Like so many successful technologies, DDR memory seems to have been with us forever. At least, thatâ€™s how it seems from within the industry. On the high street though, things move a little slower. Remarkably it was only six short months ago that a well know PC super store referred to DDR as an â€œemerging technologyâ€, as I found out when I got chatting to one of its service counter guys about which brands it stocked. And in its chain of stores it probably still was.
The pace of technology has always been brisk, but it seems the past twelve months have had even the most hardened hacks reaching for the anti-nausea pills to fend off the effects of spinning heads . Weâ€™ve moved from 32 bit to 64 bit processors, Socket 940 was joined by Socket 939, AGP morphed into PCI-express, Intel moved the pins from its processors to its sockets, and now, despite the fact that it seemed to be scaling well with maturity, it has been decided that 400MHz is plenty fast enough for DDR and we must all rush out and arm ourselves with a whole new breed of memory known rather unimaginatively as DDR2.
Developed by JEDEC and championed by Intel, DDR2 isnâ€™t just DDR running at higher speeds with the number â€œ2â€ tagged on the end. Yes it runs faster, starting at the 400MHz threshold where DDR currently ends, but there are other important technologies implemented into its design. Where the early offerings failed to impress though was in their latency ratings. Faster though the memory was, much of that speed was wasted as the module sat idle waiting for data to be read or written, more so than was the case with DDR. This in turn meant that lower speed DDR2 parts were often slower in real terms than their DDR predecessors. Fortunately, weâ€™re now seeing frequencies increase and latencies begin to fall, a combination that will redress the balance and enable DDR2 to shine.
Physically DDR and DDR2 look very similar; in fact when heat spreaders are fitted the only obvious difference is in the number of gold contacts along the lower edge. DDR sports 184 of these contacts while DDR2 has 240. The small reference notch is also in a slightly different position, which should make inserting the wrong type of memory impossible.
Other changes include the way the memory chips are attached to the PCB. DDR2 now makes use of FBGA (Fine Ball Grid Array) while DDR uses a TSOP-II (Thin Small-Outline Package). The difference between these two is that TSOP is forced to run its traces to the outer perimeter of the chip while FBGA routes its signals using tiny pads on the underside of the chip. These pads allow much more flexibility with trace lengths as they can be sited much closer to the signal source. The chip package also tends to be smaller when manufactured this way. Kingston became known for using BGA technology even before the manufacture of its DDR2 modules.
If that wasnâ€™t enough, other new technologies were introduced to enable the new memory chips to run even faster. These include On-Die Termination (ODT) and Off Chip Driver (OCD) calibration. To keep it simple, On-Die Termination takes the resistors that reside at the end of each signal line from the motherboard PCB and places them on the chip core itself. Keeping the resistors closer to the source of the noise makes for much cleaner signals.
Off Chip Driver calibration is a little more complicated but in essence it serves to clean up signal integrity using various methods to smooth out voltages and apply a reference signal.
Add to the mix greatly reduced power requirements of around 1.8V for DD2 as opposed to the 2.5V weâ€™re used to from DDR and a 4-bit prefetch in preference to a 2-bit prefetch and I think itâ€™s fair to say that DDR2 is a very different beast to DDR.