SATA Express 101: Higher computer speeds are coming

[Julio Franco]

We were recently asked if the SF3700, LSI's latest flash controller, supports SATA Express and fired away with a bunch of other questions about the standard. The depth of this customer's curiosity suggested a broader need for education on the basics of the standard.

Editor’s Note:
This is a guest post by Kent Smith, senior director of marketing for LSI’s Flash Components Division,
overseeing all outbound marketing and performance analysis for the company.

To help me with the following overview of SATA Express, I recruited Sumit Puri, Sr. Director of Strategic Marketing for the Flash Components Division at LSI (SandForce). Sumit is a longtime contributor to many storage standards bodies and has been working with SATA- IO – the group responsible for SATA Express – for many years. He has first-hand knowledge of SATA- IO’s work.

Here are his insights into some of the fundamentals of SATA Express.

What is SATA Express?

Sumit: There’s quite a bit of confusion in the industry about what SATA Express defines. In simple terms, SATA Express is a specification for a new connector type that enables the routing of both PCIe® and SATA signals. SATA Express is not a command or signaling protocol. It should really be thought of as a connector that mates with legacy SATA cables and new PCIe cables.

Why was SATA Express created?

Sumit: SATA Express was developed to help smooth the transition from the legacy SATA interface to the new PCIe interface. SATA Express gives system vendors a common connector that supports both traditional SATA and PCIe signaling and helps OEMs streamline connector inventory and reduce related costs.

What is the protocol used in SATA Express?

Sumit: One of the misconceptions about SATA Express is that it’s a protocol specification. Rather, as I mentioned, it’s a mechanical specification for a connector and the matching cabling. Protocols that support SATA Express include SATA, AHCI and NVME.

What are the form factors for SATA Express?

Sumit: SATA Express defines connectors for both a 2.5” drive and the host system. SATA Express connects the drive and system using SATA cables or the newly defined PCIe cables.

What connector configuration is used for SATA Express?

Sumit: Because SATA Express supports both SATA and PCIe signaling as well as the legacy SATA connectors, there are multiple configuration options available to motherboard and device manufacturers. The image below shows plug (a) which is built for attaching to a PCIe device. Socket (b) would be part of a cable assembly for receiving plug (a) or a standard SATA plug, and Socket (c) would mount to a backplane or motherboard fir receiving plug (a) or a standard SATA plug. The last two connectors are a mating pair designed to enable cabling (e) to connect to motherboards (d).

 

When will hosts begin supporting SATA Express?

Sumit: We expect systems to begin using SATA Express connectors early this year. They will primarily be deployed in desktop environments, which require cabling. In contrast, we expect limited use of SATA Express in notebook and other portable systems that are moving to cableless card-edge connector designs like the recently minted M.2 form factor. We also expect to see scant use of SATA Express in enterprise backplanes. Enterprise customers will likely transition to other connectors that support higher speed PCIe signaling like the SFF-8639, a new connector that was originally included in the SATA Express specification but has since been removed.

Will LSI support SATA Express?

Sumit: Absolutely. Our SF3700 flash controller will be fully compatible with the newly defined SATA Express connector and support either SATA or PCIe. Our current SF-2000 SATA flash controllers support SATA cabling used on SATA Express, but not PCIe.

Will LSI also support SRIS?

Sumit: PCIe devices enabled with SRIS (Separate Refclk Independent SSC) can self-clock so need no reference clock from the host, allowing system builders to use lower cost PCIe cables. SRIS is an important cost-saving feature for cabling that supports PCIe signaling. It doesn’t support card-edge connector designs. Today the SF3700 supports PCIe connectivity, and LSI will support SRIS in future releases of SF3700 and other products.

Why is it called SATA Express?

Sumit: SATA Express blends the names of the two connectors and captures the hybridization of the physical interconnects. The name reflects the ability of legacy SATA connectors to support higher PCIe data rates to simplify the transition to PCIe devices. SATA Express can pull double duty, supporting both PCIe and SATA signaling in the same motherboard socket. The same SATA Express socket accepts both traditional SATA and new PCIe cables and links to either a legacy SATA or SATA Express device connector.

