3D XPoint storage technology is a new type of memory jointly developed by Intel and Micron Technology Inc. And this exciting technology could fill a gap in the market between RAM (DRAM) and NAND flash storage. Now, both one company and the other work separately to develop and sell their products, with interesting consequences…
Index of contents
- How 3D XPoint memory works
- implementations
- 3D XPoint: advantages and disadvantages
- Cost
- XPoint 3D Use Cases
You may also be interested in:
- NAND memory types
How 3D XPoint memory works
In 2015, the companies Intel and Micron announced that they had created a memory technology called 3D XPoint that could be up to 1,000 times faster and have up to 1,000 times more endurance than NAND flash memory. In addition, you could get storage densities up to 10 times higher than current memories. However, the first products launched were not up to expectations.
It is true that they outperformed conventional memories, but not as much as the companies that developed 3D XPoint announced.
It must be said that 3D XPoint has a different architecture than other flash products. This new memory is based on phase shift technology, with a transistorless crossover point that places selectors and memory cells at the intersection of the perpendicular interconnects as can be seen in the image above.
Those cells, made of an unspecified material, can be individually accessed by a current sent through the upper and lower interconnects that touch each cell. To improve storage density, 3D XPoint cells can be stacked in three dimensions.
Each cell stores a single bit , that is, it can have the state 0 or 1, represented through a change in the property of the cell material that would modify its resistance. The cell can occupy a state of high or low resistance, thus representing the binary information. Because the cells are persistent, they keep their values indefinitely, even when there is a loss of power supply, that is, they would be non-volatile.
Access operations, that is, read and write, occur by varying the amount of voltage sent to each selector. For write operations, a specific voltage is sent across interconnects of a cell and a selector. This activates the selector and allows voltage to pass to the cell to initiate the general property change. For read operations, a different voltage is sent to determine if the cell is in a high or low resistance state. This is how these memories work.
On the other hand, it must also be said that 3D XPoint has the ability to write data bit by bit, an advantage over NAND . In NAND all the bits in each block must be cleared before new values can be written. In theory, this capability allows 3D XPoint to have higher performance and lower power consumption than NAND flash.
implementations
Both Micron and Intel have started to develop some interesting products based on the new 3D XPoint memory technology. For example:
- In 2017 the first 375 GB Intel Optane SSD DC P4800X based on 3D XPoint would arrive. A drive that acts as a buffer memory in computers with Intel processors and motherboards with compatible chipsets. According to Intel, the P4800X drive performed five to eight times faster than the company’s NAND flash-based DC P3700 in internal tests on shallow queues using a mixed workload. The P4800X can achieve up to 500,000 IOPS, or approximately 2 GBps, with a queue depth of 11.
- Micron QuantX would also arrive in 2017. It was a storage memory based on 3D XPoint.
However, this 3D XPoint memory has not yet caught on in the market due to various factors…
3D XPoint: advantages and disadvantages
Let’s look at some highlights from the 3D XPoint that should have powered this memory regardless:
- In the 3D XPoint architecture, data no longer has to be stored in 4KB blocks as in NAND flash memories. This stack of file I/O is slow, and with 3D XPoint everything is streamlined by being able to access small amounts of data faster and more efficiently.
- It’s not as fast as DRAM, but 3D XPoint is faster than NAND flash memory, which is a good thing considering that it’s also non-volatile memory.
- It is a versatile memory, since it can use both M.2 formats to use a PCIe bus, as well as be implemented in DIMM modules.
However, these improvements do not seem to have been enough for its expansion as we have said. And that can be due to other disadvantages such as:
- The new 3D XPoint technology is more expensive than conventional NAND flash.
- The speeds predicted by Intel and Micron have not been met. Some believe that because the PCIe bus used could be limiting its performance.
- It also has other added problems, and that is that to take advantage of it, you also have to make changes to the system. For example, the use of a compiler that can declare this memory as persistent so that applications can use system I/O to access it.
Cost
On the other hand, as we have mentioned previously, there was the issue of cost . For example, to give you an idea, we have that the 375 GB Intel Optane P4800X was priced at $1,520, that is, $4.05/GB. This is a lot compared to, say, the 400GB NAND flash-based Intel NVMe PCIe P3700 SSD, which cost $879, which works out to $2.20/GB.
I mean, it’s almost twice as expensive as NAND flash memory if you look at the cost per gigabyte…
XPoint 3D Use Cases
The 3D XPoint memory is used as an intermediate memory, a kind of cache or additional buffer between the main memory (DRAM) and the secondary storage media (NAND flash like SSDs) or also HDD drives.
In this way, the applications that are accessed more frequently or that need more performance, would be stored in this buffer 3D XPoint memory for faster access, while those that are not as frequent or do not need as much performance, would be stored. would be stored in conventional secondary memory.
Its precursors intend that this 3D XPoint memory be extended in more and more applications, but so far this has not been the case. Even Intel has ended up getting rid of the Optane division as you well know. Before that, Intel dreamed that this memory could be used for, for example:
- As part of the operating system’s virtual memory so that the operating system can move processes that don’t fit in RAM there instead of storing them in a pagefile.sys or SWAP on the slower drive.
- Optimization of specific applications.
- For fields such as data center servers and for HPC with Intel Xeons.
- Store large databases in such a way that they can be accessed quickly.
- Help overcome network bottlenecks in Big Data.
- Facilitate high-performance computing applications.
- Increase performance on cloud instances.
- Used as primary memory levels in hyperconverged systems.
