Despite the optimizations, the memory modules SDRAM still doing just a transfer per cycle, the simplest way possible. After a long period of initial cycle, the memory cells deliver a reading of data per cycle, through the output buffers and is dispatched through the data bus. All components work in the same frequency:

The DDR memory implementing a new trick, which makes them capable of performing two transfers per cycle and are almost two times faster than the SDRAM memories, even keeping the same frequency of operation and the same basic technology. Come hence the term "DDR", which means "Double Data Rate", or dual stream of data. With the release of DDR memory, the SDRAM have been called "SDR" or "Single Data Rate."

The DDR memory chips include additional circuits, which generate commands to access and receive data for the readings twice per clock cycle, performing an operation at the beginning of the cycle and another in the final. As the same tracks are used to perform both transfers was not necessary to make major changes or in modules, or on the motherboard.

Nevertheless, the memory cells themselves are still operating on the same frequency. In a DDR-266 module, for example, they operate at only 133 MHz, similar to a PC-133 module. The leap of the cat is to make each of the two commands for reading (or writing) are sent to a different address in the same line. The two readings are sent via the data bus in the form of two separate shipments, one at the beginning and the end of another cycle of clock:

The biggest problem is that the initial cycle still takes the same time as the SDRAM memories, so that the gain appears only in consecutive readings of various sectors and the transfer rate never really enough to fold, varying considerably according to type of application used.

The timing for a burst of 8 readings, using DDR memory, it would be 5 -½-½-½-½-½-½-½ (8.5 cycles) rather than 5-1-1-1-1-1-1 -- 1 (12 cycles) as a module SDR. The difference is less in smaller bursts of two or four readings.

Nevertheless, the DDR memory is just a great deal, made possible because of perceived performance gains without a considerable increase in cost. Precisely why they are popular quickly, replacing the SDRAM memory in a space of less than one year.

DDR modules can be sold both by their frequency of operation as its second transfer rate.

DDR-200 (100 MHz) = PC1600 
DDR-266 (133 MHz) = PC2100 
DDR-333 (166 MHz) = PC2700 
DDR-400 (200 MHz) = PC3200 
DDR-466 (233 MHz) = PC3700 
DDR-500 (250 MHz) = PC4000

As in the case of SDRAM modules, memory modules are DDR CL2 and CL3, with CL2 in the time of the initial access is reduced by one cycle, resulting in a small gain in performance. As the DDR perform two operations per cycle, were also CL2.5 modules, which are in the middle of the road.

The specifications of the modules indicate the maximum frequency for which its operation was demonstrated. Nothing prevents you to use the module to a frequency lower than that specified, you can use a DDR-400 module on a motherboard configured to work at 133 MHz, for example, but in this case there is no gain in performance with respect to a DDR-266 module, except for minor differences related to the CAS or the time value of the two modules.

Almost always, you can also use the module to a frequency slightly higher than that specified, making overclock. The DDR-400 module could then operate to 215 MHz, for example. Do overclock without increasing the voltage of the memory does not bring danger to the modules (even the long term), but you also have no guarantee of stability.Normally the modules CL2 or CL2.5 support the overclocks better, since the driver has more room to increase the timing of modules to compensate for the increase in frequency.

By mixing two modules of different specifications, it is necessary for low level, using the frequency supported by the slower module. Just so it is not always appropriate to use the old modules to upgrade, because you just using the new sub-module.

In cases where you have, for example, a module of 128 MB of DDR-266 memory, the penalty is worth more to sell the old module and buy a module or DDR-400 DDR-466 for 512 MB or more, than use the old together with a new module.Check only if the motherboard can adjust the frequency of asynchronous way, without increasing the frequency from the FSB (so that you can use the correct frequency for the module, without changing the frequency of other components).

