Wednesday, July 15th 2020

SATA-IO Increases Interoperability Features with Revision 3.5 Specification

The Serial ATA International Organization (SATA-IO), the storage industry consortium dedicated to sustaining the quality, integrity and dissemination of the Serial ATA (SATA ) technology, today announced the publication of the SATA Revision 3.5 Specification. Specification 3.5 introduces features that enable increased performance benefits and promote greater integration of SATA devices and products with other industry I/O standards.

"SATA-IO has a long tradition of participating in the industry and supporting new innovative technology. These new features underscore the versatility of the SATA standard," said Jim Hatfield, SATA-IO president. "SATA remains one of the most stable, yet adaptable, interfaces in the industry. We are pleased to be able to enhance SATA's compatibility benefits for our members and the industry by adding features which allow SATA and other I/O standards to coexist in a variety of environments."
As storage technologies continue to advance, the industry continues to demand improvements in performance, reliability and compatibility. To ensure that members and the industry continue to have a reliable storage solution, the SATA-IO has added new features to the ubiquitous specification that offer improvements in a number of areas including:
  • Device Transmit Emphasis for Gen 3 PHY: aligns SATA with other characteristics of other I/O measurement solutions to help SATA-IO members with testing and integration.
  • Defined Ordered NCQ Commands: allows the host to specify the processing relationships among queued commands and sets the order in which commands are processed in the queue.
  • Command Duration Limit Features: reduces latency by allowing the host to define quality of service categories, giving the host more granularity in controlling command properties. The feature helps align SATA with the "Fast Fail" requirements established by the Open Compute Project (OCP) and specified in the INCITS T13 Technical Committee standard.
In addition, Specification 3.5 incorporates all of the latest T13 standard updates to give implementors the latest requirements for drives and other products and includes miscellaneous corrections and clarifications from the previous SATA 3.4 specification.

The specification is available to SATA-IO members for free, while non-members can purchase the specification by visiting this page.
Add your own comment

9 Comments on SATA-IO Increases Interoperability Features with Revision 3.5 Specification

#1
JAB Creations
If PCI-Express can literally double the bandwidth every few years why hasn't SATA bothered to do the same? I'm switching to half NVMe drives and half SATA-SSDs later this year and it'll be an awkward position to have to work with two form factors instead of having one unified, high bandwidth and easily replaceable set of drives and interfaces.
Posted on Reply
#2
R-T-B
JAB Creations
If PCI-Express can literally double the bandwidth every few years why hasn't SATA bothered to do the same?
They did, for a while...

Short version is the connector is tapped out.
Posted on Reply
#3
Jism
JAB Creations
If PCI-Express can literally double the bandwidth every few years why hasn't SATA bothered to do the same? I'm switching to half NVMe drives and half SATA-SSDs later this year and it'll be an awkward position to have to work with two form factors instead of having one unified, high bandwidth and easily replaceable set of drives and interfaces.
The advantage of S-ata is still hotswap, meaning you can pull the drive whenever you want while it's running. I dont see this happening with NVME.

On the other hand, NVME only offers benefits in certain workloads where the large speed advantage is plus, but S-ata can still cope and still be just as fast as your latest NVME drive.

S-ata needs to step up it's game.
Posted on Reply
#4
L'Eliminateur
Jism
The advantage of S-ata is still hotswap, meaning you can pull the drive whenever you want while it's running. I dont see this happening with NVME.

On the other hand, NVME only offers benefits in certain workloads where the large speed advantage is plus, but S-ata can still cope and still be just as fast as your latest NVME drive.

S-ata needs to step up it's game.
SATA is considered legacy, even before nvme gained traction.
the cost and changes necessary to duplicate SATA speeds were not trivial nor backwards compatible thus it was decided to EOL SATA at 6gbps and focus on nvme.

and yes, nvme is hardly necessary nor notable for the vast majority of users that would never even notice the difference, but it's the future (conversely, a single lane of pcie is faster than SATA, thus "cheap" nvme drives are starting to appear more and more) and we'll reap the benefits.

sata hotplug is a very very very small niche, nvme/pcie is hotplug but requires different form factor (U.2 like enterprise SSD drives that are contained within a case and have edge connectors designed for hotswapping), not for the consumer m2 drives on bare board
Posted on Reply
#5
CheapMeat
I think the benefit has more to do with the ability to mount a lot more drives with SATA bus than diverting PCIe lanes. Eventually with PCIe 5.0 and up, x1 lanes will become very plentiful even without the need of PLX-like switches. But for now, SATA (and SAS) allows for multiple drives, even "switches" that are much cheaper and easier to handle. Most people thinking probably only think in terms of gaming with their 1 or 2 drives maximum, so it doesn't matter to those folks at all.
Posted on Reply
#6
biffzinker
JAB Creations
If PCI-Express can literally double the bandwidth every few years why hasn't SATA bothered to do the same?
There was SATA Express but it didn’t take off. It combined SATA with PCIe on a split three connectors cable plug.
Posted on Reply
#7
ypsylon
I wonder, what's the point of tweaking a - basically - dead standard. SATA as an I/O for drives should've been replaced by SAS long time ago. Especially after SAS adopted SFF-8643 connector which propagated with same external design to U.2 (SFF-8639).

It's so ridiculous that if I want to plug on a desktop 8 drives I have to use 8 physical SATA cables, instead just 2 like SAS.

SATA irrelevancy is perfectly illustrated when you look at NVMe QLC drives which cost the same (or less) and just as well work as media storage for movies, games or other rarely moved stuff. So weird SATA is still around in its archaic form from 2008 (small tweaks here and there don't change the underlying inferiority of the standard compared to SAS and NVMe).
Posted on Reply
#8
bubbleawsome
ypsylon
SATA irrelevancy is perfectly illustrated when you look at NVMe QLC drives which cost the same (or less)
I can get a 4TB spinner for the same price as a 1TB nvme drive, or 6TB if i go a little over.
Posted on Reply
#9
L'Eliminateur
ypsylon
I wonder, what's the point of tweaking a - basically - dead standard. SATA as an I/O for drives should've been replaced by SAS long time ago. Especially after SAS adopted SFF-8643 connector which propagated with same external design to U.2 (SFF-8639).

It's so ridiculous that if I want to plug on a desktop 8 drives I have to use 8 physical SATA cables, instead just 2 like SAS.

SATA irrelevancy is perfectly illustrated when you look at NVMe QLC drives which cost the same (or less) and just as well work as media storage for movies, games or other rarely moved stuff. So weird SATA is still around in its archaic form from 2008 (small tweaks here and there don't change the underlying inferiority of the standard compared to SAS and NVMe).
If you read the press release, it's an update that's basically made for manufacturers(improvements to testing and isgnal integrity) and has marginal -if none- benefit for end user, i do see the benefit for enterprise SATA(which is still widely used), for example, the ordered queue for commands could lead to small improvements in high QD and the command timeout would lead to less "raid hang" when controllers wait forever for disks to timeout.
Those changes most likely can be implemented on software/FW alone.

the thing is that the PHY change needs new silicon, so forget we'll see anything anytime soon, most likely we'll see it in enterprise sata hdd and ssd and sas switches/controllers
Posted on Reply