Tuesday, March 27, 2012

Why overclocked RAM is not worth the price and effort

I often see people spending huge amounts of money on RAM, and there is a lot of misinformation regarding RAM. So I thought I'd explain a few things regarding RAM and why you should not worry about it.

You'd think that RAM with high frequencies and low timings would be awesome performance wise, right? Wrong.
The truth is, there is next to no performance difference between for example 1333MHz CL9 RAM and 2133MHz CL7 RAM.

LinusTechTips made a video about this and the results he came up with were these:


As you can see, the difference is minimal. When it comes to computer benchmarks, the margin of error is usually considered to be about 5% These results are pretty much within the margin of error. SuperPi is a good synthetic benchmark for CPU and RAM so the difference in it should be greater than your average program, but the difference is next to none.

3DMark is a more "gaming" like benchmark. The good thing about this is that it resembles regular programs more than SuperPi. The drawback is that other factors such as GPU and CPU matters more, and the score will fluctuate (even if you run it with the exact same hardware and settings) more than SuperPi. Anyway, the score is once again pretty much the same.

How do different RAM perform in games? Here is a benchmark from Anandtech:

Again, next to no difference.

But how about heat? Overclocking RAM often have huge heatspreaders for cooling.
The truth is, the main use of those big heatspreaders is attracting customers and making them believe that the RAM is special. In reality, they do next to nothing. RAM doesn't really get hot. They don't even have heat sensors like your CPU, GPU and even motherboard do, because they don't need it.
Big heatspreaders are actually worse than small ones because the big ones can sometimes make installing big CPU coolers impossible or very hard.

Okay so the heat is not an issue, and the performance is pretty much exactly the same. But what about the quality?
This is a bit tricky actually. RAM is binned into 4 major categories.

Major brand chips
These are the highest quality chips manufactured. They are fully tested and have the manufacturer (mostly Samsung, Micron and Hynix) name on them.

eTT chips
These are basically the same quality as the major brand chips, but these chips are produced in much higher quantities and often sold premounted on modules unlike the major brand chips which are sold separately.

uTT chips
These chips are usually not fully tested and are not really that high quality. Very cheap and usually functions just fine.

Downgraded chips
These are the lowest binned chips and the manufacturer basically says "well, these chips do not perform like they should, but they work just fine if you increase the voltage or lower the speed and/or timings".
Usually works perfectly fine and are very common in most low end kits.

So what does this mean? It means that high and low quality kits are not the same quality wise. But does it matter? Not really. The chips should perform the same so the difference might be that one RAM module lasts 15 years compared to 10 years. Cheap RAM will most likely serve you just as well as high quality RAM.
Also, RAM is so cheap these days so even if your RAM were to break, you could just RMA it, or buy another kit.

So in the end, just get a cheap RAM kit. It will be just as good as an expensive kit.

1 comment:

  1. Hmm, interesting, I wasn't aware there was such little difference between them like this.

    Heh, I'm still running on 2GB right now but I play almost any new game on high anyway. I'd still like to upgrade regardless though, it will use nearly all of my RAM up in a game so I'd like to have the breathing room of more RAM.

    ReplyDelete