NVM Express 1.3 Specs released: Ready for PCIe 4.0

[WMGroomAK]

According to an article on Hexus, the NVM Express Organization has released the specifications for NVMe Version 1.3. which includes several improvements over the Version 1.2 drives that began to hit the market in the 2H of 2016, including specs to take advantage of the upcoming PCIe 4.0 standards.  

 

http://hexus.net/tech/news/storage/107323-nvm-express-13-specification-released/

Quote

The NVM Express Organisation says that it has fulfilled its promises to add capabilities to meet the needs of its Enterprise, Cloud, and Client customers. Sanitise, Virtualization and Streams functionality have received particular attention, it says. Breaking down these three enhancements: Sanitise offers a simple, fast, native way to completely erase data; Virtualization provides more flexibility with shared storage use cases; and Streams offers better endurance for NAND-based SSDs.

 

 

Some other significant new features include advanced debugging tools for SSDs and more granular control of thermal throttling. Mobile computer product designers will likely be happy to hear that the interface has been enhanced to add a Boot Partitions feature, "which enables bootstrapping of an SSD in a low resource environment".

 

None of the above features and optimisations has lead to compromises in the high performance and low latency properties, central to the NVMe proposition. Thus, NVMe devices are "ready to take advantage of next generation bus speeds, including PCIe 4.0," boasts the standards organisation.

 

In other news the NVM Express Organisation shared details of its success over recent years. The first NVMe SSDs shipped in H2 2014, it says, yet by H1 next year it will be the highest volume SSD interface on the market (in terms of gigabytes shipped).

 

Devices following the NVMe 1.2 spec only started to launch in H2 last year and a two year ramp-up is typical for products to reach end users.

So now that we have the specs released, I would think that we could hope to see the Version 1.3 of an NVMe SSD hitting the market by early 2019 or so... I do like some of the features they are listing for things like controlling thermal throttling and shared storage uses.  Maybe by 2019 or 2020, these will be high enough volume and I'll hopefully be hitting an upgrade cycle where I can get one of them. 

 

NVM Express Organization Media Release: http://nvmexpress.org/nvme-revision-1-3-expands-reach-of-fast-storage-for-enterprise-client-and-cloud-power-users/

NVM Express Revision 1.3 Specs: http://www.nvmexpress.org/wp-content/uploads/NVM_Express_Revision_1.3.pdf

 

[captain_to_fire]

When will Samsung release the refresh to their M.2 NVMe SSDs? At the moment, the Samsung 960 Pro is the fastest consumer SSD with read/write speeds of 3500/2100 MB per second which is already vert fast. 

[DocSwag]

1 hour ago, hey_yo_ said:

When will Samsung release the refresh to their M.2 NVMe SSDs? At the moment, the Samsung 960 Pro is the fastest consumer SSD with read/write speeds of 3500/2100 MB per second which is already vert fast. 

It's been around 8 1/2 months since the 960 pro and evo so I'd assume we'd be seeing them in 4 months or so.

[captain_to_fire]

2 minutes ago, DocSwag said:

It's been around 8 1/2 months since the 960 pro and evo so I'd assume we'd be seeing them in 4 months or so.

As long as they keep using V-NAND

[Taf the Ghost]

H2'2018 should see the first PCIe 4.0 devices starting to hit the consumer-level. 

[LucidMew]

I could have sworn that this is a repost, especially since I've seen that graphic before, though that just might have been me reading the anandtech article. But in any case, it's been announced for a month now. http://www.anandtech.com/show/11436/nvme-13-specification-published-new-features

[NumLock21]

Isn't this a retoast?

[DocSwag]

1 hour ago, hey_yo_ said:

As long as they keep using V-NAND

There's no doubt they will lol, if they didn't that would be a huge mistake.

 

I see no good reason for not using v nand.

[WMGroomAK]

30 minutes ago, LucidMew said:

I could have sworn that this is a repost, especially since I've seen that graphic before, though that just might have been me reading the anandtech article. But in any case, it's been announced for a month now. http://www.anandtech.com/show/11436/nvme-13-specification-published-new-features

 

30 minutes ago, NumLock21 said:

Isn't this a retoast?

