We all know that more RAM, or memory, is better. But what about the speed of the memory? It might seem like a geeky detail to consider when upgrading your computer or building a new one from scratch, but you’d be surprised at the bump in performance you can get from having faster memory. The new DDR4 standard of memory has improvements in power consumption and data transfer speed.
The DDR2 era is over, although there are plenty of computers that still hasn’t been upgraded to newer standards and still running and performing their tasks well. DDR3 memory has been the de facto standard for majority of x86 PCs out there and its successor – DDR4 SDRAM, although introduced quite some time ago, is still struggling for its place in the market due to different factors. And one of these factors that is worth mentioning, is the rapidly growing market of mobile devices and gadgets utilizing slower but low-powered memory that causes a slowing of interest in traditional desktop computers. But let’s leave alone the market and financial side of the story and concentrate on more technical details of DDR4 memory.
A little history
The Joint Electron Devices Engineering Council (JEDEC) began initial research and development of DDR4 memory standard far back in 2005 and by that time, DDR2 was ruling the market. Even DDR3 memory wasn’t available for consumers yet – serial production was just about to start. In other words, engineers already understood that DDR2 and even upcoming DDR3 capabilities might become quite limited in the future, due to fast development of other computer’s components, such as CPUs and motherboard chipsets.
In 2014, Intel has introduced their new Haswell-E processor and Intel LGA 2011-3 platform with built-in 4-channel memory controller that supports DDR4 modules. The new Skylake processors and chipsets that use the LGA 1151 socket that were introduced in 2015 all support DDR4.
New features in comparison with DDR3 SDRAM
The very first change is operating voltage. It came down to 1.2V, whereas DDR3 operates with 1.5V. Although, there are 1.35V DDR3L and 1.25V DDR3U ultra-low voltage modules currently available on the market, they cost more and have some limitations. Some might say that this tiny difference is not relevant, but they would be wrong – the total operating voltage reduction of 20% is quite an achievement. The majority of home users don’t care about how much energy their memory is consuming, as long as it’s achieving maximum performance, but for corporate environments, server clusters or datacenters the savings in electricity here would be quite substantial – roughly 30kW per 2000 CPUs. Not to mention that such a low voltage rates gives wider opportunities for overclocking in comparison with DDR3 modules.
The second big change is the speed. DDR4 has got JEDEC specifications of 1600, 1866, 2133 and 2400MHz of nominal rated speeds. DDR3 has 1066, 1333, 1600, 1866 and recently introduced 2133MHz nominal speeds. At a first glance, DDR4 has an upper hand here, but the situation is not that simple as it may seem. The thing is that for both DDR3 and DDR4 modules with similar speeds, DDR3 has lower timings and Common Access Strobe (CAS) latency (also known as CL) than DDR4, which give the same performance. For example, a DDR3 PC3-12800 module may have a speed of 1600MHz, maximum timings of 10-10-10 and maximum CL of 12¾, whereas a DDR4 PC4-1600 module with the same 1600MHz speed, may have maximum timings of 12-12-12 and a CL of 15. Both modules have peak transfer rate of 12800MB/s, hence similar performance. Where is the catch? Well, if you plan to upgrade to DDR4 with similar characteristics you won’t have memory performance improvement, but will pay more, since DDR4 modules are more expensive. Not to mention, that additional cost for a new motherboard supporting DDR4 modules might be considerable. Nevertheless, the direct comparison is not completely relevant due to variety of other factors that have to be considered when upgrading or building a new system.
Another new feature introduced in DDR4 architecture is the possibility to increase memory density in the future by using 3D-stacking of dies with TSVs (Through-Silicon Via) provisioning for up to 8 stacked dies, which would directly affect the capacity of the single UDIMM module towards bigger numbers. Simply put the capacity of a single DDR4 module for desktop platforms may vary from 4 up to 32GB. And for server solutions, there are 128GB modules already available on the market.
There are other significant differences between DDR4 and DDR3, but most of them are related with electronic engineer/design role for the memory and motherboard manufacturers, such as signal termination, extra programmable latencies and internal register adjustment.
More than a year has passed since the introduction of DDR4 SDRAM to the market practical applications for home or office users are still a bit unclear. No doubt that the DDR4 and the Intel Skylake LGA 1151 platform are a huge step forward into computer development – it has a lot of interesting wrinkles and future scalability, but all this is not really in a high demand right now within home users. The majority of real-life applications are still utilizing the DDR3 memory performance on a very good level and may even have slightly better performance rates for similar speeds due to lower timings. And even for DDR4 modules with speeds over 2400MHz (overclocked) it is still a big question whether it is worth it – a performance increment of a couple of percent wouldn’t make a significant difference, especially in games.
The real and effective field of application of DDR4 as it seen by many of analysts is overclocking enthusiasts and server solutions.
Hardware and software developers probably need more time not only to roll off this new DDR4 technologies that already presented in the market, but to optimize both hard and software in the best possible way to make it competitive with the vast DDR3 market and attractive to home and office users, not to mention the price.