Intel's new Optane memory technology could lead to 1000 times faster MacBook storage

AppleInsider

A new ultra-fast memory technology from Intel and Micron branded as Optane is expected to launch in 2017, and could pave the way for even faster storage on Apple's MacBooks.

Optane is compatible with the NVMe storage protocol that Apple already uses in some of its MacBooks, as noted by Macworld. That means it could be relatively easy for Apple to transition and take advantage of the new technology.

Using 3D Xpoint technology, Optane is said to offer speeds 1,000 times faster and more durable than common flash storage. That's because it's 10 times denser than DRAM.

Apple launched Non-Volatile Memory Express support in its new 12-inch MacBook, and enabled it with a software update to OS X last April.

NVMe is meant to replace aging AHCI technology, which was designed to maximize performance in systems using spinning hard drives. NVMe is optimized for lower latency operations afforded by solid state media, or non-volatile memory.

Compared to AHCI, NVMe reduces latency by 50 percent. For example, Intel puts latency overhead in SCSI/SAS systems at 6 nanoseconds at 19,500 cycles, but quotes NVMe at 2.8 nanoseconds at 9,100 cycles (PDF link).

Further, NVMe is designed to scale over the next decade, a lifecycle comparable to AHCI, which was introduced in 2004. And since Apple has been using the PCIe transport protocol since 2011, the company is in a good position to expeditiously roll out support across the entire Mac lineup.

boeyc15

Does the 'data bus' speeds come into play here? Or said another way(not too clear to me in article) what does this translate to in real life usage?

[Ludicrous Speed!

bsimpsen

1000x is misleading. Random accesses to bytes distributed far and wide across the memory array will be much faster, as there is no Flash page load latency. But sequential accesses, which are far more common, will go no faster than the processor's memory interface. The real-life speed improvement will be far less than 1000x, and probably far less than 10x (as already shown in the chart).

We've got half a century of computer system design (both hardware and software) wrapped around the idea of large scale high speed sequential storage. The arrival of large scale high speed random storage doesn't change all of that. It'll take time for CPU designers to adapt cache strategies (or even eliminate cache), and for OS and app designers to adapt algorithms to take advantage of this new technology.

loquitur

With mass production "12-18 months away" (http://www.eetimes.com/document.asp?doc_id=1328682), Optane won't be appearing on
any near-term Mac refresh.

ireland

Someone, please in layman's engish: what speed would a MacBook Air's hard drive get with one of these drives in the real world? Anytime I ever hear of something that will be 1000x faster I reach for the salt.

paxman

The most significant upgrade to HHD based computers is a simple HHD / SSD swap. If this increases the speed still further that is only positive though the headline is ridiculously misleading. If my next computer was a 1000 times faster than the one I have now they'd have to build in obsolescence somewhere else. The damn thing would be good for a decade at least.

appex

Instant booting. Want it.

hmurchison

bsimpsen said:

1000x is misleading. Random accesses to bytes distributed far and wide across the memory array will be much faster, as there is no Flash page load latency. But sequential accesses, which are far more common, will go no faster than the processor's memory interface. The real-life speed improvement will be far less than 1000x, and probably far less than 10x (as already shown in the chart).

We've got half a century of computer system design (both hardware and software) wrapped around the idea of large scale high speed sequential storage. The arrival of large scale high speed random storage doesn't change all of that. It'll take time for CPU designers to adapt cache strategies (or even eliminate cache), and for OS and app designers to adapt algorithms to take advantage of this new technology.

This nails it. In talking storage systems the three key metrics are IOPS, Latency and Bandwidth. It's great that Optane has a nice bump in IOPS and Latency and i'm sure dba will absolutely love it and pay the premium but the bandwidth is going to be the really expensive part and I don't think Bandwidth in laptops is going to be up for the challenge for a bit of time.

ireland

Would be super if someone could answer me.

tallest skil

That’s nice. Will we ever actually SEE this happen, and at prices someone can actually afford?

mtbnut

Show me the money.

Now, if only software can keep up with its hardware counterparts. 

All this does is enable buggy software to show you the beach ball that much faster. 

Marvin

boeyc15 said:

Does the 'data bus' speeds come into play here? Or said another way(not too clear to me in article) what does this translate to in real life usage?

This memory is supposed to have price/performance somewhere between SSD and RAM. It's not 1000x faster at transfer than SSD, it shouldn't be more than about 15GB/s, SSD just now is ~2GB/s, RAM is about 60GB/s. It depends on where they attach it. There are ones that attach to the RAM slots:

http://www.legitreviews.com/intel-shows-off-512gb-optane-drive-with-3d-xpoint-memory-that-fits-in-ddr4-slot_176826

With DDR4, the Mac Pro for example could have 2 of the RAM slots with 64GB standard DDR4 each for 128GB RAM and the other 2 can have 512GB Crosspoint DIMMs for 1TB of storage that runs at a very fast speed and they can RAID the DIMMs.

This would allow the computer to boot and run very quickly. It would be like having the entire system in RAM.

However, it would also be more expensive than SSD. DDR4 memory is about $4/GB. MLC SSD is about $0.40/GB. This memory is expected to be somewhere in between e.g $2/GB = ~$2k for 1TB of storage.

It's more durable than SSD so it might be more cost-effective for servers. Typical users just want more storage for the price at this point.