How fast can SATA Express run?

Sumit: The PCIe interface defines the top SATA Express speed. A PCIe Gen2 x2 device supports up to 900 MB/s of throughput, a PCIe Gen3 x2 device up to 1800 MB/s of throughput – both significantly higher than 550 Mb/s speed ceiling of today’s SATA devices.

Is SATA Express similar to M.2?

Sumit: There are two key similarities. Both support SATA and PCIe on the same host connector, and both are designed to help transition from SATA to PCIe over time.

SATA Express delivers the future of connector speeds today

SATA Express was born of the stuff of all great inventions. Necessity. The challenge SATA-IO faced in doubling SATA 6 Gb/s speeds was herculean. The undertaking would have been too time-consuming to support the next-generation connection speeds that PCIe answers. It would have been too involved, requiring an overhaul of the SATA standard. Even in the brightest scenario, the effort would have produced a power guzzler at a time when greater power efficiency is a must for system builders. SATA-IO found a better path, an elegant bridge to PCIe speeds in the form of SATA Express.

Permalink to story.

https://www.techspot.com/news/55662-sata-express-101.html

 

[madboyv1]

I'm moving Terabytes of data right now as I type this... I could REALLY use drives that can go 1800 MB/s.

 

[TheBigFatClown]

I hate to be a negative Nancy but I don't see the real point of this. It sounds like it's just something created to last for a short while. It just looks really confusing. I have to say 'thanks, but no thanks'.

 

[VitalyT]

I agree, it is too small a window of extra performance, to expect manufacturers to jump into this one, I don't see the point either. It's like being excited about the coming of USB 3.1 that at best will only double the performance.

This interface is supposed to provide an incentive for those who design PCI-Express cards to switch over, but it doesn't at all, the speeds are too low by comparison.

The already available Thunderbolt 2 is a better interface, with theoretical throughput of 2.5GByte per second.

 

[p3ngwin]

I hate to be a negative Nancy but I don't see the real point of this. It sounds like it's just something created to last for a short while. It just looks really confusing. I have to say 'thanks, but no thanks'.

it's intentional that it's "short-term", because it's a way of evolving two previously incompatible standards towards compatibility, and then towards the further goal of making a unified protocol for systems, mainly focussing on PCIe.

if you can remember back when PC's had dozens of protocols, from "printer ports" to "game ports", and "keyboard/mouse" PS2 ports.......it was ridiculous.

USB was a monumental paradigm shift, and we're continuing the trend towards a unified protocol for everything.

It's happening in the processors, from dedicated CPU's and GPU's towards "Fusion" style So'C, and it's happening in languages, from OpenCL, and other Heterogeneous initiatives.

This is simply one step in consolidating previously many connectors and protocols, into a more unified standard that simplifies things further.

 

[Jad Chaar]

I hate to be a negative Nancy but I don't see the real point of this. It sounds like it's just something created to last for a short while. It just looks really confusing. I have to say 'thanks, but no thanks'.

You may be right, but it is an important stepping stone between SATA and more expensive options such as PCIe SSDs.

 

[Martina Thomas]

Great news. This will speed up data transfer more which will benefit overall computing.

 

[theBest11778]

I agree, it is too small a window of extra performance, to expect manufacturers to jump into this one, I don't see the point either. It's like being excited about the coming of USB 3.1 that at best will only double the performance.

This interface is supposed to provide an incentive for those who design PCI-Express cards to switch over, but it doesn't at all, the speeds are too low by comparison.

The already available Thunderbolt 2 is a better interface, with theoretical throughput of 2.5GByte per second.

