ram faster than 1:1 wasted if don't oc?

[foruga1]

I'm new to this, but if I want to build a system with the e6850, then, if I'm not mistaken, dual channel ddr2 667 gives a 1:1 ratio between the cpu and memory busses.

I've seen some reviews where faster ram gave at least some boost in performance. I don't know if they overclocked the cpu to 'catch up' with the faster ram.

I hope this isn't a silly question, but...


without any overclocking, would I see any speed increase in combining the e6850 with ddr2-1066 instead of ddr2-667?

It seems that the theoretical formula suggests that in the e6850 + ddr2-1066 combo, the cpu is the bottleneck.

Please correct me if I'm wrong, and explanations are appreciated.

You can frame answers relative to any of the standard types of benchmarks.

 

[Mondoman]

The main thing is to have *at least* a 1:1 ratio, so the memory bus's throughput is not less than the capacity of the CPU's FSB. You're right that RAM running at DDR2-667 will saturate the throughput of a 1333MHz FSB.
You will likely see a few percent increase from using faster RAM -- it's your decision on what's worthwhile.

 

[Kari]

Or you could drop the multiplier on the proc and use higher fsb to take full advantage of the faster memory while keeping the proc close to the stock speed...
well maybe not full advantage, 1066 would need a fsb of 533, and that might be too much for the motherboard to keep up, or the proc for that matter

 

[foruga1]

Thanks for replys.

kari: thanks for the info. but the point is to keep the cpu & ram at stock. given that I have no oc experience (yet) I can't justify any risk, my first build will be too important. I will keep your info in mind for future though. The only thing I can really justify is setting the board to accept 1066, like you have to do on some of the asus, gigabyte, and evga mobos.

mondoman: thanks also for the info. Do you have any explanation as to WHY ram faster than 1:1 will give any performance increase? Can you point me to any tests that show this? I only know of one:

http://www.xbitlabs.com/articles/memory/display/
core2duo-memory-guide_ 5.html

Their methods aren't entirely obvious to me, but on their first set of benches I think everything was at stock.

Any explanation as to why ram faster than 1:1 gives even a little boost?

The reason I'm stuck on the small details is that I need to justify extra cost. It seems faster ram can't hurt when cpu\ram is at stock, and only allows for oc overhead.

I may be completely off, but the only possible explanations I can imagine are:

the 1333 and 667 numbers aren't exact, so going to 800, 1066 ram makes up for this

the ram may from time to time 'skip' or 'trip' (like when you run and slide your foot on the ground--you don't fall, just slow down a little for a second), and the faster ram compensates for this

Thanks!!

 

[Mondoman]

The reason for the small speed increase is presumably the (normally) shorter latency time of faster RAM. RAM is not always addressed sequentially; shorter latencies will become more important the less RAM is addressed as sequential blocks. Here's an article with some data (look at the result tables; the text analysis is a bit wrong): http://www.madshrimps.be/?action=getarticle&number=1&artpage=1962&ar ticID=472

 

[foruga1]

Good article, good explanation ... many thanks!

 

[BushLin]

Some processors will take a big performance hit if running on a asynchronous (not 1:1) which means you have run a substantially faster memory clock for it to be of benefit.
For Core 2 it seems that a synchronous bus (1:1) doesn't have a massive impact on performance, it's also the case that more expensive/faster memory in general results in such small performance gains that you're better off buying the next processor model up rather than sinking extra cash into a 1-3% increase in overall performance.
Those new to the game might look at the current lineup of kit, see how these variables affect performance and think of it as a rule; however, the next set of cpus and chipsets may react differently and the goalposts will be moved again.
Higher speeds and lower latencies will always be faster but how much of a bottleneck it's opening up will determine how much difference it makes overall, now Intel has a very good cache on it's processors the dependency on memory is reduced (and gains from spending loads of cash on it)