Memory Timings:

Don't matter much

By HMJ

Virtualjesus23@mail.com

The last time I started a system build one of the last pieces of hardware I bought was the RAM.  Buying memory is difficult because of all the little things associated with it.  You have multiple types of RAM, ram speed, and of course memory timings, which is what I am taking a look at today

To take a look at how RAM timings affect overall system performance, I have done some tests comparing and contrasting the performance increases and decreases associated with changing certain memory timings while leaving others alone,  in order to test each individual timings effect on the memory as a whole.  This sounds confusing I know, but wait, it will all be made clear.

For these tests I used Twin 512 MB sticks of DDR3200 Crucial Ballistix, in a DFI Lanparty Ulta-D mainboard, with an AMD Athlon 64 3000+ Overclocked to 2610MHZ (290*9), and an ATI x850 XT.  Just as a note incase your curious my memory is running at DDR 266 speed utilizing the onboard memory divider at 133 MHz.

 Note: For this article I am going to use the pretty well standard labeling or ram timings as such:

2.5, 2, 2, 6 (random example)

Cas, Trcd, Trp, and Tras respectively

Note: if you dont know what those mean, neither do I, but thats probably how they are labeled in your bios.

CAS (cas latency)

Doom III run at High Detail 1280*960

Sad to say, despite my standard opinions regarding memory timings, especially CAS, non of these fluctuations are outside of the range of random performance difference from one restart to the next. Maybe TRCD will fair better.

TRCD

Although slightly more of a difference than shown by the CAS, it is nothing extra ordinary and a similar effect could be garnered in any system via a bad restart or a few extra processes running in the background.

TRP

I dont feel the need to address this set of data, sufficient to say that it falls under a similar category as the other two timings.

TRAS

Ok still nothing, just one thing left to try, maybe the over all effect of the timings will be cumulative and if you have all crappy timings, performance will sufferlets see.

Here we see a little something but nothing all that big especially on the non synthetic bench in which we see no change whatsoever.

Mmmm'k this tells me one thing, memory timings dont mean much of crap.  Due to the data I have collected I have decided that as long as you buy good brand name ram with good chips you should be good to go.  Now this may be a forgone conclusion considering I havent gotten to test any cheaper ram as compared to my Ballistix which is one of the best brands out there. Cheaper ram may very well suffer from massive performance loss as memory timings are lessened.  Additionally my rather substantial over clock may have something to do with the lack of performance loss, but I doubt it; usually over clocking makes the rams job an ass load harder not a bunch easier.

 Remember Kids the moral of the story is; dont pay out the ass for timings, instead save your money for good name brand stuff.   Youll do better in the long run.

HMJ