There seems to be some misunderstandings about dual DDR compared to RDRAM and to what degree faster memory helps you.
DDR Stick = RDRAM Stick
A single DDR stick has exactly the same theoretical bandwidth as a single RDRAM stick at a given speed.
So a dual-channel DDR setup running at the same memory speed as a dual-channel RDRAM setup should perform about identically, and they pretty much do.
The only advantage dual DDR has over dual RDRAM is that you can run DDR chips full-tilt at higher speeds than RDRAM chips. For instance, reviewers have been able to run dual DDR configurations at 180-190MHz memory speeds, which you just can’t do with RDRAM chips running at full speed.
So dual DDR only has an advantage over RDRAM when you crank up the memory speed to levels where RDRAM can’t go.
Memory Doesn’t Help Much, Period
RDRAM PC1066 systems provides significantly more memory bandwidth than single-channel DDR systems. Nonetheless, the performance difference is relatively small (figure on average about 7%, more for some games, less for office apps).
What a dual-channel DDR configuration does is essentially bring DDR up to dual-channel RDRAM levels, so the performance gain is about the same as the difference between single DDR and RDRAM, fairly small.
This should not surprise or disappoint you.
Nor should you be surprised to find that running dual channel DDR at 200MHz doesn’t help a ton, either. That should get you on average something a bit less than 5%, and again, more for games, less for conventional office apps.
The reason for this is that CPUs and computer systems are designed wherever possible to avoid having to go to main memory to work on the next piece of data. CPUs are designed to get that data ahead of time to be ready when it’s needed.
This low scaling-up of performance by adding faster memory is nothing new. Memory has never provided a big boost to overall system performance. For instance, when memory went from 66MHz to 100MHz, that only contributed about 3% to performance. When memory went from SDRAM to DDR, the level of improvement was about 7%
This is one of the biggest myths in this field, and it’s just incorrect when it comes to the real world. The only place where faster memory gives you big improvements is in memory benchmarks. That’s it. When you get to applications and games, you’ll see some improvement, but nowhere near the increase in memory speed or bandwidth.
Are the improvements significant? Depends on what you mean by “significant.”
If you went from a single-channel 133MHz system to a dual-channel 200MHz system, we’re looking at about an 11%, maybe somewhere between 15-20% for some games.
If you want from a single-channel 200MHz system to a dual-channel 200MHz system, we’re looking at about a 7% improvement.
Do you consider that significant or not? Reasonable people, especially those who take their wallet into account, can disagree on this.
It’s not a no-brainer, it doesn’t have big bang for the buck (unless you EBay your “old” equipment all the time). It’s the kind of upgrade you do when you’re upgrading primarily for another reason.
Many may feel it would better to wait for the next generation of Intel boards due in the spring. You can read some more about those in this Inquirer article
Others more concerned about their wallets may decide they want to wait (probably quite a while) until DDR-II becomes common, maybe by the end of the year, before deciding on what kind of big upgrade to make (which might well be an AMD/Intel choice).
Any of these can be reasonable choices.
What isn’t reasonable is buying with little or bad information on what improvement you’ll get for your money.
Be the first to comment