More in

UPDATE - Intel unveils 3D XPoint based SSDs

2 mins read

It’s getting on for two years since Intel and Micron launched their 3D XPoint memory concept. Describing the approach as ‘the first new class of mainstream memory technology since 1989’, they claimed 3D XPoint would be ‘up to 1000 times faster than NAND flash’. Since then, the performance claims have been scaled back significantly.

What the companies didn’t explain then – and haven’t really done since – was how 3D XPoint operates. Most observers suggested that 3D XPoint would be an update on phase change memory (PCM), a technology developed by both companies. But Intel says no. Suggesting that 3D XPoint ‘doesn’t use electrons’, it also says that it’s not based on PCM or memristor technology.

Intel planned to launch 3D XPoint based solid state drives (SSDs) under the Optane brand ‘early in 2016’. For whatever reason, Optane SSDs have only just been launched and Micron’s version – QuantX – has yet to be seen.

The Optane SSD DC P4800X Series of SSDs is said to have the potential to transform data centre storage and memory architectures. According to Intel, the device provides a combination of high throughput, low latency, high quality of service and high endurance. In particular, Optane SSDs are said to boost performance at low queue depths, where most applications generate storage workloads.

Intel adds that Optane SSDs will drive new solutions, with applications such as artificial intelligence and machine learning, faster trading and deeper insights into medical scans. The devices are also expected to expand the reach of cloud computing solutions.

Initially, Optane SSDs will be available as a 375Gbyte PCIe format board, with ‘broad availability’ planned for the second half of 2017. Additional capacities and form factors will be available in the second half of 2017, the company noted.

Meanwhile, MRAM pioneer Everspin Technologies says it is making steady progress in its efforts to commercialise the technology and sees enterprise storage as one of its target markets. CEO Phill LoPresti said its 256Mbit device had now been released for production and that its 1Gbit part has completed in house tests and will sample on schedule later this year. “We have three design wins for our parts,” he noted. “Two are in U.2 format storage devices, the other is in an M.2 format.”

He sees MRAM as being in a similar position to NAND flash when it was launched. “The first NAND devices were low density and expensive and everyone was wondering what they could use it for. Today, we can’t live without NAND and I believe MRAM is following the same growth curve.”

He says the memory market is driven by density. “Until you have higher densities in production, you will always be limited in the design opportunities you can address. Now we have larger densities in products, market opportunities should increase.”

Process control system developer JAG Jakob is using 16Mbit parts as a main memory, while a forthcoming 256Mbit part with a quad SPI interface is attracting interest from automotive companies for data logging applications.

Everspin announced at Embedded World that its MRAM technology is now compatible with Xilinx’ UltraScale FPGAs. Using a software script to modify the Xilinx Memory Interface Generator (MIG), developers can work with Everspin’s 256Mbit DDR3 ST-MRAM and the company says a forthcoming update will expand that to include its 1Gbit DDR4 ST-MRAM.

“The compatibility of Xilinx’ MIG with Everspin’s ST-MRAM will provide systems developers the capability to implement a high speed, persistent memory system in their custom solutions,” said Manish Muthal, Xilinx’ vice president of data centre. “We are pleased to be working with the pioneer in MRAM to bring advanced persistent memory capability to modern data centre storage and computing applications."