Intel Details 2nd Gen Optane Memory 'Barlow Pass' For Ice Lake And Cooper Lake
Peering into its crystal ball, Intel
sees a future in which innovative storage products play a massive roll, both on the consumer side and especially within the enterprise and data center sectors. Related to that, Intel today outlined its Optane
roadmap, including a new "Barlow Pass" Optane DC persistent memory for its Ice Lake and Cooper Lake CPUs.
"The world is generating data at an accelerating rate, and businesses are increasingly becoming overwhelmed with how to efficiently process it. Harvesting value from all this data will be critical in separating the winners from losers. It will require cutting-edge innovation in the memory-and-storage hierarchy, which is what we are driving at Intel," said Rob Crooke, Senior VP and GM of Non-Volatile Memory Solutions at Intel.
Intel is essentially highlighting its plans in storage in the post Micron partnership
era. Whereas Intel and Micron had collaborated on 3D XPoint
memory up to this point, that partnership has now ended. Intel's interest in storage has not diminished, however.
Barlow Pass And Beyond
Intel's first generation of Optane DC Persistent Memory was codenamed Apache Pass, and it paired with the company's Cascade Lake-SP CPUs. Looking ahead, the successor to Apache Pass will be a new generation of Optane DC Persistent Memory called Barlow Pass, arriving next year in conjunction with Intel's 14nm Cooper Lake
and 10nm Ice Lake
processors. Beyond Barlow Pass, Intel is planning at least two more generations of Optane DC Persistent Memory.
This type of storage comes in the form of modules that plug into DDR4 slots (Intel may switch to DDR5 beyond Barlow Pass, though no such details have been shared). This gives servers access to gobs of non-volatile data with a close link to its Xeon
In years past, these Optane Memory products would be produced at Micron's IMFT fab in Utah. Following the separation, however, Intel is transferring its Optane technology development line to its facilities in Rancho Rio, New Mexico.
Intel is predicting enormous growth in demand in storage. In order to meet that demand, it plans on introducing 96-layer quad-level cell (QLC) drives later this year, followed by 144-layer QLC drives in 2020. Both will target data centers at the outset, before trickling into other markets.
The upcoming 144-layer QLC will retain the same 1Tb die capacity as its 96-layer and 64-layer QLC products. Intel is not sharing details on drive capacities, model names (just codenames), or pricing yet, so we will have to wait and see how things shake out.
Intel also discussed its 4 bits per cell floating gate cell tech. According to Intel, this gives it an advantage over charge trap flash tech used by the competition (such as Toshiba
), by offering bigger read windows between cells and less leakage. This ultimately results in more accurate data, and also makes it a bit easier to boost density as its Optane Memory scales.
There is only so far that QLC NAND can take Intel's storage products. As such, Intel is forging ahead to five bits per cell, or penta-level cell (PLC). This is part of a march towards increased density. Like Toshiba, however, this tech exists behind the scenes and is not yet ready to be implemented into actual products. Given that Barlow Pass is already on tap for a 2020 launch, we are probably looking at 2021 or later for PLC products to emerge.
Density and capacity are both on an upwards trajectory, and so its performance among Intel's upcoming Optane drives. Shown above is a look at the latency and IOPS of Intel's current Optane-based P4800X drive
, compared to an Intel DC P4610 NAND SSD.
If you refer back to the roadmap earlier in this article, you will see an entry for Alder Stream, a next-gen Optane DC SSD. Performance is off the chart, literally, as plotted by Intel. Whereas the P4800X hits a wall at around 500,000 IOPS, an Alder Stream drive zooms out to 800,000 IOPS and possibly beyond (the chart gets cut off, and Intel did not provide a concrete figure).
This is all exciting stuff for data centers and server applications, and it shows Intel is not slowing down its efforts in storage after splitting with Micron. On the home consumer side, we will have to wait and see what materializes. New tech has a way of trickling down over time, and upcoming Optane innovations are not likely to be an exception.