FEATURED Crucial 4x8GB DDR4-2133 CL15 1.2V - CT4K8G4DFD8213

Woomack said:

There is a chance that 2400 kit will have slightly better chips. As I said there is some difference between my sticks while matched kit will probably work about as high on all sticks. Still 2666 clock should be possible on all of them.

Click to expand...

I have the 2133mhz kit, and want to run it at 2400mhz. I cant really do much for the next 1 or 2 weeks. Still missing my pumps, and have a lot of wiring to do, then i'll get started in the ram once the rig is up and running solid.

Hello, I would like to say you do some lovely work overclocking. I found an earlier topic of yours a few weeks ago, and ended up buying a new system myself last week with the same processor, motherboard and ram as you're using here. Funny to find a new topic from you and see you have the same parts now.

I'd like to ask, have you done any more testing with these modules for 24/7 overclocks? I don't really intend to run my pc that much, but as I use it for video editing stability is still important. I was wondering if I should try getting 2666 mhz to work smoothly (I haven't so far) or stick with 2400 mhz.

More important than memory clock seem CPU Cache clock. Even DDR4-2400 provides good performance when cache clock is higher. 2666 will be for sure better but difference won't be as high as going from 3000 to 3500+ cache. I've switched motherboard to ASUS RVE and there is even higher cache clock available ( up to 4250 stable on my CPU ) but cache voltage has to be set higher.

All 4 modules of Crucial 2133 are working fine at 2666 and tighter timings than declared by manufacturer. I had to raise voltage to 1.30-1.35V to set optimal performance at 2666 and 13-12-13-28 or 13-13-14-32 timings. I had no time to make longer stability tests but 1-2h without errors is usually enough. To be sure you can leave timings at 14-14-14-34 2N what shouldn't be much slower.

I see, I have been reading up about what's new to cpu overclocking and the cache parts seemed interesting enough. From what I read it was suggested as though you could look for a stable clock, and then after you find it try putting the cache at the same speed and playing with the voltage a little. What you're saying though is that on the msi board you didn't get the cache up all that high?

I suppose for me that won't be too big of a problem, my personal goal is to see how far I can get and then I'll most likely go a bit easier on my system. Both to cut a little slack on the power and to keep the whole system from powering up the fans like crazy. I've been a quiet system purposely and if it means I won't get to push the system too far then so be it. Thanks for the tips though!

What kind of stress test would you suggest for testing the stability the memory? I have ran some memtest86+ for short times and prime95 for several hours, didn't really want to do 12 hour tests yet until I've settled on settings I'd really like to stick with. Is there something else I could use to really test the memory, or perhaps I should just run memtest86+ for a longer period of time?

I have a set of this RAM and for the price it is freaking awesome value. I ran mine out of the box at 1.2v doing 2400 the stock CL15 timings, no trouble at all.

From yesterday I have bumped the voltage to 1.35v which I am reading is safe, therefore I will now use this voltage for 24/7 usage, takes me to 2666 and timing 14-14-14-32-1T

For a new release and first batch of DDR4, the micron IC is a very sweet little number for budget RAM.

The only advantage I can see for buying more expensive RAM is guaranteed speeds and a better with higher layer PCB, other than that I see little point wasting anymore money.

Buying this cheap Crucial RAM allowed me to bridge the gap to buy my first ever extreme processor while near release time.

Deston said:

What kind of stress test would you suggest for testing the stability the memory? I have ran some memtest86+ for short times and prime95 for several hours, didn't really want to do 12 hour tests yet until I've settled on settings I'd really like to stick with. Is there something else I could use to really test the memory, or perhaps I should just run memtest86+ for a longer period of time?

Click to expand...

Quick - HyperPi32M , longer - memtest86+ in the latest version, Prime95 blend, XTU memory test ( it's based on Prime95) , AIDA64 memory test ( there is stability tab )
Can also try something like WinRAR benchmark with fully covered memory ( pick some large files ) or Intel Burn Test or anything else where you can pick memory size to test. When memory is fully covered or nearly ( liek 95% or something ) then you don't have to test it for long time and 1h should be enough.

microfire said:

I have a set of this RAM and for the price it is freaking awesome value. I ran mine out of the box at 1.2v doing 2400 the stock CL15 timings, no trouble at all.

From yesterday I have bumped the voltage to 1.35v which I am reading is safe, therefore I will now use this voltage for 24/7 usage, takes me to 2666 and timing 14-14-14-32-1T

For a new release and first batch of DDR4, the micron IC is a very sweet little number for budget RAM.

The only advantage I can see for buying more expensive RAM is guaranteed speeds and a better with higher layer PCB, other than that I see little point wasting anymore money.

Click to expand...

This is actually high quality memory and I doubt you will see any difference going on a better PCB.
Almost every DDR4 right now is based on the same memory chips. So generally there is 1 Micron, 1 Samsung, 1 Hynix IC in consumer grade DDR4 kits. The main difference is binning for higher clocked kits but as you see, even that isn't so great as most manufacturers are setting really relaxed timings for even these expensive series. I'm not sure if it's because of lack of time or cost cuts.

Probably every Crucial DDR4 is able to run at 2666 clock which is max clock without going to 125bclk strap on most motherboards. Additionally all Crucial DDR4 that I saw could make something like CL13 at 2666 and ~1.35-1.40V.

2666 clock is optimal for daily work even at high overclock of other components. The limiting factor in this case is cache speed, not the memory. Overclocked memory is mainly lowering latency which is kinda high at stock clocks. Bandwidth is mainly limited by CPU cache clock.
When you pick 8 core+ CPU then you see higher effective bandwidth in applications due to multithreading. So tests like AIDA64 will show higher max bandwidth on 8 than 6 cores.

I cant wait to fire mine up soon!

