Memory capacity
Memory modules
The number of individual modules included in RAM kit. One module occupies one slot on the motherboard, so to install the entire kit, the number of free slots must be equal to or greater than the number of modules.
If there is an even number of modules in the set, they can be paired. This mode significantly increases the speed, but is not supported by all models of motherboards, so in each case this point should be clarified separately.
Now on the market? there are modules supplied in such quantity:
one modules,
set of 2 modules,
set of 4 modules,
set of 8 modules.
Memory rank
The number of ranks provided in the memory bar.
The rank in this case is called one logical module — a chipset with a total capacity of 64 bits. If there is more than one rank, this means that several logical ones are implemented on one physical module, and they use the data transmission channel alternately. A similar design is used in order to achieve large amounts of RAM with a limited number of slots for individual brackets. At the same time, it should be said that for consumer computers, you can not pay much attention to the memory rank — more precisely, peer-to-peer modules are quite enough for them. But for servers and powerful workstations, two-, four- and even eight-rank solutions are produced.
Note that other things being equal, a larger number of ranks allows achieving larger volumes, however, it requires more computing power and increases the load on the system.
CAS latency
This term refers to the time (more precisely, the number of memory cycles) that passes from the processor's request to read data to granting access to the first of the cells containing the selected data. CAS latency is one of the timings (for more details, see the "Memory Timings Scheme" section, where this parameter is designated as CL) — which means that it affects performance: the lower the CAS, the faster this memory module works. However this is true only for the same clock frequency (for more details, see ibid.).
Now there are memory modules on the market with the following CAS latency values:
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
24,
28,
30,
32,
36,
38,
40,
42,
46.
Memory timing
Timing is a term that refers to the time it takes to complete an operation. To understand the timing scheme, you need to know that structurally RAM consists of banks (from 2 to 8 per module), each of which, in turn, has rows and columns, like a table; when accessing memory, the bank is selected first, then the row, then the column. The timing scheme shows the time during which the four main operations are performed when working with RAM, and is usually written in four digits in the format CL-Trcd-Trp-Tras, where
CL is the minimum delay between receiving a command to read data and the start of their transfer;
Trcd — the minimum time between the selection of a row and the selection of a column in it;
Trp is the minimum time to close a row, that is, the delay between the signal and the actual closing. Only one bank line can be opened at a time; Before opening the next line, you must close the previous one.
Tras — the minimum time the row is active, in other words, the shortest time after which the row can be commanded to close after it has been opened.
Time in the timing scheme is measured in cycles, so the actual memory performance depends not only on the timing scheme, but also on the clock frequency. For example, 1600 MHz 8-8-8-24 memory will run at the same speed as 800 MHz 4-4-4-12 memory—in either case timings, if expressed in nanoseconds, will be 5-5-5-15.
More features
— A series
for overclocking (overclocking). Belonging to such a series means that the manufacturer initially provided in the module the possibility of overclocking ("overclocking") — that is, increasing performance by changing the operating parameters, in particular, increasing the operating voltage and clock frequency. You can also “overclock” ordinary memory that is not related to overclocking — however, this is difficult and fraught with failures, up to complete burnout of the circuits, while in specialized series overclocking is a documented function, it is implemented quickly and simply, moreover, it is most often covered by a guarantee.
—
XMP support. Memory module compatibility with XMP technology. This technology, created by Intel, is used for overclocking (see the relevant paragraph). Its key principle is that certain overclocking profiles are recorded in the memory module — sets of settings tested for stability; and instead of manually setting individual parameters, the user just needs to select one of the profiles. This simplifies system setup and at the same time improves its reliability during overclocking. However, note that in order to use XMP, it must be supported not only by memory, but also by the motherboard.
— AMP support. Memory module compatibility with AMP technology. In terms of its main features, this technology is completely similar to the
XMP described above and differs only in the creator — in this case, it is AMD.
–
EXPO support. Memory module compatibility with EXPO technology (Extended Profiles for Overclocking). It was created at AMD by a specialist for overclocking DDR5 strips as part of Ryzen 7000 systems. At its core, this is a factory set of RAM profiles that simplifies overclocking the “RAM”. Using the technology allows you to increase performance in games by about 11% with a resolution of the broadcast image Full HD.
—
Buffering support (Registered). The presence of the so-called memory module. buffer — a section for quickly saving incoming data — between the memory controller (control device) and the actual chips (storage devices). This scheme reduces the load on the controller, thereby achieving higher reliability; on the other hand, buffered modules have slightly reduced performance due to the delay in transferring information through the buffer. Buffered memory is used mainly in server systems and is expensive. When choosing memory, note that either only buffered or only
unbuffered memory can be used in one system; it is impossible to combine these two types of memory.
—
ECC support. ECC (Error Checking and Correction) is a technology that allows you to correct minor errors that occur while working with data. To use ECC, it must be supported not only by the memory module, but also by the motherboard; Basically, such support is used in servers, but it is also found in "motherboards" for ordinary desktops.