If it is, then sorry, however, I didn't find it when I searched the Tech News & Reviews for NVM OR 1.3.

[NumLock21]

2 minutes ago, WMGroomAK said:

 

If it is, then sorry, however, I didn't find it when I searched the Tech News & Reviews for NVM OR 1.3.

I think it might also be that Anandtech article I saw a while back and thought this might be a repost.

PCIe 4.0 was briefly mentioned on the WAN Show 2 weeks ago

 

[bcredeur97]

this will probably be the thing that brings people away from their sandy bridge machines in the long run...

 

 

at least one of the things. Good platform still, great CPU's. Just lacking in features.

[NumLock21]

The Anandtech article had more detail info about it

http://www.anandtech.com/show/11436/nvme-13-specification-published-new-features

[Antistatic12]

More news about NVMe and more people taking notice is good. The adoption rate seems pretty decent too. I see a few people in these forums with m.2 drives.
For me the SATA interface have been an issue of speed and aesthetics. I really look forward to building a new PC in a few years with only M.2 drives, and not having to deal with hiding SATA cables.

 

[tlink]

3 hours ago, Antistatic12 said:

More news about NVMe and more people taking notice is good. The adoption rate seems pretty decent too. I see a few people in these forums with m.2 drives.
For me the SATA interface have been an issue of speed and aesthetics. I really look forward to building a new PC in a few years with only M.2 drives, and not having to deal with hiding SATA cables.

then drive cages will be fully obsolete for most builds. thats a pretty huge development in computing if you think about it.

[leadeater]

The stream separation is pretty cool.

[captain_to_fire]

4 hours ago, DocSwag said:

There's no doubt they will lol, if they didn't that would be a huge mistake.

 

I see no good reason for not using v nand.

They shouldn't given that V-NAND is superior than TLC which is used by other inferior SSD vendor

[leadeater]

10 minutes ago, hey_yo_ said:

They shouldn't given that V-NAND is superior than TLC which is used by other inferior SSD vendor

@DocSwag was saying they will use V-NAND 

[DocSwag]

6 hours ago, hey_yo_ said:

They shouldn't given that V-NAND is superior than TLC which is used by other inferior SSD vendor

???

 

V-NAND is a completely different concept to TLC. Saying V-NAND is better than TLC is like saying 3 ALUs per core is better than 1 MB L2 cache per core. They're completely different metrics and comparing them makes no sense.

 

Samsung uses TLC in the 850 evo and 960 evo, both of which use V-NAND.

Edited June 28, 2017 by DocSwag
Wrote tlc instead of v-nand at one point :P

[leadeater]

5 minutes ago, DocSwag said:

Samsung uses TLC in the 850 evo and 960 evo, both of which use TLC.

Still it's 3D TLC and not planar TLC which is about as much difference as MLC vs TLC when talking performance and endurance. Isn't V-NAND Samsung name for 3D NAND technology, both TLC and MLC?

 

Edit:

They just call it 2bit V-NAND or 3bit V-NAND

[DocSwag]

Just now, leadeater said:

Still it's 3D TLC and not planar TLC which is about as much difference as MLC vs TLC when talking performance and endurance. Isn't V-NAND Samsung name for 3D NAND technology, both TLC and MLC?

Yeah it's just 3D NAND.

 

I'm not sure if 3D NAND actually directly impacts performance THAT much. I know for certain though that it's not actually the 3D-ness itself that increases endurance. It's rather that, for Samsung at least (I'm not sure if other nand manufacturers are doing what they did), because they used 3D NAND it mean they were able to reduce die space a lot so they basically said screw it and instead of using newer <20nm processes they went back to 40nm, which naturally greatly increase endurance.

 

I wouldn't really say planar mlc is comparable to 3D TLC, first of all you have to take process into account, and even then something like the 850 evo falls behind higher end planar mlc drives out there (Sandisk extreme pro for example). 

 

I still think the reason for such high performance on newer Samsung drives is maybe a bit because of 3D NAND but if I had to take a guess I would say that for the most part it's because of a better controller and firmware.

[Doobeedoo]

The stream separation is neat. By then we'll see some amazing SSDs!

