So is SATA really done at 6?

[Ketchup]

I was just a little surprised. Things were going along fine with SATA: 1.5, 3, 6, then

But there was a problem. SATA 6Gbps still wasn't fast enough to meet the needs of SSD manufacturers as they were already able to saturate it. SATA-IO was given a difficult task: they would have to come up with a new standard with drastically better performance only a few years after the previous strandard had been announced. Not only would it have to be faster, but it also needed to be cost and power efficient. Instead of developing the SATA protocol further, which would have been expensive and time consuming, SATA-IO decided to utilize an existing interface found in every mainstream computer: PCI Express.

Part of me just wants a bit more explanation I suppose. What makes going from 6 to (throw in number here) harder than going from SATA to a completely new interface like M.2 (and getting Intel and manufacturers to go along with it).

So does anyone want to show me what I'm missing?

 

[dave_the_nerd]

I was just a little surprised. Things were going along fine with SATA: 1.5, 3, 6, then



Part of me just wants a bit more explanation I suppose. What makes going from 6 to (throw in number here) harder than going from SATA to a completely new interface like M.2 (and getting Intel and manufacturers to go along with it).

So does anyone want to show me what I'm missing?

Well... one thing you might be missing is that SATA-IO is an industry organization made up of a bunch of companies.

http://www.serialata.org/membership-listing

So, basically, when SATA-IO makes a decision, it's because Intel, etc., have already decided to go along with it. There may have been some companies that didn't think it was a good idea, but when the 800 pound gorillas make a decision, well, you follow or you die. Yay capitalism.

As far as why it's easier to do one thing than another, I don't really know. I could speculate, but I won't because I'll probably just make an ass of myself.

I'm sure the engineers have their reasons.

 

[Cerb]

What makes going from 6 to (throw in number here) harder than going from SATA to a completely new interface like M.2

M.2 is not a new [electrical] interface. It's two existing interfaces. For SATA, it's just another physical form factor, like mSATA.

So does anyone want to show me what I'm missing?

What do you need a dedicated SATA interface for, when you have all those PCIe lanes? Oh, and those PCIe lanes keep getting faster and faster, too.

Having SATA means more space, more connectors, and more power, and decades of ATA protocol baggage, plus the R&D costs of improving the bandwidth yet again (they could have ripped off SAS 1.2Gbps...). For HDDs, no big deal. For SSDs, go check out the low-QD performance of the new NVMe server drives. The future is even faster SSDs than those...but not with ATA.

 

[Tsavo]

I'm still waiting for SATA floppy drives.

This industry moves so slowly sometimes.

 

[master_shake_]

well sata express is 10gbps...

 

[Eeqmcsq]

Maybe they don't simply bump the SATA speed up is due to heat? Kind of like for desktop CPUs, we don't simply increase the GHz of the CPU to 8 GHz or 10 GHz due to heat.

 

[Ketchup]

I'm still waiting for SATA floppy drives.

This industry moves so slowly sometimes.

I went ahead and pre-ordered my M.2 CD-ROM. Thanks for the insight. So, I suppose since it has been decided by the big boys, and the big big boys have decided to be the first to include it on a chipset for once, it will be around for a while.

It will be interesting though to see how far it goes. Like server. I don't see those changing from traditional SATA, unless the mechanical drive market starts changing just for consistency.

 

[Jeff7181]

It's technology. It evolves.

These types of discussions happen constantly in the IT world. Why go from ISA to PCI? Why go from PCI to PCI-Express? Why go from AGP to PCI-Express? Why go from PATA to SATA? Why go to DDR4 rather than just make DDR3 faster? Why do I need Windows XP when Windows 2000 is running just fine? Why make Windows Vista or Windows 7 - why not just make Windows XP better?

The answer is always the same. Eventually technology reaches a limitation (whether that's a physical limitation, a practical limitation, etc.) and a new technology is required to overcome that.

To address your question specifically, there's no reason SATA/SAS can't be extended to 12 Gbps, in fact, you'll likely see 12 Gbps SAS for enterprise level equipment. But why bother if there's a lower latency, higher bandwidth interface that can be used?

Honestly, now that flash is more affordable and prevalent, M.2 or a similar interface just makes sense. With spinning platters, a cable connection was basically a necessity - you can't really plug a 1-2 pound spinning disk drive into a DIMM slot without significant support for it. But with flash, it's just another set of chips. I wouldn't be surprised to see something more like a traditional DIMM slot for RAM replace M.2 and SATA will likely stick around to support legacy devices like optical drives and external drives. Or maybe they'll just switch to USB 3 (or 4, or 5 or whatever).

