn't really affect performance. Your settings look good so as long as all is stable and you are satisfied with the results then all is fine :)

What is interesting is that I set 4th timing to 40 in BIOS and somehow it was 38 once in Windows, very odd. Anyhow, yes I suppose I'm ok with this but honestly I am wanting a little more and am tempted to try for 3800 or 3900 MHz at 1.45v with the same timings but I'm afraid that that may be too much voltage for 24/7 usage and, given your expertise and reviews of similar memory (the 8GB variant with same timings but Samsung B-Die) I'm worried that increasing the voltage from 1.4 to 1.45v may not help much with attaining more freq. from these modules and this may be as good as it gets.

How do you feel about 1.45v for 24/7 use?

Thanks again!

Edit: I also have another problem about voltage, if I set the DRAM voltage to a particular value in BIOS, say 1.4v, is that value dynamic or (when the memory is of low load) or it static.

I can't seem to figure out how to determine what the voltage value is with software, Hwinfo64, CPU-Z, Thaipoon, nothing seems to have this value.
 
Last edited:
There are many memory kits with a lifetime warranty and 1.45-1.50V. Pretty much every 4500+ memory kit has 1.40-1.50V. I assume that this voltage is safe as no one would put a lifetime warranty on something that could bring loses and affect a brand's reputation.
 
Thanks for the reply! I will see what this memory can do at 1.45v and report back my findings.
 
I was able to get 3600 MHz @ 17-20-20-38 @ 1.4v:

this works to me, i have i9 9900k + Maximus XI Code and F4-3600C18D-32GTZN.

But i want to try 4000mhz, and cant find any timing/config.

and i think i made a mistake, i bought a 16gb Kit(F4-3600C18D-16GTZRX) with the same timing however it is B-Die. the system is a little unstable, I got this config(18 22 22 42), however after turning off the PC and only turning it on the other day dont post(it is nothing more than the post), you need to change to auto, start the windows and then go back to mobo and redo the config, I don't know what can be. I'm thinking of returning the 32gb kit and buying another 16gb if someone can help.
###Edit:

test on 18-22-22-42 and 1.410v :
XqI5QnG.png



is a so serious problem mix F4-3600C18D-32GTZN + F4-3600C18D-16GTZRX ?

FLP
 
Last edited:
In general, if both kits work at the same timings then it should be fine. Clearly something doesn't work with sub timings if it's unstable. Check if it works the same if you switch memory sticks so if you had 2x8GB in first slots then install 2x16GB in them. It's because the motherboard is usually setting timings for the first memory which is recognized and then uses it for all other sticks.

G.Skill is using every available IC in these 3600 CL18-22-22 kits. It can be Micron, Hynix, or Samsung. There is no guarantee you will get a second Samsung kit if you decide to buy it. Also, why do you want to use 4x8GB when 2x16GB is the same and it should work without issues?
 
the 32gb kit doesn't really work with the i9 9900k, i will return it to the seller. in the first test I did when putting the kit at 4000mhz it worked, I didn't change the timings I left the default and it worked. I intend to lower it a little. With the 32gb kit I couldn't even get 3800 with the same timing or even increasing.


uNlIf57.png

do you recommend any timing for this model(F4-3600C18D-16GTZRX) at 4000mhz?

i have i9 9900k + Maximus XI Code


##### edit :

and I notice that sometimes the tabs of the opera crash at 4000mhz, is this connected?

and again when turning off the PC it is necessary to at least enter the bios for the computer to turn on. I increased the voltage to 1,450 so let's see if it improves.

FLP
 
Last edited:
Run the AIDA64 stability test. For memory use memory and cache tests only, others are not important. Usually ~1h of this test is enough to tell if memory is stable or not. The cache test is added because memory works close with cache and if memory controller doesn't have enough voltage then it will crash too.

16GB or 32GB memory modules are harder to OC. Usually, they're not passing much above 3600. On your motherboard, as long as the memory kit is good, you should reach 3866. I didn't have this motherboard and MXI Gene, that I have, is tuned for the maximum clock so it works a bit different.

I still would recommend 2x16GB, maybe something from the QVL to be sure it will work. If you can't return both kits then get a second one 2x8GB, the same as you have.
 
Back