How To Choose RAM (Random Access Memory) is important computer hardware that will directly affect your user experience.
There are two types of RAM: DRAM and SRAM, the former being the most commonly used one. Although technically DRAM can be implemented as SRAM, it sacrifices speed for lower cost due to the higher complexity of circuits and therefore generally you only buy DRAM nowadays. The main differences between DRAM and SRAM include:
DRAM needs refreshing periodically while SRAM doesn’t need refreshing at all; this results in high access latency for DRAM but almost zero latency for SRAM, which accounts for up to 99% of CPU’s waiting time for RAM access.
DRAM uses quite a lot of power, usually at least 100 times more than SRAM which reduces battery life if using a laptop.
SRAM is significantly faster than DRAM thanks to its simpler structure and lower latency. However, the complexity of circuits increases with size up to a certain point which makes it impractical to implement large capacity SRAM due to higher cost per bit. Also as mentioned above, although theoretically possible, currently there are no single-chip SRAMs larger than a few KB in capacity normally used in desktop/notebook computers due to high cost per bit meaning that even with a high level of integration you are unlikely ever going to see 32 MB or 64 MB single-chip SRAM for desktop PC. Therefore most DRAMs available on the market are very large capacity, e.g. 2 or 4 GB per stick of 128 bit wide DRAM for desktop/notebook PCs. This makes them much cheaper compared to SRAM because in addition to a higher level of integration you don’t need any decoding circuitry for each individual bit which would be required in case if you wanted to make single-chip SRAM with the same width as a wide DDR3 module.
How To Choose RAM?
RAM stands for Random Access Memory. This type of memory is volatile, meaning that when the power source is cut or lost, the data in the best Gaming RAM will be lost/discarded as well. That’s why it’s called ‘random access’ and not ‘persistent storage’. It works using a series of transistors and capacitors, which hold either a 1 or 0 (on or off). When you want to read from this memory, it takes microseconds to find out what was stored where and retrieve that information. However, when you need to write something into memory, it can take weeks before all the cells are refreshed correctly after each bit flip due to interference from passing cosmic rays (which happen more than any other form of radiation on Earth).
RAM is used as the main memory for any programs/processes running on your computer. If your applications don’t have enough RAM available, they’ll begin to make use of something called paging, which will start writing data onto the virtual memory (a section of space on your hard disk) and then reading it back when needed. This results in slower performance. A lot slower. For example, if you’re using a pagefile with 1024MB worth of memory, loading a large image may take half a second instead of instantaneous since it’s written into that file fullscreen before being read back again each time it is opened or displayed on the screen. We humans usually notice changes in delay below 100ms – anything shorter than this is effectively instantaneous, which is precisely why you won’t notice pagefiles in action.
Why are they needed?
RAM is the only universal storage medium that can be accessed at random – hence its name. That’s also why it has to exist in ‘volatile memory’ form. One of the most common misconceptions about RAM is that having more will always speed up your computer. This is not true because each type of data takes a different amount of time to access, so there isn’t an I/O multiplier or something similar to How To Choose RAM. The fastest operation would involve reading from RAM over and over again without any paging involved since the latency of accessing information from RAM is lower compared to accessing information from HDD (hard disk drive) and flash storage (SSD). While reading from RAM will always be faster than reading from HDD, it’s not 4x, 8x, or even 16x as fast. It just takes a certain amount of time to read anything regardless of where it is situated on your computer.
Laptop and Desktop RAM
For example, if you’re using an app that needs to display multiple images at once or in sequence, it needs a certain amount of time’ to do so. That is why many printing companies will use a sequence of already printed photographs on their products – it reduces manufacturing costs by minimizing the number of prints they make. You would also see an improvement when opening more than one window of the program, especially if they are heavy – like Adobe Photoshop or Google Chrome with 50+ tabs open.
There are no hard and fast rules for RAM allocation, but generally, 2GB is enough for light browsing while 8-12GB of DDR3 can handle almost any day-to-day task even when dealing with multiple windows/programs at once. As far as 4K video editing goes, you would want at least 16GB of DDR4 to ensure your system doesn’t use pagefiles during those long stretches of rendering out those split-second vocal effects. Hopefully, by now you can see that more isn’t always better – it’s down to what the user actually wants and needs to do with their computer since not everyone does the same things with their computers.
Laptop vs Desktop?
There is a debate going on about using LPDDR3/DDR3L for laptops, versus DDR4 for desktops. At first glance, there seems to be a difference in latency but LPDDR3 only takes 1 clock cycle longer than DDR4, so this argument is moot from the get-go. There was an argument about LPDDR3 being better or worse for laptops, but considering its low voltage requirements compared to DDR4/DDR3, it’s no surprise that laptop manufacturers are opting for this over desktop RAM. It also runs at a lower voltage so it doesn’t contribute to heat build-up inside the chassis – perfect for any ultraportable device!