 

[BrightCandle]

Along with the move away from SATA we are also replacing the protocol to be something far more sensible for an SSD. SATA and all its IO protocol make lots of sense when you have a spinning disk that takes 10ms to find anything and has to reorder its reads to get the best performance but an SSD just doesn't have those limits. So along with going for a higher bandwidth and lower latency connection, and dropping into an existing standard for PCI-E we are also going to getting a protocol adjustment to make the use of an SSD much more efficient.

You can keep using SATA and cables with SATA express, its not quite as fast and its not as efficient but it will be the budget option compared to PCI-E and M.2 drives in the future I suspect. Finally they just decided it was time to give SSDs a nice primary place on the machines to make them better. Happens all the time, PATA only lasted about 15 years and SATA has had a similar length of time in the industry. These things just change.

 

[VirtualLarry]

To address your question specifically, there's no reason SATA/SAS can't be extended to 12 Gbps, in fact, you'll likely see 12 Gbps SAS for enterprise level equipment. But why bother if there's a lower latency, higher bandwidth interface that can be used?

This is what I don't get. Why didn't they released 12Gbit SATA at the same time as 12Gbit SAS? Only because technology companies are getting really greedy, and don't want "consumers" to get the same technology as "enterprise". That's why we don't get ECC anymore (and suffer from random memory errors), and why consumer PCs will be exclusively Atom-based in the future, you can already see the writing on the wall, with Atom PCs taking over the low-end, with future Celerons and Pentiums being exclusively Atom-based, with Core being reserved for the higher-end. Eventually, Core will only be found in servers, and Intel's server-derived Enthusiast line, with CPU prices north of $500.

Edit: And they wonder why the consumer PC industry is in a slump...

 

[Ketchup]

OK, I think I get it now. Since SATA was already there, SSD manufacturers just adapted to that until it was able to convince the industry that SATA wasn't cutting it.

This would make me wonder if, as we make the transition to the new interface, the drives will start getting faster again. Not that they are slow now, but seemed like they hit a speed wall a couple years ago IMO.

 

[Jeff7181]

This is what I don't get. Why didn't they released 12Gbit SATA at the same time as 12Gbit SAS? Only because technology companies are getting really greedy, and don't want "consumers" to get the same technology as "enterprise". That's why we don't get ECC anymore (and suffer from random memory errors), and why consumer PCs will be exclusively Atom-based in the future, you can already see the writing on the wall, with Atom PCs taking over the low-end, with future Celerons and Pentiums being exclusively Atom-based, with Core being reserved for the higher-end. Eventually, Core will only be found in servers, and Intel's server-derived Enthusiast line, with CPU prices north of $500.

Edit: And they wonder why the consumer PC industry is in a slump...

That's a pretty pessimistic view of the industry. I don't think it's accurate.

Regardless... the majority of consumers won't benefit from 12Gig SATA. The ones that would will have M.2 flash drives or whatever the standard becomes.

Enterprise gets the 12Gig SAS because they use it. There are disk shelves getting as dense as 48 3.5 inch spindles in 4U. You need a low latency, high bandwidth interface for that. Even with multi-lane SAS, when you pile that many spindles on 4 SAS channels and do a ton of sequential IO, you're quickly going to have a bottleneck at the uplink.

 

[Jeff7181]

OK, I think I get it now. Since SATA was already there, SSD manufacturers just adapted to that until it was able to convince the industry that SATA wasn't cutting it.

This would make me wonder if, as we make the transition to the new interface, the drives will start getting faster again. Not that they are slow now, but seemed like they hit a speed wall a couple years ago IMO.

The answer is simply, yes. Flash won't get any slower and once there's an interface that allows lower latency and higher throughput, manufacturers will use it. There's no point in engineering a single consumer SSD that can do 2000 MB/sec at sub-ms latency with SATA3.

OCZ did this early on with their Revo Drive or whatever they called it. NAND flash on a PCI-Express card. Did at least 2-3 times the IO and throughput as SSD on 3Gig SATA at the time.

 

[VirtualLarry]

The answer is simply, yes. Flash won't get any slower and once there's an interface that allows lower latency and higher throughput, manufacturers will use it. There's no point in engineering a single consumer SSD that can do 2000 MB/sec at sub-ms latency with SATA3.

