Micron Said to Beat Competitor’s QLC SSDs with Innovation

By Steve Hanna, leads product management, value-focused data center SSDs, Micron Technology, Inc.

For years, there’s been a pervasive problem with quad-level cell (QLC) NVMe data center SSDs – the specs have been too low to warrant broad adoption. Users were looking for budget SSDs, but the QLC technology sacrificed performance, endurance and power efficiencies, which didn’t make the SSDs a good deal. Simply put, they weren’t the greatest value. Micron solves this problem in data center SSDs.

It announced shipment of the world’s first 200+ layer SSD ⁽¹⁾ – the company’s 6500 ION SSD – that solves the many drawbacks associated with competitor’s QLC SSDs by using the firm’s history of NAND layer leadership to finally fuse the value of QLC with the performance of TLC technology. It’s another firm’s innovation that comes barely a year after launching the world’s first 176-layer data center SSD, the world’s first 176-layer QLC SSD, and 5 years after launching the world’s very first QLC SSD – and a data center one at that. As the product manager for that first QLC SSD, I have a unique perspective on the intriguing story behind the 6500 ION SSD’s origins. It’s a story that goes all the way back to May 21, 2018, when we launched the QLC-class category with the 1st ION drive, the Micron 5210.

When we first launched QLC, the world was excited about the technological leap and the opportunity to get higher capacity, lower cost SSDs. But customers were rightly concerned about the hidden costs of adopting QLC data center SSDs. While QLC allowed them to save on the upfront drive purchase price, some customers ended up paying higher operating costs because QLC consumes more power, due to the complexity of its cell programming. As a result, competitor’s QLC-based SSDs today deliver 5-10x lower endurance, 10-20x lower random write performance, and at least 20% lower specs in all areas outside of sequential reads than our TLC-based equivalent. As QLC matures, and as future SSDs and storage systems are designed to optimize the higher capacities that QLC uniquely enables, then the drawbacks of today’s QLC in the data center will become less relevant. However, most of today’s data centers aren’t designed to overcome the current challenges of QLC-based SSDs, so customers must ask themselves a few key questions: ‘what do I pay for this – and what do I get in return? And is what I get sufficient to meet my needs?’ When that answer is yes, then they buy it.

Across the 4-year lifespan of our original QLC product, we found that many customers were willing to make tradeoffs, particularly when they were replacing legacy 10,000rpm HDDs. Our April 2020 press release proved this by noting that a majority of the world’s OEMs had qualified our QLC products for this purpose. While we saw strong adoption, we wondered if it was possible to help customers save at an even greater scale – what if we could find a way to deliver the upfront QLC cost savings, and lower the annual power and cooling costs that come with QLC in the data center?

That ‘how might we?’ question is what led to the development of the 6500 ION SSD. It offers the cost savings and capacity scaling that customers love about QLC. And it comes with a 20% reduction in power consumption while providing a higher performance and endurance profile that address customer longevity concerns about QLC. With the ability to sequentially write 30TB of data/day (1 SDWPD, or sequential drive write/day) and to randomly write 9TB of data/day (0.3 RDWPD, or random drive writes/day), the classic QLC endurance objection is now off the table with the 6500 ION. It delivers more endurance at 0.3 RDWPD than a 1 DWPD 7.68TB standard SSD. ⁽²⁾

This is a huge paradigm shift – exactly what we set out to do. Since launching the first QLC SSD, our mission has been to raise the bar and create the best value in data center SSDs the world has ever seen – regardless of whether the NAND inside is TLC or QLC. Rather than fall in love with a type of technology, we fell in love with solving customer problems, like lowering total storage costs. By scaling NAND layer counts faster, we’re ultimately able to deliver a sustainable cost structure compared to the competitor’s QLC SSDs – without the compromises that customers experienced in the past. The interest we’re seeing from many of the world’s largest data center customers speaks for itself.

How users benefit from 6500 ION SSD relative to competing 30.72TB QLC SSDs:

34% better average read latency ⁽³⁾
58% faster sequential writes ⁽⁴⁾
Up to 62% more 4KB random read IO/s ⁽⁵⁾
Over 30 times more 4KB random write IO/s ⁽⁶⁾
10x more 4KB random write endurance for improved longevity and flexibility ⁽⁷⁾
Future-proofed feature set: OCP 2.0, NVMe 2.0, NVMe-MI 1.2b, SRIS, and more
Leading security feature set (FIPS, SPDM 1.2, SHA-512, and more)
Get all the above using 20% less active power (20W vs. 25W)
Secure your supply chain: multiple manufacturing sites and Trade Agreements Act (TAA) compliant SKUs
Saturate your network with ease: ⁽⁸⁾
- - 2 drives saturate 50GbE
  - 3 drives saturate 100GbE
  - 7 drives saturate 200GbE
  - 13 drives saturate 400GbE

6500 ION goal: Provide all benefits of TLC, with none of compromises associated with QLC NVMe data center SSDs.
When you’re deploying workloads with constantly scaling capacity needs, such as object stores, general purpose cloud storage, all-flash arrays, SDS capacity tiers, NoSQL databases, content delivery networks and AI/ML data lakes, give the Micron 6500 ION a try.

Additionally, see the recommended resources below for more information on the 6500 ION SSD and its companion drive, the company’s XTR NVMe SSD, which delivers extreme endurance for workloads or all-flash configurations that require even more endurance.

Click to enlarge

⁽¹⁾ Based on similar use SSDs with NVMe available on the open market as of the date of this document’s publication.
⁽²⁾ 1 DWPD for a 7.68TB SSD user capacity means you write 7,680GB/day whereas a 0.3 DWPD for a 30.72TB SSD user capacity means you can write 9,216GB/day.
⁽³⁾ Average read latency for 4KB 100% random read at queue depth (QD) of 1.
⁽⁴⁾ 58% improvement measured at QD32 and QD64. Measurements made at QDs 1 to 256 yielded a range of improvements from -7% to 58% when compared to the competitor’s drive.
⁽⁵⁾ 62% improvement measured at QD32. Other measurements were made at QDs ranging from 1 to 256 and yielded a range of 18% to 62% improvements over the competitor’s drive.
⁽⁶⁾ 30x more 4KB random write IOs at QD128 and over 10x more 4KB random write IO/s at QD1.
⁽⁷⁾ Solidigm documentation states that its endurance is rated using a 64KB 100% random write workload and is 0.41 DWPD. Rated Solidigm D5-P5316 endurance at 64KB transfer size is estimated to be 1/16 of the rated endurance for a 4KB transfer size, yielding a 4KB value of 0.0256 (0.41/16). 6500 ION rates endurance using 4KB random write workloads and is 0.3 DWPD.
⁽⁸⁾ Saturation metrics shown for sequential reads, sequential writes, and random reads: the most common customer saturation goals. Values shown not based on random write performance saturation. 100% network efficiency assumed vs. the typical ~80% network efficiency seen in real-world deployments. You may need fewer drives than shown as values are only intended for conceptual comparison.