Choosing the right type of DDR
The choice of memory modules depends on your motherboard. What you really need to know is how many slots there are and what types of chipsets are installed in each chipset slot, including their latency rates which are expressed as CL timing figures (eg 17-17-17). It’s also good to buy memory that falls within the operating range of your chipset. For example, Intel boards will recommend 1.35-1.5V SDRAM while AMD recommends only 1.5V on their AM4 platform (socket FM2+ and later). If you run on the lower end of this spectrum, there is a chance you’ll be able to squeeze in some extra hours of battery life which can make all the difference when it comes to getting work done while traveling or commuting with your laptop.
Every RAM chipset manufacturer advertises four figures for every type of module they produce: capacity, speed, CL latency, and voltage – but not all manufacturers list their figures the same way since the nomenclature used can be very confusing for newbies. Two of the most common figures you’ll see are 16GB per stick and CL17 timing, but which one is better – the former or latter? The answer lies with your motherboard’s memory compatibility list – you need to find out what type of RAM it supports and whether is it single-sided (one chip on each side), double-sided (two chips on each side), or quad-channel modules.
Motherboard Form Factor and Slot Count
The above question can also be answered by checking your motherboard’s form factor and how many slots it has. If you bought a Mini ITX board it is likely to be single-channel while a Micro ATX will support dual channels if it has two or four memory slots. More than that and you will be looking at quad-channel boards which are available in E-ATX, ATX, and Extended ATX form factors.
CPU and OS Limits
Then of course there’s the CPU itself – some CPUs don’t support memory modules running at full speed, so they may only run DDR4/DDR3L/LPDDR3 at 2133Mhz instead of their maximum speeds (eg 2666MHz). For this reason, it pays to check your CPU’s specification sheet before buying anything since faster speeds on slower rated RAM won’t help boost performance by much if at all. As far as Windows 10 goes, it is not limited to dual-channel memory modes on any platform so you don’t have to buy 4x8GB kits of memory either.
Bandwidth – Single Channel vs Dual-Channel vs Quad-Channel
The number of channels simply refers to the amount of data that can be read or written to your modules. Single-channel memory is very basic and runs at half bandwidth, while dual-channel doubles this figure but requires two RAM modules for it to work – one on each side of the board. Quad-channel takes things even further by doubling up again with four modules installed. The difference between single vs dual vs quad channel modes comes into play when you are editing video footage in high-resolution formats like 4K where memory speed has a direct effect on how fast you can encode these files on your system. As mentioned above, our Intel Core i5 CPU does not handle higher clock speeds on slower rated RAM so the quad-channel mode is out of the question for this build. In fact, as per these official AMD figures, you would be better off with two memory sticks in dual-channel mode than four sticks in quad-channel mode since the latter has significantly lower bandwidth – even on a Ryzen 1800X CPU.
Timings and Latency
Another crucial factor to take into consideration is RAM timings which consist of five sub-categories: TCL (CAS), tRCD (RAS to CAS Delay), tRP (RAS Precharge), tRAS (Active to Precharge Delay), and tRFC (Row Refresh Cycle Time). The names are quite self-explanatory; CL refers to column latency which measures how long it takes data signals to travel between the memory controller and the individual DIMM modules while RAS refers to row latency which is another word for memory speed. The lower the timings, the faster your system will be but this comes at a cost of increased power usage. You’ll notice some DDR4/3L RAM kits advertise speeds higher than their standardized ones without any additional CL or timings – that’s because they are overclocked already, so check with your motherboard manufacturer to see if you can do this too before buying something like this kit which has no specs listed.
As mentioned above, 8GB of RAM in dual-channel is better than 16GB of RAM in single-channel mode since bandwidth per module is doubled instead of doubled again. This is why we opted for a 4x8GB kit of Corsair Vengeance Pro DDR4 RAM rated at 2800MHz and 16-18-18 timings (an additional +400Mhz and +12 timing increase can be applied to this memory but this will push the voltage up as well).
As you can see, there are several different aspects to take into consideration when How To Choose RAM memory such as form factor, slot count, maximum capacity (in GB), bandwidth per channel, speed (in MHz), and latency (in CL value) – all of which depend on your motherboard’s chipset and CPU architecture. Armed with the above knowledge we hope you now have a better idea regarding what to look for when choosing DDR4 memory which you will need to run at a minimum of 2133MHz, however, there are some users who can get away without using high-speed RAM in their systems. Those with older CPUs or laptops will have the least amount of problems since even slower modules will still help improve performance by a noticeable margin. As always, take your time when making a purchase and visit this post frequently as we continually add more useful information about RAM so you won’t have to learn everything from scratch.
Charles Grice is a civil engineer, urban planner, and blogger. He is keen on exploring and writing about the technical hacks for iOS, Android as well as Windows.