Is 1.45V safe on DDR4 RAM?

Status
Not open for further replies.

steffeeh

Reputable
Feb 12, 2016
265
1
4,815
You see my build in my sig if that helps

So I have a pair of Corsair Vengeance DDR4 2400 MHz CL14 8 GB (=16 GB) at 1.2V stock, which I may have succesfully reached 2666 MHz CL13 at 1.45V on.
Touching the RAM sticks right after a gaming session on a RAM heavy game, they surely feel hot, but not burning hot and there's no problem touching them without it feeling to hot to touch them.

Before doing this last overclocking step on the RAM, I did some reading and found an older Q&A with Corsair on the new DDR4 sticks, and they suggested going no further than 1.4-1.45V, or there is a risk of harming the RAM sticks, and even the CPU IMC if you're unlucky (combine this with some other instances saying that 1.35V is totally fine, and 1.4V is still good).
However that thread is a bit old probably, and I still want some second opinions... is this voltage setting on the RAM sticks safe?
Are there any examples of massive failures caused by RAM overvolting? If you'd google failures caused by CPU or GPU overvolting you'd get plenty of scary results, but I don't find anything similar with DDR4 RAM.

Also, is CAS latency also affected by the voltage, meaning cutting the latency requires extra voltage to maintain stability, just like increasing the frequency eventually requires more voltage?
 
Solution
You mentioned liquid cooling at 1.32, which to most means water. Effectively saying that off the shelf DRAM that calls for 1.35 needs liquid....please, that's rather ridiculous...and odd as i have 1.35 DRAM in both my Z170 and my X99 and they don't get hot. And yes there are people out there that want to water cool their, most for appearance, again somewhat ridiculous and a waste of money


I actually read that one, though it's 2 years old... wouldn't the recommended max voltage have been updated somewhat since then, either increased or even decreased?
 
So by now I've received a lot of input on this topic on multiple forums, and overall people say that 1.45V is in the greyzone and wouldn't be recommended for 24/7 usage, so I guess I'll avoid that amount of voltage. So for now I've bumped it back to 1.35V until I've done some more research.
I've realized too that my frequency boost shouldn't normally require such high voltage boost, and it turns out this is because I've adjusted the settings poorly, such as having a too short true latency along with other timings, as well as overlooked other voltage settings of different parts that communicate with the RAM, which if configured right contribute to the overclock stability (=less DRAM voltage is needed).
So by sacrificing some latency, and adjusting different other voltages more than DRAM voltage, I should be able to reach a good overclock with 1.35V.
However, is 1.4V also in the greyzone, or is it okay to use in the long run? I have at least one person who said that 1.4V is no problem but that I shouldn't go past that, but I want to hear from more people (some people have even said that 1.45V is okay for heavier overclocks as long as they don't get too hot, obviously as heat degrades hardware, but they scratched their heads over that I used such heavy voltage for so little overclock).

The different voltage settings people have told me about are:
 - System Agent Voltage (aka VCCSA): Which most people recommend setting to 1.1V, however I'm a little bit confused here as this voltage is by default actually higher than 1.1V already (currently at 1.176V = the same voltage as the CPU core stock voltage). What should I do here?
 - VCCIO Voltage: Should also be set to 1.1V according to most people, currenly set to 1.05V by default. The odd thing here is that there is both a VCCIO for the CPU and another one for the PCH - which one should I change (not sure why increasing one of the chipset voltages would matter)?
 - VTTDDR Voltage: Should be half the current DRAM voltage, so I need to bump it up from 0.6V to 0.68-0.7V
 
I've been trying out some overclocks just to check stability, and I just can't seem to reach 3200 MHz, despite having edited the necessary voltages (+/- 1.15V VCCSA, 1.1V VCCIO, 0.725V VTTDDR), using 1.45V, having timings 20-24-24-45, having overclocked the CPU to a stable 4.4 GHz.
These settings may seem ridiculous, but I just use them as a failsafe when checking compability. The furthest I could go was 2800 MHz (i.e 3000 MHz didn't work either).

Any tips on how I may achieve 3200 MHz?
The reason why I aim at this specific frequency is because there seem to be compability issues with X99 chipset when you overclock the CPU or using all RAM slots (in this case the CPU is overclocked), so all bandwidths above 2666 MHz until 3200 MHz will only result in the RAM sticks disappearing one by one until a bluescreen appears.
 
General rule of thumb is never go over 10% unless your liquid cooling. That puts it at 1.32 max to begin liquid cooling. I would say 1.45v is way to far overclocking. Most I would do is 1.35 given some companys are a using bit higher voltage than Jedec spec.
 
You mentioned liquid cooling at 1.32, which to most means water. Effectively saying that off the shelf DRAM that calls for 1.35 needs liquid....please, that's rather ridiculous...and odd as i have 1.35 DRAM in both my Z170 and my X99 and they don't get hot. And yes there are people out there that want to water cool their, most for appearance, again somewhat ridiculous and a waste of money
 
Solution
First and foremost, disable xmp and memory try it. Once done you'll start to have a good time. Loosen timings, increase voltage to max safe, 1.35, increase soc voltage to 1.1 or 1.2, set sub timings to auto and then crank up the mhz baby.
 
I have been running 2 x 16gb Team Vulcan 2400mhz 15-17-17-35 to 2933mhz 16-16-16-36 @ 1.4V for 2 years now, no issue what so ever, the system is rock solid, I've done countless numbers of aida64 stress test/benchmarks. Just sharing what I've experienced.
 
Status
Not open for further replies.