OCZ did this early on with their Revo Drive or whatever they called it. NAND flash on a PCI-Express card. Did at least 2-3 times the IO and throughput as SSD on 3Gig SATA at the time.

You do know that OCZ Revodrive was just a couple of SATA SSDs, minus cases, on a card with a PCI-E RAID controller, right?

 

[VirtualLarry]

That's a pretty pessimistic view of the industry. I don't think it's accurate.

Regardless... the majority of consumers won't benefit from 12Gig SATA. The ones that would will have M.2 flash drives or whatever the standard becomes.

I'm not so sure that the consumer is going to benefit from the proliferation of "next-gen" SSD interfaces either. They will have to purchase new motherboards (with new CPUs), and new storage devices as well.

SATA was and is a well-established standard, and one of it's benefits is backwards and forwards compatibility, much like PCI-E.

12Gbit SATA would have been a natural, easy, evolution, and most of all, wouldn't be disruptive to the ecosystem.

Can you buy an M.2 SSD and put it on an older Core2Quad system? NOPE!

SATA offers flexibility, and one of the popular uses of SSDs, is to upgrade older systems and increase their performance.

With these newer interfaces, and mfg's spending R&D on them, then they won't support R&D for older interfaces, and older systems will be left out in the cold.

Edit: And I can't help but think that SSD mfgs are heading straight towards a brick wall, by adapting their products to utilize PCI-E in all it's many forms, since Intel's goal is to ELIMINATE PCI-E on consumer platforms, at least. (In order to cut off AMD's and NVidia's "air supply". You can thank Nvidia, and a lawsuit, and a federal judgement, for the fact that current PCs even still HAVE PCI-E slots.)

 

[Jeff7181]

You do know that OCZ Revodrive was just a couple of SATA SSDs, minus cases, on a card with a PCI-E RAID controller, right?

Of course, that's pretty much the only difference. The fact that it was connected to a PCI-Express bus instead of SATA allows lower latencies and higher throughput.

 

[Jeff7181]

I'm not so sure that the consumer is going to benefit from the proliferation of "next-gen" SSD interfaces either. They will have to purchase new motherboards (with new CPUs), and new storage devices as well.

SATA was and is a well-established standard, and one of it's benefits is backwards and forwards compatibility, much like PCI-E.

12Gbit SATA would have been a natural, easy, evolution, and most of all, wouldn't be disruptive to the ecosystem.

Can you buy an M.2 SSD and put it on an older Core2Quad system? NOPE!

SATA offers flexibility, and one of the popular uses of SSDs, is to upgrade older systems and increase their performance.

With these newer interfaces, and mfg's spending R&D on them, then they won't support R&D for older interfaces, and older systems will be left out in the cold.

Edit: And I can't help but think that SSD mfgs are heading straight towards a brick wall, by adapting their products to utilize PCI-E in all it's many forms, since Intel's goal is to ELIMINATE PCI-E on consumer platforms, at least. (In order to cut off AMD's and NVidia's "air supply". You can thank Nvidia, and a lawsuit, and a federal judgement, for the fact that current PCs even still HAVE PCI-E slots.)

SATA's not going away tomorrow. It'll be here for years to come. That doesn't mean newer and better technologies shouldn't be investigated. You're back to the same "Windows XP worked fine, why do I need a new OS?" argument.

 

[tweakboy]

You got ASUS dual 790GTX ,gl

 

[Cerb]

Can you buy an M.2 SSD and put it on an older Core2Quad system? NOPE!

You might want to search before you go there (P.S. PCIe ones exist, too)...because, well, you can 🙂.

 

[Cerb]

Of course, that's pretty much the only difference. The fact that it was connected to a PCI-Express bus instead of SATA allows lower latencies and higher throughput.

No, it didn't. Latencies were nothing special, and throughput was increased because it was a RAID 0. Intel and LSI RAID 0s generally outperformed them. The RevoDrives were SATA SSD RAID 0s, crammed into one or two boards, running on a SATA RAID controller..

 

[Lonyo]

I'm not so sure that the consumer is going to benefit from the proliferation of "next-gen" SSD interfaces either. They will have to purchase new motherboards (with new CPUs), and new storage devices as well.

SATA was and is a well-established standard, and one of it's benefits is backwards and forwards compatibility, much like PCI-E.

12Gbit SATA would have been a natural, easy, evolution, and most of all, wouldn't be disruptive to the ecosystem.

Can you buy an M.2 SSD and put it on an older Core2Quad system? NOPE!

