nOOb question

[GeorgeKellow]

What are ram timings? Do I need to mess with theM? 

[Bob Jim]

No.

[FloRolf]

nah. they make pretty much no difference

[Enderman]

no

 

https://en.wikipedia.org/wiki/CAS_latency

[Prysin]

nah. they make pretty much no difference

they make A LOT of difference if you know how to tune them

 

@MageTank

can explain this better

[FloRolf]

they make A LOT of difference if you know how to tune them

 

@MageTank

can explain this better

yeah i'd like to know that one programm that it does a little difference.

[Godlygamer23]

They are defined as latency with certain things in memory.. You can find out more information here: http://www.hardwaresecrets.com/understanding-ram-timings/

[vanished]

they make A LOT of difference if you know how to tune them

 

@MageTank

can explain this better

I really want to see this too

[MageTank]

yeah i'd like to know that one programm that it does a little difference.

 

I really want to see this too

While "A Lot" is a large order to fill, they do indeed make a decent difference. Depending on what we are talking about (DDR4, DDR3, etc) the differences vary.

 

That being said, I'll share what i know on this subject.

 

First of all, Primary timings, while the most advertised, are not the "be all, end all" timings. Sure, getting these low is nice, but if we are talking raw bandwidth, they won't be doing that much for you in the long run. Bandwidth comes from your tertiary timings. The timings that are often overlooked and ignored even by ram manufacturers within XMP profiles. Now, how much of a difference are we talking? Well, a few factors govern this. However, the most important is your CPU IMC. Tertiary timings are directly related to the IMC and VCCIO/SA voltages. Which is why optimizing tertiary timings can take stress off of your IMC. Here is the difference between optimized tertiary timings, and untouched tertiary timings.

 

Before:

 

After:

 

Now, that is an 8GB/s difference. Notice memory speed, CPU speed, and primary timings are exactly the same. 20% difference in read bandwidth just from tertiary timings. Now, let me explain why this difference is so vast on my kit. My kit is normally a 2133 CL13-13-13-35-CR2 kit. That is its stock XMP profile. This means when I overclock the ram speed, change voltage, etc, the board will auto-train the rest of the values. Depending on the board, your autot-trained values can vary drastically. For exmaple: EVGA and ASUS boards are praised for having great auto-training on timings, but ASRock (board i use)... not so much. So if I ignore my tertiary timings, and just focus on primary + raw clock speeds, I am still losing 20% of my potential bandwidth. To put this into perspective, 400mhz of memory clock is roughly 6.2GB/s of peak theoretical bandwidth. To lose 8GB/s bandwidth, you are looking at a difference of about 550mhz. Meaning in order to hit the bandwidth in my "After" image with the same timings from the "Before" image, you would need 3750mhz memory. You people had to have seen the videos comparing 2133 to 2666mhz memory from DigitalFoundry. You saw how much of a difference 500mhz made for them. Imagine the difference 550mhz worth of bandwidth will make if you just leave it lost within your tertiary timings.

 

In case you did not see, check my signature and the following links: 

 

http://www.overclock.net/t/1487162/an-independent-study-does-the-speed-of-ram-directly-affect-fps-during-high-cpu-overhead-scenarios

 

Yes. Bandwidth does impact gaming performance. To the tune of about 10% higher minimum FPS. Not only in my results, but in the DF videos, the OCN thread (which btw, came out before DF and my tests) and even a few forum users, such as @SteveGrabowski0. 

 

To ignore all of this information without testing it yourself would be silly. If you have XMP ram, simply default to JEDEC speeds for your kit, and monitor minimum FPS. Then load your XMP profile, and do the same test. It's that simple. 

 

For iGPU's, this raw bandwidth will matter even more, as it won't just be a crutch for situations with high CPU overhead. It will be used for pretty much every situation that the iGPU is used in. Also, any program that uses heavy compression will show a large difference in performance from faster memory. 

 

Now, on to OP's question.

 

 

What are ram timings? Do I need to mess with theM? 

Need? No.You do not. If you want to, sure, you can get some extra performance out of your primary, secondary, and tertiary timings if you take the time (put aside a few days, it will take a while to stress ram to make sure its really stable) or you can just buy a decent XMP kit, and do it that way. Ram timings are far more complicated than traditional overclocking of CPU's and GPU's. You cant just change one thing and test it at a time. Some timings rely on each other, and to change one, you have to change the other or else you will not post. It is a fine balance that is hard to explain, as IMC quality, ram quality, and even board manufacturer all impact the timings names, their min/max values, and how much tolerance you will have when going outside of certain thresholds. DDR4 and DDR3 differ, in that DDR4 is far more forgiving on timings, and will still post even if timings are not exactly great, the board will just override your manual input and train it to a value that can boot.

 

I would say, unless you have the time, and are doing things such as heavy compression, or gaming on an iGPU, I would just stick with XMP settings. If you want the most performance out of your hardware, then yes, thoroughly overclocking your ram can give you more performance.

 

Hope this helped, and sorry for over-complicating it.