Does RAM burn in over time?

[Carfax83]

Question, does RAM burn in over time? The reason why I'm asking is that when I first got my memory, I tried to run it at DDR4 2666 14-14-14-32 1T but it ultimately proved to be unstable after testing it with several memory hogging games like Batman Arkham Knight and Witcher 3.. So I backed the timings off slightly to 14-15-15-32 1T and it was rock solid stable.

Just recently though after fiddling with the UEFI bios, I decided to try 14-14-14-32 1T again. To my surprise, I haven't had any crashes or lock ups yet after playing several games back to back, including Batman Arkham Knight which previously had crashed without fail about two or three minutes of Batman gliding around Gotham..

I also played Rise of the Tomb Raider, which uses quite a bit of RAM as well and no crashes or lock ups of any sort.

So is it fairly common for RAM to do this? To be able to run tighter timings over time?

 

[VirtualLarry]

Did you do any BIOS updates in the interim?

 

[Carfax83]

Did you do any BIOS updates in the interim?

No, no BIOS update. The only thing I've done differently other than tighten the timings is I've also increased my uncore frequency from 3.7ghz to 4ghz.

But that should make the RAM more unstable you would think, not less..

 

[Carfax83]

I got a memory error last night when playing Rise of the Tomb Raider, but I think I know why I've been able to run for much longer periods with stability at 14-14-14-32 1T compared to when I first built this system..

I neglected to mention that I had swapped out GTX 980 FTW SLI for a single Zotac GTX 980 Ti Amp Extreme. I think doing so lessened the pressure on my system ram as I'm now running in single card configuration, and I also have 2GB of extra VRAM..

So while it took much longer to error this time, it still eventually happened..

 

[tweakboy]

LOL no ,,,,,, it can live for 10 years or 1 year,, you just dont know... gl

 

[cretinbob]

yes

 

[AlienTech]

I have noticed speed changes with chips after a while, but usually it is to a lower degree because the circuits are made from deposits and those flake and burn off so the circuits are less reliable.. So this might be a case of higher voltage to the chips or some other variable like line capacitance or inductance changes due to dust and dirt etc..

 

[destrekor]

I got a memory error last night when playing Rise of the Tomb Raider, but I think I know why I've been able to run for much longer periods with stability at 14-14-14-32 1T compared to when I first built this system..

I neglected to mention that I had swapped out GTX 980 FTW SLI for a single Zotac GTX 980 Ti Amp Extreme. I think doing so lessened the pressure on my system ram as I'm now running in single card configuration, and I also have 2GB of extra VRAM..

So while it took much longer to error this time, it still eventually happened..

I think you have it exactly right.

SLI will utilize system memory much more extensively than a single GPU, thus stressing it harder.

Bummer that you couldn't get your timings down with a single GPU, but all the same, I wouldn't worry about it. Memory timings are simply not a concern compared to what they had been in the early days of SD and DDR-RAM.

The system bus and memory speeds in this day and age make higher memory timings largely a non-issue. Outside of very specific benchmarks, I doubt one would ever truly notice a difference.

 

[razel]

There's no burn in. You wouldn't want burn in. If you're unstable, but are now stable even then you are in that bleeding edge of stability. One day when the power going to the RAM is perfect you'll be fine, the next day when it's not perfect you'll be unstable. Just throttle back a bit and consider that as 'stable.' But if your RAM isn't even performing at it's rating specs and you didn't punish it, then just claim warranty and move on. No need to try to find a ghost in the machine.

 

[nerp]

I don't believe burn-in for components is really a real thing.

 

[Mushkins]

I don't believe burn-in for components is really a real thing.

It's not. A component either works or it does not, and as it's used more it degrades.

The only PC component that needs to "burn in" would be some types of thermal paste that require the heating and cooling cycles of the processor to cure the paste.

 

[destrekor]

I don't believe burn-in for components is really a real thing.

It's not. A component either works or it does not, and as it's used more it degrades.

The only PC component that needs to "burn in" would be some types of thermal paste that require the heating and cooling cycles of the processor to cure the paste.

Yeah PC components don't need burn-in, it's just not a thing for ICs. As Mushkins said, thermal paste does need heating cycles, but that's it.


Now, the concept of "burn-in" is a real thing for other devices. I know all displays need time to settle in before committing to calibration. And speakers are usually advertised as requiring some time, but I don't buy that considering [the good] speakers are tested extensively before being packaged up.

 

[nerp]

People like comparing processors to engines, and I think people have always tried to put mechanical physics on electronic devices. I swear, when I dust off my computer, it runs faster!

 

[AlienTech]

I don't believe burn-in for components is really a real thing.

This is definitely a problem with like UPS's and such where power spikes will burn of a little bit of the chips each time and they still work as long as things are normal but act flaky when things are not perfect. After using half a dozen of them things, I learnt to not rely on them after like there is a lightning strike close by and change them even though they still work. I dont think I have ever directly and I mean immediately blown one up. They happen gradually and then one day stop working.. But they are flaky for a while before they die. Now the same thing can happen to a CPU because there are so many parts to it, parts of it can stop working but you only notice it when those particular parts are being used. Like ALU problems only show up when you use floating point and one day you try to encode some video and it shows dots and glitches but it seems to work in your normal every day web surfing. Space probes have this problem a lot.. Those high intensity cosmic rays will blow of parts of the CPU but it still functions until those particular parts are needed..

