We haven’t heard much from Intel’s storage division lately, but there are rumors that the company will make several announcements in the not-too-distant future.
All of these rumors are from BlocksandFiles.com. Reportedly, Intel is working on a new type of QLC NAND that would see it stretch to 144 layers, up from 96 layers today. Such parts wouldn’t be likely to ship before later 2021, but they’d be a significant capacity improvement over modern hardware. 3D die stacking has been driving the capacity improvements in NAND for the past few years, and multiple manufacturers like Samsung, Intel, WD, and Micron are all moving to 100+ layer designs.
Intel is also still reportedly working on its PLC (penta-layer-cell) technology for NAND, which stores five bits of data per cell, compared with 1-bit for SLC, two bits for MLC, and three for TLC NAND. Each bit of additional storage capacity has always required steep sacrifices in terms of drive performance and long-term program-erase cycles. QLC NAND, for example, is generally recommended for “cool” storage drives and projects with limited write workloads, due to the small number of program/erase cycles available.
PLC NAND would require 32 charge levels to be stored in each cell. Intel, however, seems to think it has a path towards commercializing the technology, and an additional 1.25x increase in data storage per cell would be welcome. Theoretically, the lower write cycles can be compensated for with larger pools of replacement NAND on-drive, and TLC and QLC drives commonly boost their performance by dedicating small amounts of the drive as SLC cache.
New Optane Coming on PCIe 4.0
There’s no support for PCIe 4.0 baked into Intel’s upcoming Comet Lake CPUs, but motherboard manufacturers are loudly signaling that their Comet Lake boards are all equipped for the feature. That means we can expect an Intel chip that does support PCIe 4.0 to show up at some point — and there’ll be Optane storage ready for the standard when it does.
Bringing Optane to PCIe 4.0 shouldn’t change the standing between it and equivalent NAND in any particular. Here’s what I mean: A PCIe 4.0 NAND SSD is faster than a PCIe 3.0 NAND SSD, but it doesn’t automatically offer dramatically better latency than a PCIe 3.0 SSD. A PCIe 4.0 NAND SSD supports the same type of read/write operations as a PCIe 3.0 drive, and it has the same strengths and weaknesses. Intel hasn’t given many details on what to expect from next-generation Optane, but we’re assuming that the storage medium will continue to offer its historical strengths (very strong performance at low queue depths) and that Intel will continue to position it as a DRAM alternative in certain server installations.
I’m very curious to see what Intel has in mind as far as PLC storage, and I’d like to see second-generation Optane create a more meaningful gap between itself and NAND. Optane, I think, has to be graded on a bit of a curve. It’s very difficult for any company to ramp a brand-new storage product up to compete with already-established solutions, which is why it took NAND so long to emerge as a viable competitor to hard drives. To-date, Optane has demonstrated some superior capabilities to NAND in specific use-cases, but neither NAND nor Optane has upset DRAM’s position in the memory hierarchy. Intel wants to push Optane into RAM sockets going forward, so the more DRAM-like its overall performance, the better.
- Intel Had Record-Breaking 2019, but Optane Refresh Could Slip to 2021
- Scientists May Have Discovered Universal Memory, DRAM Replacement
- Intel Releases Specs for Its Optane+QLC NAND H10 Memory