Do you have any settings you can start me off with Woomack? Since im running the same motherboard you started out with.

cpu:
- raise cpu ratio when you notice instability raise voltage ( reasonable max voltage will be about 1.35V and temps below 80*C )
memory:
- memory clock -> raise memory ratio, when you notice instability raise voltage ( reasonable max voltage about 1.35V )
- timings -> lower timings, when you notice instability, raise voltage
cache:
- raise cache ratio, when you notice instability raise voltage ( reasonable max voltage about 1.35V ), this will stop at about 3500MHz on MSI

Set LLC to 100%. On MSI it will give about the same voltage as you set in BIOS.

simple as that

Guides could be made for 4.7GHz+ but till that clock it's about the same as 1150 CPUs or even earlier as you don't have to change much.

I was thinking mainly ram, Like a few settings that know should work without anything, 2400mhz LC something.

Woomack said:

This memory is not getting hot even after overvoltage so it doesn't require any additional cooling. You can use heatsinks for looks but I'm not sure if it's worth to pay for that.
Right now most DDR4 kits on the market base on similar Micron or Hynix IC. Depends from binning you can find it in lower or higher clocked kits. Micron can usually run at tighter timings ( at least from what I saw and tested ).

Crucial kit from this thread is running pretty stable @2666 13-13-14-32 1N ~1.4V. I didn't check lower voltages but I guess I can try that when I switch platforms. Now I set all for 1150 tests.

Click to expand...

I have a question here, how have you tested your memory temperatures? I've read some people don't like going over 1.35v for their memory but you're saying these sticks don't get too warm. Wondering if you used a tool to measure temps manually.


I did have an early bsod before any overclocks last week and code suggested it was probably memory. Ran HCI Memtest for 400% and then memtest86+ for 4 passes, latter took like 13 hours, pff. Shame that as you say this motherboard will probably not get over 3.5 uncore. If I had known before I might have gone for a better mobo, but I do really like the look of this one! Personally some things got in the way so I am still searching for optimal core overclock, after that I'll do memory and lastly uncore.

Has a lot of info in memory overclocking changed recently or could I still use the memory overclock guide from your signature? One more thing that confuses me a little is the CH_A/B and C/D, is this about phases or memory channels? Will I be fine if I just raise them both to the same levels as if raising one single voltage control?

You can touch memory under full load and it will have low temps. I didn't check it with any thermometer but it was cold enough to not to use any fan. DDR4 are not much different from DDR3 so you can expect similar temps at similar voltages. DDR3 could run at least up to 1.75V without additional cooling and here we have 1.2V memory.

Soem users say that anything above 3500-3700 uncore is not stable even though you can run benchmarks at 4500+. I haven't been testing full stability so I can't tell. At least in my tests, uncore clock up to 4200 seems stable but above that there were problems.

There are separate phases for A/B and C/D channels. You can set higher vrm frequency and optimized profile on MSI board. Probably you won't see any difference anyway but in some cases it may improve stablity after overclocking.

How on earth do i go about overclcoking my ram? I cant get into the advanced menu on the ram? MSI x99s sli plus

EDIT:

Got that one figured out, and got the ram overclock started, but when you write e.g 2800 12-11-12-28 1N 1.5v, what is the 1N??

1N or 1T = Command Rate.
On MSI only x21.33, x24.00, x26.66 memory ratios are working. If you want to set higher then use lower memory ratio and 125bclk+ strap. 2666 clock seems optimal for my Micron kit but I saw some users with more luck.

I do not plan on going over 2666. So far i have it stable at 2400 13-13-13-30 1t 1,35 same as your 2400clock.

I seem to be unable to go to 2666.

2666 14-14-14 1N 1.5v doesnt boot
2666 15-15-15 1N 1.5v doesnt boot

Can you recormend a starting point and which timings to lower first?

EDIT:

BTW, with the 1.5 bios, default now hits 3200mhz @ 266mhz settings. Just noticed i was running 1.4 and updated.

Check Micron memory profile or just rewrite all timings from my results ( including sub timings ) and it should run at 2666. Above can be a problem. Stick to 1.35-1.4V max.
Also if you want to set manual settings then disable XMP or enable Micron profile and change some timings manually.

Can also try 14-14-15-32 1N 1.4V tRFC auto or something near 300 ( 280-320 ). For 2666 can also set ~1.05V VCCSA.

Been while, but bringing this thread up again as is a decent one and I am still running the same Micron modules.

I was have troubles with computer now waking from sleep at 2666 14-14-14-32 1T using 1.35v, although otherwise fine during stress testing and memory tests. Every now and then the computer would loop at cold boot. I took off the overclock and just ran the Micron at stock 2133 CL15, and everything is way more stable for power ups and wake from sleep, seems overclocking the RAM was causing other issues.

What I have found that just running at stock and tweaking the cache speed make a way better different than increasing the RAM speed alone, although latency is little bit loose but not to much. Also the stability is back for the system.

This is the benchmark running at 2133 CL15 1.2v, with cache bumped up to 4000MHz rather than the stock 3000MHz, which provides some impressive results. (Ignore the 1400MHz in the benchmark, it is actually 4000MHz, that is speed-step not running up to full speed)

I think you have to play with sub timings or additional options to make it run stable at higher clock.
That kit was running in my brother's PC for some time on MSI board @2666 14-14-14 1.35V without issues but of course not all kits are the same.

Not all motherboards can make more than 3500MHz cache clock so I'm testing everything at 4.2GHz CPU and 3.5GHz cache clock. At these settings it's easier to see differences in performance between memory kits and all readers can check it at home.
Sometimes I'm adding results at higher cache clock to show max bandwidth or just to compare something.

What means 1N and 2N after Timings?

they are wait latency. with timings the lower the better, as long at them memory is stable at said timings