[leadeater]

23 minutes ago, DocSwag said:

Yeah it's just 3D NAND.

 

I'm not sure if 3D NAND actually directly impacts performance THAT much. I know for certain though that it's not actually the 3D-ness itself that increases endurance. It's rather that, for Samsung at least (I'm not sure if other nand manufacturers are doing what they did), because they used 3D NAND it mean they were able to reduce die space a lot so they basically said screw it and instead of using newer <20nm processes they went back to 40nm, which naturally greatly increase endurance.

 

I wouldn't really say planar mlc is comparable to 3D TLC, first of all you have to take process into account, and even then something like the 850 evo falls behind higher end planar mlc drives out there (Sandisk extreme pro for example). 

 

I still think the reason for such high performance on newer Samsung drives is maybe a bit because of 3D NAND but if I had to take a guess I would say that for the most part it's because of a better controller and firmware.

True most of it is going back to 40nm but that is only possible due to going 3D which is why I say 3D TLC is roughly the same as 2D MLC, just have to be careful how you compare it etc.

 

It's also faster as well going back to 40nm

Quote

The larger cell structure also enables higher performance because it takes less iterations to program a cell. With planar TLC NAND it took multiple very high voltage pulses as well as numerous verification process to reach the right charge, but with looser voltage distribution the programming process has less steps and thus takes less time.

http://www.anandtech.com/show/8747/samsung-ssd-850-evo-review/3

 

As for 850 EVO vs SD Extreme Pro the SD isn't much faster in tests where it does win and it's behind in others. Other factors that make the 850 EVO is the far superior I/O consistency and I also believe higher steady state performance but I can't find the source that shows that (I'm sure I've seen it directly compared). A small amount of over provisioning also tips the scale even more to the 850 EVO.

 

I do think Samsung has the better controllers too but their NAND is slightly better than others and since they make all components of the SSDs, DRAM/Controller/NAND, I think that is the main reason they have the best I/O consistency being that they can really optimize and tweak everything the whole way through.

 

As you can tell I'm a Samsung SSD fan 

[BuckGup]

So we can finally run 960 pros in raid 0 and not get bottlenecked?

[DocSwag]

31 minutes ago, leadeater said:

True most of it is going back to 40nm but that is only possible due to going 3D which is why I say 3D TLC is roughly the same as 2D MLC, just have to be careful how you compare it etc.

 

It's also faster as well going back to 40nm

http://www.anandtech.com/show/8747/samsung-ssd-850-evo-review/3

 

As for 850 EVO vs SD Extreme Pro the SD isn't much faster in tests where it does win and it's behind in others. Other factors that make the 850 EVO is the far superior I/O consistency and I also believe higher steady state performance but I can't find the source that shows that (I'm sure I've seen it directly compared). A small amount of over provisioning also tips the scale even more to the 850 EVO.

 

I do think Samsung has the better controllers too but their NAND is slightly better than others and since they make all components of the SSDs, DRAM/Controller/NAND, I think that is the main reason they have the best I/O consistency being that they can really optimize and tweak everything the whole way through.

 

As you can tell I'm a Samsung SSD fan 

I was not aware larger process node meant higher performance.

 

In that case it probably would be fair to say tlc based on a larger process node would perform close to planar mlc based on a smaller process node.

 

I'm not completely certain but if I had to take a guess I would probably say most of the reason for Samsung's V NAND performing well would be from the larger process node not the 3D NAND, since a lot of other SSDs based on 3D NAND with a process node don't perform that well. However the 3D NAND is what enables the larger process node so... ¯\_(ツ)_/¯

[leadeater]

2 minutes ago, DocSwag said:

I was not aware larger process node meant higher performance.

 

In that case it probably would be fair to say tlc based on a larger process node would perform close to planar mlc based on a smaller process node.

 

I'm not completely certain but if I had to take a guess I would probably say most of the reason for Samsung's V NAND performing well would be from the larger process node not the 3D NAND, since a lot of other SSDs based on 3D NAND with a process node don't perform that well. However the 3D NAND is what enables the larger process node so... ¯\_(ツ)_/¯

Yep, chicken and egg something something breakfast lets enjoy the food . I just hope SSDs start getting cheaper again.