Memory DDR3

The launch of DDR2 memory has a different impact for Intel and AMD. For Intel, the migration to DDR2 memory was easiest, as the memory controller is included in the chipset, so join a new technology demand only changes in chipsets and boards.

Intel supports DDR2 memory in its chipsets since the i915P, launched in 2004. Initially, the chipsets offer support for both DDR memory in DDR2, so that was charged to the manufacturer choose which standard use. There are also hybrid cards, which support both standards, such as ECS 915P-A, which have two slots of each type, allowing you to choose which standard use. From a certain point, however, the DDR2 memory fell in price and almost all socket 775 boards now come with exclusive support for DDR2 memory.

For AMD, the change was more tortuous, as the Athlon 64 and derived using a memory controller built directly into the processor, developed to minimize the time of access.

On the one hand this is good because it offers a real gain in performance, but the other is bad, because any change in the type of memory used demand changes in the processor and the socket used. It was precisely what happened when AMD decided to make the migration from DDR memory to DDR2. Apart from internal changes in processor and memory controller, the socket 754 was replaced by Socket 939 and AM2 by then, breaking compatibility with the plates old.

With the adoption by the AMD, the demand (and consequently the production) of DDR2 memory has increased considerably, so that prices would fall quickly.

From the end of 2006, prices of DDR2 memory modules (the U.S.) have fallen to the point of going to be cheaper than the regular DDR modules. As always, the changes coming to Brazil with some months of delay, but from the beginning of 2007, the DDR2 memory began to be found at prices lower than DDR here too. As always, high performance modules (like the Corsair Dominator) cost to reach twice as expensive, but in the case of generic memory, prices have dropped a lot.

Another important issue is the popularization of DDR2-1066 modules and also modules for low latency, with timing of 4-4-3-11, or 3-3-3-9. Unlike the initial generation of DDR2 modules, offering more bandwidth, but in compensation working with access times of very high, this second generation of DDR2 modules is undoubtedly faster. The gain in practical use DDR2 memory is not very high because the bus with the memory is only one factor that determines the performance of the PC. But from the moment the DDR2 modules will be cheaper, any gain on the lower it is, is very welcome.

While still used in the DDR2 memory, Intel shook the step and included support for the following standard P35 chipset, officially launched in June 2007.

As logic suggests, the DDR3 memories carry 8 accesses per cycle, compared to 4 hits per cycle of DDR2 memory. As with the previous technology, the drives are made to address underlying, so that there is no need to increase the frequency "real "memory cells.

Initially, the modules were launched in DDR3 version DDR3-1066 (133MHz x 8) and DDR3-1333 (166 MHz x 8), followed by the standard DDR3-1600 (200MHz x 8). The three patterns are called (respectively) PC3-8500, PC3-10667 and PC3-12800, in this case emphasizing the theoretical transfer rate.

Despite the increase in the number of transfers per cycle, the buffers of data are still working just double the frequency of memory cells. That is, the internal frequency (cell memory) of a DDR3-1600 module is 200 MHz and external frequency (of buffers of data) is 400 MHz The memory cells hold 8 transfers per cycle of clock (instead of 4, as in DDR2) and the buffers of data (twice the operating frequency) perform 4 transfers per clock cycle, instead of just two, as in DDR2 modules.

With this, we come to the frequency of 1600 MHz disclosed that, as you can see, is obtained by increasing the number of transfers made by cycle and not by increasing the clock "real" cell memory or the data buffers. If the changes stop there, not DDR3 modules offer huge gains in practice, since the time of initial latency still be the same as the modules DDR2 (since there was no change in the frequency of memory cells). If a DDR3 module operating with access times of 10-10-10-30, the gains would be small compared to a DDR2 5-5-5-15, since we only gain access on subsequent üloved.

To avoid this, the DDR3 modules include an integrated system of calibration of the signal, which would bring considerable improvements in the stability of signals, allowing the use of times of lower latency, although the stability is compromised.

The DDR3 modules also use 8 instead of 4 banks, which helps reduce the latency time in high-capacity modules. They also brought a further reduction in tension used, which fell to only 1.5V rather than the 1.8V used by DDR2 memories.The reduction in voltage causes the electric consumption of the modules fall proportionately, which makes them more attractive for manufacturers of notebooks.

Together these improvements, the time of access "real" modules have been significantly reduced. Instead of working with times of access 10-10-10-30, the initial generation of DDR3 modules are able to work with 9-9-9-24 timing, or even 7-7-7-15.

Nevertheless, many high-performance modules may need higher voltages such as 1.6V or 1.7V to work at maximum frequency. As for DDR2 modules, manufacturers can adjust the tension of operation according to the needs of the project and you can also use higher voltages for yourself when you overclock.

DDR3 modules use the same 240 contacts of DDR2 modules and maintains the same format. The only visible difference (outside of codes and identification tags) is the change in the position of the bevel, which is now positioned closer to the corner of the module. The scarf is just to ensure that modules of different technologies are embedded in cards 

The DDR2 memory took almost 3 years to become popular since the introduction of the i915P chipset in 2004. The DDR3 memory to go through a similar path, with modules costing much more expensive initially and fell to the same level of price of DDR2 modules only around 2009. There is nothing fundamentally different in DDR3 modules that become more expensive to produce, the price is determined primarily by the volume of production.

As in the case of DDR2 memory, the highest transfer rate offered by DDR3 memories results in a relatively small gain in overall performance of the system, only 1 to 3% in most applications, so it is not worth paying much more expensive for DDR3 modules to buy. While they are substantially more expensive, keep buying (and giving) cards with support for DDR2 modules.

In cooking, who is rushed eat raw. In the world of computing the side effects are worse, because who rushed pay more. :)

So far, the use of DDR3 memory is limited to the Intel platform, the chipsets P35, G33, G35 and X38. In the case of AMD, adoption is still uncertain, as the support for DDR3 memories depends on a change in the new integrated memory controller, and (possibly) in the socket used by the processor. Moreover, history shows that Intel is almost always the first to adopt and popularize new technologies (such as DDR2 memory and PCI-Express), while AMD generally prefers to wait until they are mature and the cost has fallen.