Still, almost all the memory modules DDR SDRAM or have a chip for identification called "SPD" (Serial Presence Detect), which stores the identification codes of the module, details on the frequency, time of access, CAS latency and other specifications . This information is displayed by programs of identification, such as CPU-Z and Sandra. On Linux, you can read the information stored on the chip using the script "decode-dimms.pl" (you can find it using the command "locate") as part of the package "lm-sensors." He returns a long list of information about each of the modules installed on the machine, as in this example:

Memory Serial Presence Detect Decoder 
By Philip Edelbrock, Christian Zuckschwerdt, Burkart Lingner, 
Jean Delvare and others 
Version 2.10.1

Decoding EEPROM: / sys/bus/i2c/drivers/eeprom/0-0050 
Guessing DIMM is in bank 1

---=== SPD EEPROM Information ===--- 
EEPROM Checksum of bytes 0-62 OK (0x8C) 
# Of bytes written to SDRAM EEPROM 128 
Total number of bytes in EEPROM 256 
Fundamental Memory type DDR SDRAM 
SPD Revision 0.0

---=== Memory Characteristics ===--- 
Maximum module speed 400MHz (PC3200) 
Size 512 MB 
TCL-tRCD-TRP-Reverse 3-3-3-8 
Supported CAS latencies 3, 2.5, 2 
Supported CS latencies 0 
Supported latencies WE 1 
Minimum Cycle Time (CAS 3) 5 ns 
Maximum Access Time (CAS 3) 0.65 ns 
Minimum Cycle Time (CAS 2.5) 6 ns 
Maximum Access Time (CAS 2.5) 0.7 ns 
Minimum Cycle Time (CAS 2) 7.5 ns 
Maximum Access Time (CAS 2) 0.75 ns

---=== Manufacturing Information ===--- 
Manufacturer Kingston 
Manufacturing Location Code 0x04 
Part Number K 
Manufacturing Date 0x001E 
Assembly Serial Number 0x6B376D48

For information, we can see that this is a module DDR-400 (PC3200) from 512 MB of Kingston. Check that the module supports the use of CAS 3, 2.5 or 2, but then it is specified that the minimum time for access using CAS 3 is 5 ns and using CAS 2 is 7.5 ns. That is, the module is only able to use CAS 2 at lower frequencies. When operating at 200 MHz, the nominal frequency, it is automatically using CAS 3. Although the specifications are a bit confusing, they indicate that in fact I have in hand a CL3 module.

The SPD is a small chip EEPROM memory, with only 128 or 256 bytes, which can be easily located in the module:

Thanks to him, the motherboard may automatically use the recommended settings for the module, facilitating the setting. But you can disable the automatic configuration (By SPD) and specify your own configuration through Setup.

Most current boards allow the memory to operate asynchronously with the clock of the motherboard, allowing the motherboard operates to 166 MHz while the memory runs at 200 or 233 MHz, for example. By using an old module, you can also do the opposite, keeping the motherboard and 200 MHz, but setting the memory to operate at 133 MHz, for example:

Many plates beyond, allowing you to manually adjust the value of CAS memory. This can be useful when you overclock, because a DDR-400 module, can not work stable at 233 MHz (for example), while the CAS in 2 days, but can work well if the time is increased to 3 times. The reverse is also possible. A module DDR-400 CAS3 might work stable with CAS 2 if the frequency was reduced to 166 MHz, for example, offering a good flexibility for when you have time and want to get the best possible performance.

Some cards go further, offering a complete range of adjustments, as in the second screenshot:

Playing with the frequency and time of access of the memory is not liable for the equipment. At most you may need to clean the setup for the microwave re-boot after trying to use an unsupported configuration for the modules.

The greatest risk is to increase the voltage used by the modules (Memory Voltage). It is shown that small increases in tension increases the possibility of stable work modules to higher frequencies, especially in the DDR2 modules, which dissipate more heat. The problem is that it can also shorten the life of the modules, so it is not always a good idea.

Increases of up to 5 to 6% are within the limits of tolerance of the circuit and do not offer big risks. You can use a 2.65v or 1.9v DDR module in a DDR2 module, but from then there is injury to the life. Increases from 20% can really burn the modules in a few hours, so the options are not normally available.