Last year, Intel and Micron announced that they’d developed a new memory standard. This new memory, 3D XPoint (pronounced “crosspoint”) is a non-volatile memory that Intel is advertising as the first major memory breakthrough in 25 years. Early speculation was that 3D XPoint would be based or at least related-to phase change memory, but Intel has denied that this is the case without specifying the exact details of how Optane actually works. Intel has been claiming that 3D XPoint (marketed as Optane) would deliver up to 1000x the performance of NAND flash, and the company actually demoed the new technology live at Shenzhen IDF this week.
The video shows an Intel Optane drive pounding the heck out of an SSD with both drives connected via Thunderbolt. But there are some serious disparities between the two configurations. The NAND drive is copying from a SATA SSD to an external SSD connected via Thunderbolt, but the sustained copy rate of 283MB/s suggests a serious bottleneck in the system somewhere; SSD-to-SSD copy rates should be higher than that. The Optane-to-Optane copy used a PCI Express-based drive as the internal storage, which means it ought to have been compared against an SSD configured in the same fashion. As PC Perspective points out, Intel’s own top-end products can beat Optane in a head-to-head NAND-vs-whatever-the-heck-Optane-is shootout.
Early days for Optane
The short explanation for this sleight-of-hand is that Intel is still ramping up Optane (mass production is set for the end of 2016) and the hardware’s performance is likely still in early days. Furthermore, there are advantages to a non-volatile memory pool that aren’t just performance related — if Intel can build a storage array that’s far more durable than NAND with better access latencies, than it may not matter if maximum NAND performance is able to keep up with Optane. Most of the innovation happening in NAND these days is aimed at increasing its density rather than pushing performance.
It’s possible that Optane will kick off at a point similar to high-end NAND, but scale more effectively in the long term or by offering yet another storage tier between high-speed, low-density RAM and traditional hard drives.
There’s no word on whether or not Optane will come to consumer products or how useful it would be if it did. The performance gap between SSDs and HDDs is still larger than the gap between even an older SSD and a modern high-end drive. Optane isn’t expected to match DRAM performance, and that’s more or less what it would need to do to give users a performance boost to match what SSDs offered over traditional hard drives.
The video shows an Intel Optane drive pounding the heck out of an SSD with both drives connected via Thunderbolt. But there are some serious disparities between the two configurations. The NAND drive is copying from a SATA SSD to an external SSD connected via Thunderbolt, but the sustained copy rate of 283MB/s suggests a serious bottleneck in the system somewhere; SSD-to-SSD copy rates should be higher than that. The Optane-to-Optane copy used a PCI Express-based drive as the internal storage, which means it ought to have been compared against an SSD configured in the same fashion. As PC Perspective points out, Intel’s own top-end products can beat Optane in a head-to-head NAND-vs-whatever-the-heck-Optane-is shootout.
Early days for Optane
The short explanation for this sleight-of-hand is that Intel is still ramping up Optane (mass production is set for the end of 2016) and the hardware’s performance is likely still in early days. Furthermore, there are advantages to a non-volatile memory pool that aren’t just performance related — if Intel can build a storage array that’s far more durable than NAND with better access latencies, than it may not matter if maximum NAND performance is able to keep up with Optane. Most of the innovation happening in NAND these days is aimed at increasing its density rather than pushing performance.
It’s possible that Optane will kick off at a point similar to high-end NAND, but scale more effectively in the long term or by offering yet another storage tier between high-speed, low-density RAM and traditional hard drives.
There’s no word on whether or not Optane will come to consumer products or how useful it would be if it did. The performance gap between SSDs and HDDs is still larger than the gap between even an older SSD and a modern high-end drive. Optane isn’t expected to match DRAM performance, and that’s more or less what it would need to do to give users a performance boost to match what SSDs offered over traditional hard drives.