 

[AlienTech]

For an year I drove a car with only 3 pistons working.. It was slow and a pain to get started but heck it worked and I thought it was normal.. And yea I had no idea until someone else told me after checking the spark plugs.. And I dont go make claims about such things if I dont know about them. Computers have a lot of things in them that work because many other things work properly, even supposedly unrelated events can make them behave strangely. Many of the things that make computers work are not easily measurable and can not be seen or heard but that dont mean they dont affect it. Unlike driving a car with only 3 pistons working anyone who has been around engines can hear the difference without even looking closely.

 

[tortillasoup]

Question, does RAM burn in over time? The reason why I'm asking is that when I first got my memory, I tried to run it at DDR4 2666 14-14-14-32 1T but it ultimately proved to be unstable after testing it with several memory hogging games like Batman Arkham Knight and Witcher 3.. So I backed the timings off slightly to 14-15-15-32 1T and it was rock solid stable.

Just recently though after fiddling with the UEFI bios, I decided to try 14-14-14-32 1T again. To my surprise, I haven't had any crashes or lock ups yet after playing several games back to back, including Batman Arkham Knight which previously had crashed without fail about two or three minutes of Batman gliding around Gotham..

I also played Rise of the Tomb Raider, which uses quite a bit of RAM as well and no crashes or lock ups of any sort.

So is it fairly common for RAM to do this? To be able to run tighter timings over time?

Compare room temperature when you first conducted the test and the room temperature recently, report back. Ambient air temperature can definitely have an effect on whether something works properly or not. Even 10 degrees can make a difference.

 

[DigDog]

no, but yes. what is slowly failing is the mobo, as it ages it has difficulty pushing the correct voltages. generally, bumping the voltage in bios will sort out the problem.

 

[tortillasoup]

no, but yes. what is slowly failing is the mobo, as it ages it has difficulty pushing the correct voltages. generally, bumping the voltage in bios will sort out the problem.

it has been my experience that these motherboards primarily 'fail' because of crappy electrolytic capacitors. Maybe OP could upgrade the capacitors in his board and see if that allows him to improve performance.

 

[DigDog]

yep, it's the capacitors. but, you can't really replace them (or at least i don't think you can), you need to buy a new board.

when a voltage fails, you raise it in the bios .. and all you are doing is putting it back to where it was originally. the bios reads it wrong. or rather, it reads it fine, but if the board is old, some of that voltage never makes it to the component.

so for example, when my E6600 needed 1.32v to overclock, and then decided to crap out, i raised it to 1.36v, but the actual voltage the cpu got was still 1.32v. it just "leaks" out of the mobo in the form of heat.

later tested the cpu with a new board, and it was as good as new - on 1.32, it did 3.2Ghz, same as day 1 on the old board.

 

[tortillasoup]

yep, it's the capacitors. but, you can't really replace them (or at least i don't think you can), you need to buy a new board.

when a voltage fails, you raise it in the bios .. and all you are doing is putting it back to where it was originally. the bios reads it wrong. or rather, it reads it fine, but if the board is old, some of that voltage never makes it to the component.

so for example, when my E6600 needed 1.32v to overclock, and then decided to crap out, i raised it to 1.36v, but the actual voltage the cpu got was still 1.32v. it just "leaks" out of the mobo in the form of heat.

later tested the cpu with a new board, and it was as good as new - on 1.32, it did 3.2Ghz, same as day 1 on the old board.

they're definitely replaceable. It's also very important to get low ESR capacitors as well.

 

[DigDog]

they're definitely replaceable. It's also very important to get low ESR capacitors as well.

hey, looks like you are right.

https://www.youtube.com/watch?v=ZP9tfu3Xjso

 

[Soulkeeper]

Some memory errors can be hard to trigger and only happen once ever few weeks under the right conditions. IE: a few degress ambient change, power draw from other components at the same/right time dropping the voltage just enough.
Part of the joys of extreme memory overclocking and custom timings.

I spend lots of time stability testing before I confirm it as stable. My favorite is compiling source code for hours on end, most memory that can make it thru that error free is usually ok.

 

[TheRyuu]

I got a memory error last night when playing Rise of the Tomb Raider, but I think I know why I've been able to run for much longer periods with stability at 14-14-14-32 1T compared to when I first built this system..

I neglected to mention that I had swapped out GTX 980 FTW SLI for a single Zotac GTX 980 Ti Amp Extreme. I think doing so lessened the pressure on my system ram as I'm now running in single card configuration, and I also have 2GB of extra VRAM..

So while it took much longer to error this time, it still eventually happened..

Yea it just means that it's unstable in it's current configuration and you'll have to back off something somewhere (not sure what else you have overclocked). I suppose if you're 100% sure your CPU overclock is stable then I guess it's the RAM so you should back off on it. I doubt it'll make any difference in performance. Wouldn't these RAM errors show up in a long session of memtest?

The only "burn in" that will happen in a computer is the thermal paste and even then it will affect certain kinds/brands more than others (Arctic Silver being a famous example of one which required some amount of burn in).

 

[Deders]

It's not burn in as such, but a lot of motherboards do have memory training whereby they adjust the timings at startup to ensure stability.

So whilst you will be adjusting one or two of the main timings, the bios might be adjusting some of the minor timings in the background which could have a knock on effect to the major ones, or be completely independent.

Many people upon hearing this decide they want to turn memory training off, but it's not a good idea unless you are absolutely sure that the current settings are 100% stable. There are a few factors like environment temperature that could affect stability. Memory training can help compensate for this.

 

[sporkster]

I have yet to experience this with my new dell gaming pc when running lol in it.