Quote:
|
Originally Posted by antikristuseke
Well actualy DDR memory is pretty much near its theoretical limits with AMD K8 architecture systems. The theoretical limit of DDR400 is 3.2gb a second and 6.4gb a sec in dual channel mode. I have DDR500 ram sticks which reach 7.5gb of their 8gb theoretical maximum bandwith use. Besides nowadays there are low latency DDR2 modules available for reasunable prices making the advantages of DDR near non existant in that area and giving you more bandwith. Right now for gaming, that is the best way to go. As time passes DDR3 latencies will come down because production technologies and procedures mature and with that prices allso come down.
|
Sort of. I guess I have to base my opinion on dual cores which most people have these days, but I bet if you ran testing on it, your reads and writes would probably never hit the full 6.4 GB/sec on a given day. A Cl 2 in latency is going to greatly outperform any speed advantage that any additional pipe size is going to get you for real world and real gaming work. I've yet to see actual DDR2 modules be below CL4 for the most expensive type available, and that is only half the speed of a mid range priced DDR module. That's a significant loss for many small transactions as done by system memory, giving DDR a significant performance edge in doing normal work on any given day.
People don't seem to realize that latency is more important than throughput when it comes to system memory. Both are important of course, but you have to look at what your RAM is doing all day to figure out why you want better latency over throughput. It's like giving a dragster 20,000 horsepower, with only 100 ft/lbs torque. It can go real fast with a great top speed, but can't accelerate for crap.
-S