Intel said they'd be able to produce a 10TB SSD around now and they said they'd have disruptive pricing:

http://www.pcworld.com/article/3032984/storage/intel-ssds-may-get-speed-capacity-boost-with-new-micron-chips.html

There's a 13TB 2.5" SSD from another manufacturer but $13k is a bit much:

http://www.pcworld.com/article/3021886/hardware/the-worlds-first-13tb-ssd-is-here.html

Micron is using 3D memory this year so we'll see if it has any benefits. The following page says that in June this year, the first-gen 3D NAND will offer 25% price drop and the second-gen a further 30%:

http://www.anandtech.com/show/10028/micron-3d-nand-status-update

Right now, Intel has drives like the following, 1.6TB for $969:

http://www.amazon.com/Intel-S3510-Internal-Solid-SSDSC2BB016T601/dp/B00WPCBFJ6

3D NAND in June would mean $727 and then next year $509 (~$0.30/GB). This would mean a 10TB 2.5" drive would be about $3k. Right now an 18TB Pegasus is $2600 and the SSD would be a very portable bus-powered 2.5".

sflocal

loquitur said:

With mass production "12-18 months away" (http://www.eetimes.com/document.asp?doc_id=1328682), Optane won't be appearing on
any near-term Mac refresh.

That's only 1yr-1.5yrs.  It'll be here before you know it.

Looking forward to that day!

sflocal

ireland said:

Someone, please in layman's engish: what speed would a MacBook Air's hard drive get with one of these drives in the real world? Anytime I ever hear of something that will be 1000x faster I reach for the salt.

that's kind of misleading because a Macbook Air hasn't hard a physical hard drive since what?  2008 I think?

Right now my late 2015 5K iMac does about 1.5GB/s (read) on the SSD.  So if I'm reading this right, does that mean we can expect a (roughly) 1.5TB/s read speed in 2-3 years?  If so... that's goes way past "ludicrous" speed..

dobby

I wonder if the technology is from the 3D RAM technology that IBM (and other companies) were researching in the Almaden labs in 2010. 

linkman

mtbnut said:

Show me the money.

Now, if only software can keep up with its hardware counterparts. 

All this does is enable buggy software to show you the beach ball that much faster. 

Oh, those programmers will be able to keep up. A 1000x improvement in hardware definitely means that the software can be 1000x slower to compensate. All improvements in hardware are negated by poor software. It's like an arms race. Think that stuff like Visual Basic and .Net were put in to make software perform faster? No way. It makes for some seriously lazy programmers.

staticx57

Marvin said:

boeyc15 said:

Does the 'data bus' speeds come into play here? Or said another way(not too clear to me in article) what does this translate to in real life usage?

This memory is supposed to have price/performance somewhere between SSD and RAM. It's not 1000x faster at transfer than SSD, it shouldn't be more than about 15GB/s, SSD just now is ~2GB/s, RAM is about 60GB/s. It depends on where they attach it. There are ones that attach to the RAM slots:

http://www.legitreviews.com/intel-shows-off-512gb-optane-drive-with-3d-xpoint-memory-that-fits-in-ddr4-slot_176826

With DDR4, the Mac Pro for example could have 2 of the RAM slots with 64GB standard DDR4 each for 128GB RAM and the other 2 can have 512GB Crosspoint DIMMs for 1TB of storage that runs at a very fast speed and they can RAID the DIMMs.

This would allow the computer to boot and run very quickly. It would be like having the entire system in RAM.

However, it would also be more expensive than SSD. DDR4 memory is about $4/GB. MLC SSD is about $0.40/GB. This memory is expected to be somewhere in between e.g $2/GB = ~$2k for 1TB of storage.

It's more durable than SSD so it might be more cost-effective for servers. Typical users just want more storage for the price at this point.

Intel said they'd be able to produce a 10TB SSD around now and they said they'd have disruptive pricing:

http://www.pcworld.com/article/3032984/storage/intel-ssds-may-get-speed-capacity-boost-with-new-micron-chips.html

There's a 13TB 2.5" SSD from another manufacturer but $13k is a bit much:

http://www.pcworld.com/article/3021886/hardware/the-worlds-first-13tb-ssd-is-here.html

Micron is using 3D memory this year so we'll see if it has any benefits. The following page says that in June this year, the first-gen 3D NAND will offer 25% price drop and the second-gen a further 30%:

http://www.anandtech.com/show/10028/micron-3d-nand-status-update

Right now, Intel has drives like the following, 1.6TB for $969:

http://www.amazon.com/Intel-S3510-Internal-Solid-SSDSC2BB016T601/dp/B00WPCBFJ6

3D NAND in June would mean $727 and then next year $509 (~$0.30/GB). This would mean a 10TB 2.5" drive would be about $3k. Right now an 18TB Pegasus is $2600 and the SSD would be a very portable bus-powered 2.5".

3D NAND has been out for a while now.

http://www.anandtech.com/show/8239/update-on-samsung-850-pro-endurance-vnand-die-size

lkrupp

The physical spinning hard drive is going the way of the Dodo bird and soon, just like the 3.5mm headphone jack.

blastdoor

How about a fusion drive with 64 GB of 3D-XPoint, 256 GB of "regular" SSD, and 4 TB of spinning platters?

cash907censored

As others have pointed out, the current bottleneck in regards to storage read/write speeds exists not on the storage medium itself but the interlink between processor and storage medium. It's awesome what Intel has done here, but until they either increase the amount or speed of the lanes between the two, you're not going to see any real world performance increase from this, as the current SSD drives are saturating the heck out their SATA 3/6GB or PCI Express connections right now. It's like owning a Ferrari when you live in NYC and never leave the city. 

VisualSeed

loquitur said:

With mass production "12-18 months away" (http://www.eetimes.com/document.asp?doc_id=1328682), Optane won't be appearing on
any near-term Mac refresh.

As slow as Apple is at refreshing hardware, it could very well be the next update.