Newb memory question

This is probably a really newbish question, but oh well.

I know when it came to A64 overclocking RAM that overclocked well wasn't really that important. For example, if you were on a budget, you could get away with just buying cheap value RAM and running a divider. Thus, the memory didn't really hold back your overclock.

However, now I'm looking at building a C2D rig and all the overclocking guides I read all talk about overclocking the RAM 1:1 with the FSB. If this is the case I'd imagine RAM it is actually quite a bit more important on a C2D rig and could hold back your overclock?

Any insight on this and maybe some suggestions on what I should be looking for in RAM? Maybe just a quick rundown on what's good? I'm not really looking at suggestions for a particular set of sticks, but more the tech stuff. Like should I be getting DDR2-800, what types of IC's should I be looking for, etc.? I found this RAM list that lists the IC's.

http://ramlist.ath.cx/ddr2/

If you are looking for ddr2, crucial ballistix pc6400 are decent and they still have much headroom for oc. http://www.newegg.com/Product/Product.aspx?Item=N82E16820146565

What about DDR2-1000? Will you have a 1333FSB C2D?

Thanks for the memory suggestion, but right now that's not what I'm looking for. I'm more of the learning type. I'm not really look for "This RAM overclocks well, buy it." If something sets it apart from others I want to know WHY, not just because it overclocks well. I also want to know if you're on a budget, does it matter? Like I said in my first two paragraphs...

I know when it came to A64 overclocking RAM that overclocked well wasn't really that important. For example, if you were on a budget, you could get away with just buying cheap value RAM and running a divider. Thus, the memory didn't really hold back your overclock.

However, now I'm looking at building a C2D rig and all the overclocking guides I read all talk about overclocking the RAM 1:1 with the FSB. If this is the case I'd imagine RAM it is actually quite a bit more important on a C2D rig and could hold back your overclock?

Click to expand...

So could somebody explain to me if there's any difference between A64 and C2D in terms of memory? Also, RAM suggestions would be nice, but I'm not looking for the typical "This RAM is good" responses.

What about DDR2-1000? Will you have a 1333FSB C2D?

Click to expand...

I'm really not sure. I may not even upgrade anytime soon. It was just kind of a thought and I came to the realization that when it comes to memory, I don't know crap. So this topic is more for me learning more, rather than a memory recommendation. If I do decide to finally make the plunge and go with a C2D setup, it would probably be something cheap like an E4300/E2160 since I'm more of the "best bang for your buck" type of person.

Thanks in advance for any responses!

BrutalDrew said:

So could somebody explain to me if there's any difference between A64 and C2D in terms of memory? Also, RAM suggestions would be nice, but I'm not looking for the typical "This RAM is good" responses.

Click to expand...

BrutalDrew,

I'm slowly testing for my overclocking guide to add C2D to the A64 stuff that is already there so don't have a complete detailed answer for you. But as you know there are hardware differences between the two systems, e.g. C2D uses Northbridge as memory controller while the A64 has an onboard memory controller.

However in terms of how they behavior and how you overclock them, I don't see that there is much difference. You do get higher bandwidth with higher FSB on both (see attached for the C2D results). What I don't have yet is how this effects performance (the real question of course). I can say that if I hold the cpu's speed constant and change FSB for different cpu multipliers, I don't see major difference -- I don't have statistics yet so can't make too strong a statement here.

By the way, the "hole" in the data is due to bad behavior of my motherboard. This seems to be typical of a number of 775 motherboards, not just the eVGA 680i I'm using.

Sorry that isn't a complete answer to your question but best I can do for now. Hopefully some else can add additional info for you.

So here is the effect on performance measured using Pi (the 1M test) when memory speed and CPU speed are held constant (several different speeds shown). In this case the memory speed is 1000Mhz (stock for my PC2-8000), double data rate of course.

Thus you see that increasing the northbridge speed alone while memory and cpu speeds are constant has no effect on performance in the ranges tested. That is the result that one would hope to see.

I'll do some more testing when I get a chance.

Final bit of info. This shows the gains of higher memory speed at a fix CPU speed of 3.5Ghz using Pi 1M as a metric. Keep the y-axis the same so you could see the changes relative to plot above easier. Memory timings are held constant in these studies. Separate study not shown indicates timings help but aren't major players unless you change them greater than increments of 1

As you can see, higher memory speed does improve performance. But the difference between memory at 800 and 1000 is minimal.

From these 3 sets of studies, I conclude
1) CPU speed is most important and greatest payback
2) memory speed is next but not as significant
3) FSB alone has minimal value.

Anyway, those are basics with the regions I've tested.

sPi likes tight timings/optimized timings too.

JamesXP said:

sPi likes tight timings/optimized timings too.

Click to expand...

Not a great deal of payback for tight timings for C2D and DDR2. Maybe I'll add that study later.

So here is the last bit of info on memory timings. Interesting detective story associated with the initial part of investigation.

The table shows results using my Mushkin that is rated at 4-5-4-11. Two sets of tests associated with memory timings are shown. One test ran Pi using 1M digits with 5 samples and the other 2M digits with 10 samples. The 1M case is often used as a reference test by a number of folks including myself.

To make a long story short, the results using 1M digits and 5 samples do not produce results that we would expect. For example 5-5-5-12 is faster than 4-4-4-11, very questionable result. However if you look at the standard deviations of the tests (called sigma in table), you see that the uncertainty associated with the numbers indicates that while the results aren't what we expect it is a reasonable result. (If you have a collection of data from a Normal Distribution then approximately 66% of the data should fall within one standard deviation of the mean.) Thus 66% of the average values would lie between 14.7689 (average - sigma) and 14.8835 (average + sigma) if I had an infinite number of tests. My result for 5-5-5-12 lies within that spread so I can't say it is wrong. Thus my metric as applied doesn't appear to be good enough! You always have to make sure you're measurement reflects what you think it does.

From this I required a better metric so decided to try Pi with 2M digits and taking 10 samples. Took half a day testing to run those cases in combo with a few honey-dos too. Anyway now looking at those results, they are more typical of what we expect, i.e. the ordering of what is important is typical for the various settings.

So what is the maximum improvement that I might reasonably expect with better memory timings? 66% of my maximum performance improvement will be less than 1% for a single setting change to all timings, i.e. going from 5-5-5-12 to 4-4-4-11. I get this number by doing this (37.4248+.0.1108)/(37.2565-.08831)=1.0099 or 1%.

How does this relate to better memory? If I had DDR2 rated at 800 and went to memory rated at 1000 using results I've not shown you, (15.019+0.017)/(14.854-0.034)=1.0146 or 1.5%. Thus better (faster) memory is better, in this case 800 at 4-4-4-11 will be better than 1000 at 5-5-5-12. Of course taking bigger steps in memory speed or wider spreads in timings would have to be tested, guessing the answer isn't worth much.