DDR6 RAM Explained: Release Date, Price, and DDR6 vs DDR5

DDR6 is the next generation of PC and server memory, and it is not on shelves yet. JEDEC, the industry group that writes the actual DDR6 specification, has not published a final version of it as of late September 2026, and the three companies that will build the chips — Micron, SK hynix, and Samsung — have given the public wildly different timelines, from cautious 2027 server pilots to a 2029–2031 mainstream window in SK hynix's own newest roadmap. When it does land, DDR6 is expected to roughly double DDR5's top speed, push operating voltage lower, and move to a new four-channel internal layout built around the CAMM2 module shape. None of that changes what to buy today: DDR5 is mature, widely available, and the only real option for a PC built in 2026.

What is DDR6?

DDR6 (Double Data Rate 6) SDRAM is the sixth generation of the DDR memory standard that has shipped in essentially every PC, laptop, and server since the early 2000s. Like every DDR generation before it, DDR6 is being defined by JEDEC as the successor to DDR5 SDRAM, which itself debuted in July 2020, according to Crucial's own DDR5 explainer — the same kind of generational leap covered in Crucial's broader guide to different types of RAM. Each generation roughly doubles peak data rate while lowering operating voltage and reworking the internal channel architecture to keep signal integrity intact at higher speeds — DDR6 is expected to follow that same pattern, according to the Wikipedia entry for DDR6 SDRAM, which lists it as the direct successor to DDR5 with a projected transfer rate of 8,800–17,600 MT/s and bandwidth up to 134.4 GB/s.

It is important to separate "DDR6 exists as a technical project" from "DDR6 is a product you can buy." Prototype chips have reportedly been fabricated by all three major DRAM makers, and platform-level testing with CPU vendors is underway, but no JEDEC-ratified, warranty-backed DDR6 module has shipped to a single consumer as of this writing.

DDR6 vs DDR5: what actually changes

The headline change is speed, but the more interesting shifts are in how DDR6 gets there. Rather than simply cranking the clock on DDR5's existing layout, DDR6 restructures how data moves in and out of each module.

Two of those changes deserve more explanation: the channel split and the module shape.

Speed and bandwidth, explained simply

"MT/s" (megatransfers per second) measures how many data transfers a memory chip can perform per second, not clock frequency directly — it is the number search engines mean when people search "DDR6 speed." DDR5 modules on the market today, per Crucial's own DDR4-vs-DDR5 comparison chart, start at a standard 5,600 MT/s and reach considerably higher in overclocked, XMP- or EXPO-enabled kits. DDR6 is projected to start where today's fastest DDR5 kits top out and scale up to roughly 17,600 MT/s — around double DDR5's baseline. In practical terms, that bandwidth increase matters most for workloads that are memory-bound: AI training and inference, multi-core CPU rendering, and data-center virtualization, which is exactly why servers and AI systems are expected to get DDR6 years before ordinary desktops do.

If you already run a DDR5 system and want to check what speed your own memory is actually running at — many kits ship well below their rated speed until XMP or EXPO is turned on — Pandromeda's guide to checking RAM speed in Windows 11 and enabling XMP/EXPO covers that on today's hardware, separate from anything DDR6-specific.

Voltage and power efficiency

Every DDR generation has lowered operating voltage: DDR4 ran at 1.2V, and DDR5 dropped that to 1.1V, according to Crucial's DDR4-vs-DDR5 spec comparison. JEDEC has not published a final DDR6 voltage figure, so treat any specific number you see elsewhere as an unconfirmed projection rather than a spec. What is consistent across industry roadmap discussion is the direction: DDR6 is expected to push voltage lower still, both to control heat at nearly double DDR5's transfer rate and to suit AI data centers, where memory power draw at scale is a real operating cost.

Architecture: four channels, CAMM2, and signal integrity

DDR5 already split each module into two independent 32-bit sub-channels instead of DDR4's single 64-bit channel — a change that improved how efficiently a CPU could queue up memory requests. DDR6 goes further, splitting each module into four narrower 24-bit channels, per the DDR6 SDRAM entry on Wikipedia. Narrower, more numerous channels are a way to keep signals clean at speeds nearly double DDR5's, and DDR6 is also expected to add Decision Feedback Equalization, a signal-processing technique that cleans up a data signal as it arrives rather than requiring it to stay clean the entire way — both changes exist specifically to make ultra-high transfer rates reliable.

The module shape is changing too. DDR6 is expected to standardize on CAMM2, a compact, socketed module format that already ships in a handful of DDR5 laptops today, rather than the traditional DIMM stick. CAMM2's shorter signal paths are better suited to DDR6's higher speeds than a full-length DIMM, and JEDEC has separately floated optional 3D-stacked DRAM support for the standard, aimed squarely at the same AI and data-center workloads driving the speed increase in the first place.

JEDEC standardization: where things actually stand

This is the part search results tend to get confidently wrong. As of late September 2026, JEDEC has not published a finalized, publicly available DDR6 specification the way it has for every prior DDR generation. Development work has reportedly been underway since 2024, and DRAM makers are building prototype silicon against draft specifications, but "prototype chips exist" and "the standard is ratified and modules are shipping" are two very different milestones — and DDR6 has only cleared the first one. (JEDEC's own site was not reachable for direct verification while researching this piece; the standardization status above reflects manufacturer and industry reporting, not a JEDEC-published announcement.)