You're actually upset about doubling performance? Imagine if Intel's next chips doubled performance across the board without increasing TDP. That would be insane. It's logical that everything needs to move to PCI-E, but we all have SATA drives, and bet you don't want to have to rebuy 2x 2TB or 4TB HDDs, AND replace your SSDs (or whatever configuration you're running.) I know I don't. Storage drives don't come close to pushing SATA II let alone SATA III. No point in rendering current tech obsolete without a replacement. The day is coming where 1-2TB SSDs will be "affordable" Sub $300 range, but that's not for a few years (node changes.) Considering I converted all my media to a server I need 8TB minimum for that, and 4K/UHD will jump that towards 16-32TB of storage. Mechanical HDDs aren't going anywhere so this is a great compromise.

 

[VitalyT]

Imagine if Intel's next chips doubled performance across the board without increasing TDP.

That's not an appropriate comparison. Increase in CPU performance doesn't require anyone to change anything or to buy new equipment, whereas the change in storage protocol is a huge implication of changes for the hardware to support. And such change need to be justified, and by today's perception that means a good future provision, which this new interface doesn't offer.

 

[lipe123]

Imagine if Intel's next chips doubled performance across the board without increasing TDP.

That's not an appropriate comparison. Increase in CPU performance doesn't require anyone to change anything or to buy new equipment, whereas the change in storage protocol is a huge implication of changes for the hardware to support. And such change need to be justified, and by today's perception that means a good future provision, which this new interface doesn't offer.

The whole idea behind this is NOT to have to change anything, did you read?
You'd be able to use existing SATA drives on the new tech once it comes out.

 

[VitalyT]

The whole idea behind this is NOT to have to change anything, did you read?
You'd be able to use existing SATA drives on the new tech once it comes out.

I read it alright, the SSD manufacturers will have to make the changes to be able to take advantage of the new interface, or it's all for naught.

 

[lipe123]

The whole idea behind this is NOT to have to change anything, did you read?
You'd be able to use existing SATA drives on the new tech once it comes out.

I read it alright, the SSD manufacturers will have to make the changes to be able to take advantage of the new interface, or it's all for naught.

You're completely missing the point, this is a interim solution that allows drive manufactures direct access to PCI-E bus (which thunderbolt also works from so no way in hell is it going to be better than direct access to the same bus) and still allow older sata drives to work.

Instead of doing a major break like the changeover from IDE to SATA this gives people backward compatibility!

 

[TheBigFatClown]

You're completely missing the point, this is a interim solution that allows drive manufactures direct access to PCI-E bus (which thunderbolt also works from so no way in hell is it going to be better than direct access to the same bus) and still allow older sata drives to work.

Instead of doing a major break like the changeover from IDE to SATA this gives people backward compatibility!

I don't like it. It will confuse people more than anything. Why worry about backwards compatibility? They don't most of the time. Intel makes you buy a new motherboard every year or so. Constantly upgrading hardware all the time is expected in the world of technology. Why so much emphasis on a short-term implementation that will simply confuse most people for the sake of keeping an old SATA hard drive.

Unless I am way off base on what this actually is. Would it be correct to say that it is nothing more than an DVI-HDMI adapter of sorts? If that's the case then I guess it would be acceptable. But, only if, that's the case.

 

[cliffordcooley]

I'm with TBFC, this all seems as if it is a desperate attempt for faster speeds and a shortcut till the next best thing. This will likely only cater to the high-end/commercial user.

 

[Guest]

I wish this was the case a few years ago for laptops. Last year I bought my current laptop to replace my old, heavily-used netbook. It used a PCIe card as its hard drive and came with a small amount of storage, so I paid $250 and change for a larger one (in 2010 dollars for 64GB). Today, that very same drive costs $120 because it's PCIe. The price went down, but still, compared to mSATA, it's too much (right now I could get the same size in mSATA for $70).

My current laptop is a ThinkPad. One of the deciding factors for my purchase was that it came with an mSATA slot. I figured "that's great, I can just take the card out of the old computer and put it into the new one and not have to set anything up (other than a few drivers, maybe)". You already know what happened next. I found out the hard way that it wasn't compatible, despite the perfect fit into the slot. So now I just use the PCIe card as an external backup. It's mostly sufficient because I don't have very many files.

