SK Hynix 16 Gb DDR5 RAM modules will enter production this year, 4,800 - 8,400 MHz expected frequency range
We could say that smartphones already have it better than computers, as Samsung has been producing LPDDR5 memory chips since mid-2019. Of course, the DDR5 standard will clearly be faster than LPDDR5, but computer motherboard integrators and CPU manufacturers are taking their sweet time with proper implementations, waiting for a sign from RAM producers like SK Hynix. Well, the green light is finally on and the first SK Hynix 16 Gb DDR5 modules built on 10 nm nodes are scheduled to enter mass-production this year. The South Korean company just released a more detailed outlook explaining how it intends to proceed in the coming years, plus some of the immediate advantages brought with the jump to DDR5.
When SK hynix first announced the development of DDR5 chips back in 2018, the top frequencies were expected to hit 5,200 MHz. However, these numbers have meanwhile been bumped to 8,400 MHz. This is probably the theoretical maximum, and we shouldn’t expect memory kits to hit those frequencies this year. SK Hynix is aiming for a 50% increase in frequency at launch time, meaning that the first kits should operate at 4,800 MHz (DDR4 JEDEC maximum is 3,200 MHz). Sure, we already have DDR4 kits that can be pushed to 5,000 MHz, so it will not be long before 6,400 MHz DDR5 variants hit the market.
Memory density will be improved, as well, amd we should see 24 GB to 32 GB per single DIMM more often, with the possibility of reaching 64 GB per single DIMM as the technology matures. Other DDR5 improvements over the current DDR4 standard include doubled bank sizes (32 vs 16), doubled burst length (16 vs 8), Same Bank Refresh function for improved access times, lowered operating voltage from 1.2 V to 1.1 V, plus on-die error correction circuitry for reduced server implementation costs.
According to IDC, DDR5 adoption should rise to 22% by 2021, then 43% by 2022. The first DDR5 modules launching this year will most likely be aimed at server-grade solutions like the AMD’s gen 3 EPYC Genoa and Intel’s Sapphire Rapids Xeon platforms, but the first consumer-grade implementations may not be too far off, probably scheduled for late 2021.