SATA offers flexibility, and one of the popular uses of SSDs, is to upgrade older systems and increase their performance.

With these newer interfaces, and mfg's spending R&D on them, then they won't support R&D for older interfaces, and older systems will be left out in the cold.

Edit: And I can't help but think that SSD mfgs are heading straight towards a brick wall, by adapting their products to utilize PCI-E in all it's many forms, since Intel's goal is to ELIMINATE PCI-E on consumer platforms, at least. (In order to cut off AMD's and NVidia's "air supply". You can thank Nvidia, and a lawsuit, and a federal judgement, for the fact that current PCs even still HAVE PCI-E slots.)

SATA 12GBps is already too slow for next gen SSDs. So it's utterly pointless. It would be "sufficient" for most drives for now, but then we will need an ever faster interface almost instantly to maximise the potential of drives.

So why not use SATA 12GBps? Because it's already too slow.
1100MB/s from SATA 12GBps (SATA6 tops out around 550MB/s)

http://www.anandtech.com/show/8152/computex-2014-gskill-demos-pcie-ssds
1800MB/s from a PCIe drive.

http://www.anandtech.com/show/8147/the-intel-ssd-dc-p3700-review-part-2-nvme-on-client-workloads/3
1500MB/s

You are saying they should use a 1100MB/s interface for the next gen drives that can already exceed that. So basically we already need SATA-18GBps right now.

And as for M.2 on a Core 2 system, um... you don't need that much speed. Also, you can get an M.2 -> PCIe adapter card if you really want, so yes, you can use it on your Core 2 system, although it would be a complete waste.

Instead, they introduced SATA-Express, which is SATA or PCIe, so it's backwards compatible, and will work up to 1969MB/s, which is almost already too slow.

Instead of the otherwise usual approach of doubling the speed of SATA interface, PCI Express was selected for achieving data transfer speeds greater than 6 Gbit/s. It was concluded that doubling the native SATA speed would take too much time, too many changes would be required to the SATA standard, and would result in much greater power consumption when compared to the existing PCI Express bus

https://www.sata-io.org/sites/default/files/documents/SATA Express - CS 2013.pdf

Oh, and if you want faster interface speed, YOU NEED TO BUY A NEW MOBO ANYWAY. You can't upgrade your SATA-6 to SATA-12 just by magic. Your speed will be limited to your SATA interface on your motherboard.
Manufacturers aren't stopping making SATA drives because they can also make PCIe/M.2 drives, they co-exist and you buy what you need.
You don't seem to be thinking things through sufficiently.

 

[Cerb]

12Gbit SATA would have been a natural, easy, evolution, and most of all, wouldn't be disruptive to the ecosystem.

That I don't disagree with. But, it looks like the storage companies decided they wanted to be a bit disruptive, and I'm sure they didn't decide to go this way without weighing their collective long-term objectives and the costs (not yours or mine, but theirs). I would have liked to see one more SATA gen before it going away, but they've decided that layering ATA onto PCIe is the way to keep ATA alive, and that it's time to move off of ATA for SSDs en masse (the latter is definitely a good thing).

 

[Cerb]

Edit: And I can't help but think that SSD mfgs are heading straight towards a brick wall, by adapting their products to utilize PCI-E in all it's many forms, since Intel's goal is to ELIMINATE PCI-E on consumer platforms, at least. (In order to cut off AMD's and NVidia's "air supply". You can thank Nvidia, and a lawsuit, and a federal judgement, for the fact that current PCs even still HAVE PCI-E slots.)

...and what were they going to replace PCIe with? PCIe is used even in dirt-cheap notebooks.

 

[Cloudfire777]

Take your pick.
SATA have become useless for SSDs. It was designed for HDDs and the controller adds latency.
NVMe is here to replace it and is tailored for fast NANDs. In fact, the Intel NVMe SSD is so fast, the CPU have become the bottleneck.
Makes you realize how good NVMe standard is.

With 9 series Intel chipset with native support for use of 2-4 lanes of PCIe lanes dedicated for bootable PCIe SSDs, its now up to the SSD OEMs to push out new drives with the NVMe standard. As of today, Intel is out with one consumer SSD, Samsung is soon out with SM951 and several other have announced they are working on drives.

Make note. This is the future starting this year.

 

[Ketchup]

As as these driver progress, are we going to have to start worrying about them taking bandwidth from other technologies (say SLI/ Crossfire)? Or are we not even in the same realm here?