 

[Guest]

So as I am understanding it, they are trying to move storage drives to a cable system that is directly connected to the PCIe bus? This sounds awesome! I can imagine now that video cards could be connected directly to the PCIe bus via a cable! That means that we could get a much needed shift from the ATX form factor that is so restricting in space and positions. This could finally take Personal computers and servers out of the stone age of computers, finally doing away with the age old ATX form factor! I am very interested in seeing how this turns out!

 

[Guest]

SATA would be 550MB/s, not Mb/s

 

[Darth Shiv]

I hate to be a negative Nancy but I don't see the real point of this. It sounds like it's just something created to last for a short while. It just looks really confusing. I have to say 'thanks, but no thanks'.

The point is of course to move (or remove) a bottleneck. It has been a VERY long time since disk bus interfaces were a big bottleneck. With SSDs, the SATA3 interface has been saturated for 3-4 years. Pretty poor.

We could be running SSDs that are 2-3+GB/s rather than 600MB/s if the interface wasn't holding us back.

Then you ask "What would I do with that speed?" Well without having the speed, you'll never know. People are change averse. Google is rolling out 10gbit fibre because they have next gen services that need that sort of speed. Not CURRENT services. You can't rollout these new services if the infrastructure isn't there to make it so there are enough customers to make it economically viable. You have to build the capability/capacity first *and* have a lot of people start using it.

So in short, unless you are a visionary with great accuracy, who knows what *you* will find useful with the extra bandwidth.

I know I use large databases and run large programs. Load times make a significant difference to my day. Restoring a database backup and updating it can take me 10 minutes to an hour. The machine is practically unusable during this time... very light tasks possible. If you halve or quarter that, the productivity boost is enormous...

To prove how much of a difference this will make to yours, you could run a ramdrive and see how much of a bottleneck your disk/current SSD is.

 

[TheBigFatClown]

The point is of course to move (or remove) a bottleneck. It has been a VERY long time since disk bus interfaces were a big bottleneck. With SSDs, the SATA3 interface has been saturated for 3-4 years. Pretty poor.

We could be running SSDs that are 2-3+GB/s rather than 600MB/s if the interface wasn't holding us back.

Then you ask "What would I do with that speed?" Well without having the speed, you'll never know. People are change averse. Google is rolling out 10gbit fibre because they have next gen services that need that sort of speed. Not CURRENT services. You can't rollout these new services if the infrastructure isn't there to make it so there are enough customers to make it economically viable. You have to build the capability/capacity first *and* have a lot of people start using it.

So in short, unless you are a visionary with great accuracy, who knows what *you* will find useful with the extra bandwidth.

I know I use large databases and run large programs. Load times make a significant difference to my day. Restoring a database backup and updating it can take me 10 minutes to an hour. The machine is practically unusable during this time... very light tasks possible. If you halve or quarter that, the productivity boost is enormous...

To prove how much of a difference this will make to yours, you could run a ramdrive and see how much of a bottleneck your disk/current SSD is.

Please forgive me if I ever gave the impression that I was against improvements in technology. The words 'faster' and 'cheaper' always give me multiple orgasms. I simply think the technology we are talking about is confusing. JEDEC have just released the final specs for DDR4. To that I say 'hell yeah'!! Easy to understand, 4th generation DDR memory that is faster than the previous generation. If this Sata Express is the future than so be it. I just don't quite understand it. I hope it's not just a temporary stop gap. I guess if you wanted to be a wise guy you could say DDR4 is just a stop gap until we get to DDR5 but that's going a little too far in my opinion.

 

[Guest]

This is basically a way to lug sata drives into a pcie connector.. and theyb will be able to go faster, if possible.
gfx card wont work as they are not self=clocking. But yeah this is a midway step to full pcie. Cheapness of cabling is a consideration. Also 1Tb affordable SSD will be here before spring 2015,
cheap sleep in Ram is best tho.