Intel reveals it's first Optane SSD that has high memory capacity

Intel has launched its Intel Optane SSD series of memory for servers, created to be capable of driving demanding artificial intelligence and machine learning platforms.

The new Intel Optane SSD DC P4800X Series, which optionally comes bundled with Intel Memory Drive Technology, provides 375GB of fast storage and caching tier functionality.

While the focus has been on 3D XPoint memory, Intel insisted that Optane would be a "bundle" - alongside 3D XPoint, there is a memory controller, a PCIe interface, plus software and some other hardware elements. This means that the data can be written and read in small sizes, leading to faster and more efficient read/write processes. It expands the reach of cloud computing solutions.

At last, Intel is making a stand-alone drive based on its extremely fast 3D storage technology. although you're probably not about to pick one up yourself.

Intel claims the raw 3D XPoint technology is 1,000 times faster than the NAND flash commonly used in storage drives, though still about 1% the speed of temporary memory called DRAM, or dynamic random-access memory.

Intel's first Optane product, the P4800X, is designed for data centre use and can add 375GB of high-speed storage or pseudo-DRAM to a system via the PCI Express bus. In the case of NAND, its low price is offset by latency, endurance and density issues.

Intel argues that current storage approaches based on DRAM and NAND are contributing to the current datacenter storage gap, and that storage platform increasingly need to behave "like system memory". Micron, which will market 3D XPoint under the brand name QuantX, plans to sell products based on technology this year. Whereas server DRAM modules now top out at 128GB, the first P4800X drive has 375GB of capacity.

In all, Intel's $1520 375GB drive offers a combination of high throughput, low latency and high QoS, while being built to endure and alleviate the bottlenecks data centres now face.

Intel also offered figures in support of its claims, with latency times quoted as less than 60 or 100 microseconds for 99.999 percent of the operations with a queue depth of 1.

  • Tracy Ferguson