DDR4 arrives in the mainstream
One of the major steps forward for Intel Z170 is the support for DDR4 memory, which marks the first time that DDR4 support will be available on a chipset other than the ultra-enthusiast Intel X99. In fact, Skylake motherboards will dispense with support for DDR3 entirely, and will support either DDR4 or DDR3L RAM. But because the slots are different for each memory type, motherboards will only be able to support one of them, and the overwhelming majority of boards are going to opt for DDR4, which offers several advantages over older DDR3 modules.
For starters, DDR4 has a lower operating voltage of 1.2V, down from 1.5V on DDR3. It also boasts higher data transfer rates – Skylake’s memory controller supports DDR4 2,133MHz by default, as compared to DDR3 1,600MHz on Haswell CPUs and Intel Z97. Furthermore, DDR4 enables higher memory densities, which means up to 16GB modules. Intel Z170 supports up to 64GB across four DIMMs, so you can really load up those RAM slots if you’re of the persuasion that there is no such thing as too much RAM.
To cap things off, manufacturers are also releasing kits at frequencies that DDR3 would be hard-pressed to match. For instance, the new G. Skill Trident Z memory is available in kits clocked up to 4,000MHz, nearly double that of the 2,133MHz base specification.
These advantages aside, DDR4 does have a slightly higher CAS latency than DDR3 because of the higher memory frequencies. CAS latency is a measure of the time – or more precisely the number of clock cycles – taken by the memory to send data back to the processor when it receives a command. On the other hand, memory frequencies indicate how quickly the system can access data on the RAM. So while it might seem like higher CAS latencies would mean that DDR4 is slower, the outcome isn’t always so straightforward because the two work together to influence memory performance.
Memory performance is determined more accurately by the actual latency time, which is measured in nanoseconds (ns). Faster memory clock speeds result in shorter clock cycles between when the memory receives a command and when it actually responds, so even though the CAS latency goes up – more cycles – the actual latency time still decreases, resulting in overall faster memory.