How much memory does Vista really need? What about Windows XP? And what the heck does DDR mean? How much money will a memory upgrade cost, and where do you buy it? We’ll answer these questions and more, briefing you on the essentials of PC memory technology and saving you a great deal of money and hassle. Whether you’re building a new system from scratch or upgrading, you’ll find PC Magazine Labs-tested advice you won’t get anywhere else.
If you know what you’re doing, now is a great time to buy more memory because prices have never been lower. If you don’t know what you’re doing, though, you could end up spending a bundle on memory you really don’t need. (More on pricing later in this story.) Let’s get started with a refresher on memory basics. – next: DDR and Other Inscrutable Acronyms
DDR and Other Inscrutable Acronyms
Any computer that you bought in 2001 or later uses a type of dynamic random access memory
(DRAM) known as double data rate, commonly referred to as DDR. It simply refers to memory in which two data items are transmitted with each clock signal. It replaced single data rate SDRAM. (The S stands for synchronous.) DDR memory is also synchronous, and you may sometimes see DDR memory referred to as DDR SDRAM.
DDR also replaced RDRAM, so called because Rambus Corp. developed the standard. Rambus memory offered substantially higher memory bandwidth than SDRAM. However, RDRAM was always more expensive than SDRAM, and DDR’s lower cost and equivalent throughput eventually canceled out any performance advantages RDRAM offered.
In 2003, a new generation of DDR memory, DDR2, appeared with Intel Pentium 4 processors. Because AMD Athlon 64 CPUs had integrated memory controllers, DDR2 support didn’t appear in AMD processors until the launch of a new lineup in 2006 that used the AM2 socket.
DDR2 differs from DDR in that the bus carrying the data runs twice as fast as the actual memory clock. For example, older DDR memory that is clocked at 200 MHz clocks the I/O bus at 200 MHz and can transfer data at 3.2 gigabytes per second. Why? Well, the clock speed determines the speed of the I/O bus; that’s why it’s called synchronous DRAM. The calculation for determining transfer rate (3.2 GB in this case) is:
(bus clock * data rate * data width
Of memory module) ÷ 8
Meanwhile, DDR2 memory that runs at the same 200 MHz runs the I/O bus at 400 MHz. Each I/O bus clock cycle still carries two items of data, but the increased bus clock effectively doubles the maximum throughput to 6.4 GBps.
The various flavors of DDR2 are confusing because JEDEC, the standards group for the semiconductor industry, has issued variations of the specification. For example, some high-performance memory is rated at specs beyond the official JEDEC rating.
To hit those high specs, computer users need to go into the BIOS to increase voltage beyond the standard 1.8V. Increase the voltage and you increase the throughput. We’ve seen people buy DDR2 memory that manufacturers say can run at up to 1,066 MHz and be surprised that it runs only at DDR-667 speeds because memory is shipped at officially rated speeds. (The numbers following the hyphen represent the bandwidth in megahertz.) For example, DDR-400 has a maximum bandwidth of 3.2 GBps, which is also 3,200 MBps. So you’ll see DDR-400 also called PC3200.
In PC Magazine Labs, we’ve seen DDR2 max out at DDR2-800 speeds at the rated 1.8V, but an effort is in the works to create a JEDEC standard for DDR2-1066.