Intel plans to launch a new line of solid state storage devices under the Intel Optane brand in 2016. They’ll be the first solid state drives to use 3D Xpoint technology, which is said to offer up to 1,000 times the speed, 1,000 times the durability, and 10 times the density of current solutions.
3D Xpoint is a new type of non-volatile storage, which means that unlike DRAM, it can retain data without any power. This makes it more like the NAND flash storage used in today’s solid state drives. But it offers speeds that are closer to what you’d expect from DRAM than NAND (although DRAM is still faster).
What does that mean in terms of real-world performance? At the Intel Developer Forum keynote in San Francisco, Intel showed off a demo of a PC with an early version of an upcoming Intel Optane drive.
It was about 5 to 7 times as fast as an Intel P3700 series solid state disk using NAND flash storage, depending on the test. And this is just an early sample.
Intel says the first Optane chips will be “a new line of high-endurance, high-performance Intel SSDs” that will be available for a wide range of devices including services, desktops, and portable computers. It will be available both as a PCIe SSD and as a DIMM for use with Intel Xeon-powered products.
and 50 times the price
Not even close to that.
Wow, this is actually happening. I hope the price is right!
It won’t be, its going to be a PCI-e SSD. Its going to be very expensive.
OMG. RAM DRIVE SPEED?
PCIe becomes the bottleneck in the same way SATA3 did?
I had the same thought. At some point we rethink what kind of protocols and interfaces with these devices.
Their top end PCIe uses, what PCIe 3.0 x4 right? To get even x5 that performance you’d need x20 PCIe lanes. That’s all the lanes on a z170 mobo. So if Intel are launching this product next year it’s going to interesting to see how many lanes they build in to Skylake-E or the next gen of Xeon platform. 40x anyone?
Some day, but for now this technology will probably be around 4GB/s. Sata 3.0 x16 can do up to 32GB/s (real world speed will be lower, but still even at 75% of that speed, we still have a ways to go)
Speed is important yeah, but sounds like we’ll be able to take advantage of the increased density as soon as this comes out.
So long mechanical hard drives. It was nice knowing you.
You really thing PCI-e SSDs are going to put mechanical drives out of business?
These are going to be priced into the $1000’s.
We will need to wait for this memory technology to make its way to 2.5″ SATA SSDs. But they will be pushing the boundaries of SATA 3.0. SATA 3.2 will be needed.
The only thing that will put mechanical drives out of business is if normal 2.5″ SATA SSDs reach a place when they are no more than 4x more expensive than HDDs of equal storage.
On the client side, we could see the end of mechanical drives as a standard in PCs and laptops within two or three years, but it’s going to be a long time before solid state drives can compete on price when it comes to mass storage in the cloud.
SSDs already kills mechanical on speed. The reason for still using mechanical is price (either people want a cheapo laptop, or you want lots of storage and SSD speed isn’t needed, e.g., for an additional drive). This news doesn’t change any of that.
Intel’s earlier announcement said “up to 1000x faster than NAND”. I suppose that means that the technology may developed one day to reach that level of performance, but that is quite the difference.
I suppose this technology won’t benefit any consumers if they are only going to sell it in PCI-e, and DIMM varieties.
1000x most likely means write speed, which is the biggest bottleneck that NAND has (in the microsecond range whereas DRAM is in the nanosecond range). In the real world you’re never going to see anywhere close to that improvement due to the NAND controller hiding most of that time by using interleaving and buffering and other tricks.
The more I think about it, the less realistic a x1000 performance boost is for a *storage* device.
If Intel are creating faster SSDs with 3DXP, I cant see how the prototype could be using any interface other than a x16 lane PCIe 3.0 slot.
In that case the Optane will be limited to being 4 times faster than the current Intel P3700 which uses a x4 lane PCIe 3.0 slot. Any other gains must be from other characteristics like access latency I suppose.
To see the real speed of 3DXP, it would need to be built in to a CPU or some new interface on the mobo where it can act like a slow but persistent RAM disk. Hibernate to 3DXP for hardly noticeable low-power/zero-power modes, perhaps?
Traditional PCs are not going to get the biggest benefit from 3DXP. It’s going to be more useful in dedicated devices for things like image capture IMHO.
But Intel SSDs usually carry an un-Optane-nium pricetag unless there was a massive clearance sale.
Comments are closed.