Realistically, expect platform-level validation and certification work to continue through 2026 and 2027, first-wave DDR6 products to target servers and AI accelerators rather than PCs, and ordinary desktop and laptop DDR6 kits to arrive only after that server ramp is well underway — likely 2028 at the earliest, and possibly later depending on which manufacturer's roadmap proves accurate.

What Micron, SK hynix, and Samsung have actually said

Of the three major DRAM makers, SK hynix is the one with the clearest public acknowledgment of DDR6 work in progress. In a profile published on the SK hynix Newsroom, senior manager Myung Chan "Mickey" Choi describes part of his job as "working on the pathfinding for future products like DDR6" — direct confirmation that early-stage DDR6 development is underway inside the company, without any commitment to a launch date. Separately, industry reporting on SK hynix's DRAM roadmap presentation at its AI Summit in November 2025 placed DDR6 in a 2029–2031 window, aimed at the server market first — a notably more conservative estimate than the 2027–2028 figures that circulated earlier in DDR6's development.

Micron has not published a DDR6-specific roadmap page on its own site. Its current public DRAM messaging tops out at DDR5-generation technology such as MRDIMM, a bandwidth-boosting DDR5 module format, and its standard DDR5 SDRAM product line — both signs the company's public-facing roadmap has not yet moved on to DDR6 specifics, even as it is reported to be validating prototype silicon alongside Samsung and SK hynix.

Samsung is the most quoted on DDR6 timing in industry press, with unconfirmed reports attributing a roadmap presentation to the company that targets DDR6 speeds beyond 10 Gbps by 2027 — earlier than SK hynix's own newest public figures. Samsung's own semiconductor newsroom was not reachable for direct verification while researching this piece, so treat that 2027 figure as an industry report rather than a confirmed Samsung commitment.

The short version: all three companies are working on DDR6, none has published a firm consumer ship date, and their own public signals disagree with each other by two to four years.

DDR6 price: what to expect

There is no DDR6 price to report yet, because no JEDEC-compliant DDR6 module has shipped to any customer, consumer or enterprise. Anyone quoting a specific DDR6 price per gigabyte right now is guessing. What is knowable is the historical pattern: every new DDR generation has launched at a meaningful premium over the outgoing generation's mature pricing, then converged toward parity as adoption, yield, and competition among Micron, SK hynix, and Samsung increased. DDR5 followed that curve after its 2020 debut, and there is no public signal suggesting DDR6 will behave differently. Expect early DDR6 — likely server and AI-accelerator modules first — to carry a steep premium, with anything resembling mainstream desktop pricing several years further out still.

Which motherboards and CPUs will support DDR6

None do yet, and none will for some time. As with every past DDR transition, DDR6 will require CPUs built with DDR6-native memory controllers and motherboards designed around DDR6's new channel layout and, likely, CAMM2 sockets — exactly as DDR5 required new CPUs and boards that could not accept DDR4 modules. Every current consumer platform, including the newest desktop chips compared in Pandromeda's Core Ultra 7 270K Plus vs Ryzen 7 9800X3D vs 9850X3D comparison, runs DDR5. Given that server and AI silicon is expected to get DDR6 memory controllers first, watch for data-center CPU platforms from the usual vendors to lead, with consumer desktop and laptop chipsets following only once the server ramp is established — likely not before 2028.

Should you wait for DDR6, or buy DDR5 now?

Buy DDR5 now if you need a PC. That is true whether you are building a gaming desktop, buying a laptop, or specifying a workstation: DDR5 is a mature, fully validated, widely stocked standard with years of runway left, and no DDR6-compatible CPU or motherboard exists for consumers to buy at any price. Waiting for DDR6 today is functionally waiting for the next one or two CPU platform generations to arrive, not for a memory kit to go on sale.

The only readers who should actively track DDR6 timing rather than just buying DDR5 are those planning a system build two-plus years out, or anyone specifying enterprise/AI infrastructure where server-first DDR6 rollout is actually relevant on a realistic 2027–2029 procurement horizon. Everyone else gets more value from tuning the DDR5 system they already have than from waiting on a standard that is not close to shipping to consumers. That is free performance most systems are leaving on the table right now — and if you are not sure what else is inside your current build, Pandromeda's guide to checking what graphics card you have is a good companion piece for the same "know your hardware" habit.

What's next

Watch for three concrete signals rather than more speculative roadmap slides: a JEDEC announcement that the DDR6 specification is formally ratified and published, the first Micron, SK hynix, or Samsung press release naming an actual DDR6 server or AI-accelerator product (rather than a prototype or roadmap slide), and the first Intel or AMD platform disclosure naming DDR6 memory-controller support. Any of those would be the first real evidence that DDR6's timeline is firming up rather than sliding, the way it has moved several times already since early development began. Pandromeda will update this explainer as JEDEC, Micron, SK hynix, and Samsung make it official.

Written byMara Lindqvist

Mara Lindqvist edits the hardware desk. She covers graphics cards, processors, memory and storage, the foundries and chip designers behind them, and what the numbers on a spec sheet mean for people choosing a PC. Specifications in her stories come from manufacturer spec pages and datasheets.

More from Mara LindqvistAll Hardware coverage