DDR3-1333 I think I understand that something like 5 of 1gig = 4 of these 1gig DDR3-1600 Is this a practical way of seeing it?
Somebody please correct me if I'm wrong.. I think "memory speed" is mostly a marketing scheme and a TOTAL CROCK! The specs indicate a MAXIMUM of data transfer/sec. I understand that. If you're gaming at high FPS and high resolution, perhaps this makes a difference. But what about when the data stream is but a mere TRICKLE of the RAM's maximum capability? As would be seen in a trading environment.... My contention is that "RAM speed makes no difference at all to trading... because the data stream is such a small percentage of the RAM's capability".... That is, beefing up your rig to DDR3 1600 is a total waste of time and money vs. DDR2 800... That's my hypothesis until somebody changes my mind....
i think your right. only need a huge pipeline if you have a ton of water to pass through. you don'r want it to damn up. but the 'trickle' flows just as fast in a smaller pipeline as it does in the larger. or does it? is this analogy valid?
Probably a fair assessment if you are talking charting, which in the bigger scheme of things really is just processing a trickle of data. These days, for charting, just about anything should do the job. However if you are processing a large amount of data in time series analysis eg backtesting, simulations etc, it may be another matter. Even then, the use of expensive memory may only yield incremental improvements. More important, IMHO, is the design of the memory subsystem eg LII cache size and design etc. I would guess that the new Intel i7 probably has an edge here because of the on-chip controller and the three channel memory. Possibly better to have an i7 with cheap memory rather than a C2D with expensive. I once compared the performance of some backtesting code I wrote in Java on an AMD Sempron and Intel C2D. The C2D was twenty to thirty times faster. I attribute a lot of this to design of the memory system, LII cache size and so on.
Doesn't seem like it would make a difference even for video editing. After all, the video speed is only about 30 fps, right?
This comparison did not look too favorable for ddr3 to ddr2. Although, the theoretical speed of ddr3 is much faster due to smaller processing technologies, there are still limitations due to overall system interaction/configuration. The review found a much greater improvement was due to the newer controller chipset, rather than the memory module differences. http://www.anandtech.com/memory/showdoc.aspx?i=2989&p=1
It should be noted that the more recent processors (for instance 45nm intel E series (wolfdale)) allow DDR3 to stretch out a bit more in that they allow higher clock speeds and higher FSB speeds for linking and syncing ram in a 1:1 ratio (resulting in higher bandwidth). But generally I agree - where a northbridge is involved, you will not fully realize the bandwidth capabilities of DDR3. DDR3 will really stretch its legs in the latest chipsets (i.e. X58) where top of the line DDR3 (i.e. 50nm Elpida) is currently capable of bandwidth >30GB/s (that is fast!)