Micron Pulls Ahead on DRAM
DRAM nodes are going back to the beginning of the alphabet.
Micron Technology has unveiled its 1-alpha node DRAM, which the company said offers a 40% improvement in memory density over its 1z node DRAM, as well as a 15% improvement in power-savings for mobile devices. This latest memory node supports densities from 8Gb to 16Gb, and Micron has started volume production of DDR4 memory for compute customers and Crucial consumer PC DRAM products on the new process node, while LPDDR4 is being sampled to mobile customers for qualification.
Aside from mobile devices, the company sees the density, reliability, and power efficiency its 1-alpha DRAM appealing to customers who value longevity, including embedded automotive solutions, industrial PCs, and edge servers as they typically have longer lifespans, said Thy Tran, Micron’s vice president of DRAM Process Integration in a briefing with EE Times. The 1-alpha moniker reflects perceived scaling limitations of the 10-nanometer class process node that ended up being dubbed “X,” she said. “The scaling path was unclear.”
That node was a pivotal one in that Micron was behind in its DRAM development compared to competitors and constantly playing catchup, said Tran. The company was making changes to its cell layout and digesting the Elpida technology to align it with native Micron technology to take the best of each organization. “We didn’t want to rehaul all the tool sets,” she said, and the cell layout change was seen as a way to allow Micron to scale further. “We were always trying to close the gap instead of trying to get technology leadership until we overcame those hurdles.”
These growing pains have been worth it, said Tran, as the technology and design process is more holistic. While there was density targets and cost reductions in mind for 1-alpha, Micron didn’t just focus on the memory cell array but also other areas including the circuit area—everything needed to keep up with the advances made on the memory density front. That included optimizing the process steps to meet the electrical requirements as well while bearing in mind the need to be able to manufacture in high volume.
In addition, she said, the impact of the process capability and tooling improvement on earlier products still in production had to be considered, as well as looking forward. “When we transition to either brand new, leading edge tool, or processes or new module integration, we scope out how scalable that is. What we want to avoid is a one hit wonder,” said Tran. The 1-alpha node, for all its achievements in features and manufacturing advances, is also a steppingstone to 1-beta and beyond.
Ultimately, Micron must find a balance between being evolutionary and disruptive. “That’s how we determine our strategy,” said Tran, and thinking holistically must include not only design and product engineering and technology development, but also, manufacturing costs, as being too aggressive means not getting the biggest bang for the buck from existing tools.
That thinking is also why customers will see 1-alpha DRAM in DDR4 and LPDDR4 first, rather than the latest and greatest JEDEC DDR5 specification, and not just because that’s what customers are predominantly buying right now, she said, but because it’s mature and proven. “We can leverage previous DDR4 designs, for example.” And if you’re going to introduce a new node in manufacturing, from a cost perspective it’s better to have an existing demand to support the ramp up, said Tran.
Micron being first out of the gate with a new DRAM node is just as much about executive management’s focus on good execution as it is about technology advancements since CEO Sanjay Mehrotra took the reigns, said Jim Handy, principal analyst with Objective Analysis. While Samsung and SK Hynix focus more on market share as a metric for success, Micron has focused on profitability and has put itself in a position where it’s able to re-invest in the business—they can invest more in research and development to stay ahead, he said. “The most important thing now is that if they’re getting 40% more gigabytes per wafer that basically sets them up to be that much more profitable than the next guy making DRAM.”
Micron’s business discipline has also made it a leader on the GDDR and NAND flash front, noted Handy, which is impressive, adding that GDDR is a higher profit business and can become more profitable by pushing costs down by moving to a more advanced process node. Similarly, they can get a higher price in the enterprise SSD market than the consumer PC market with lower costs and get better margins.
Just after Micron made its 1-alpha node announcement, both SK Hynix and Samsung announced earnings, with SK Hynix stating it “the memory market showed sluggish trend,” but for the fiscal year ending 2020 it had mass-produced its main products such as 1Znm DRAM and 128-layer NAND Flash. “’Stably mass-produced’ sounds to me like the product finally became production-worthy,” said Handy.
Samsung also reported its fourth quarter 2020 earnings, and seemed to respond to Micron’s recent 1Znm, and that it had produced the Industry’s first 1Anm DRAM with multiple EUV layers, he said, although he’s not sure why a customer would care how much EUV was used to make the part, as long as it does the job and is competitively priced.
Gary Hilson is a general contributing editor with a focus on memory and flash technologies for EE Times.
Micron Leapfrogs to 176-Layer 3D NAND Flash Memory
DDR5 Brings Design Ops, Challenges
Emerging Applications Could Transform GDDR Market
Micron Closes Elpida Acquisition
IMAGE 1 CAPTION Micron Technology has started volume production of DDR4 memory for compute customers and Crucial consumer PC DRAM products on the new 1-alpha process node, while LPDDR4 is being sampled to mobile customers for qualification. This latest memory node supports densities from 8Gb to 16Gb (Courtesy Micron)