SATA3 vs SATA2 RAID 0

[thelostswede]

You're somewhat mistaken, as you bond the channels together inside the controller and you can easily exceed the speed of a single port this way.
Here's an example of some SSD's in RAID for example http://www.bjorn3d.com/2012/04/2x-kingston-hyperx-240gb-3k-ssds-in-raid-0-configuration/
That's hitting over 1GB/s which is much faster than what a single SATA 6Gbps port can deliver.
You'd see similar benefit using hard drive, albeit it won't be quite as obvious, since mechanical drives are that much slower.
Admittedly it's all very dependent on the drives you're using etc. but yes, you can easily exceed the speed of a single SATA port.

 

[ggeorge]

You're somewhat mistaken, as you bond the channels together inside the controller and you can easily exceed the speed of a single port this way.
Here's an example of some SSD's in RAID for example http://www.bjorn3d.com/2012/04/2x-kingston-hyperx-240gb-3k-ssds-in-raid-0-configuration/
That's hitting over 1GB/s which is much faster than what a single SATA 6Gbps port can deliver.

You'd see similar benefit using hard drive, albeit it won't be quite as obvious, since mechanical drives are that much slower.
Admittedly it's all very dependent on the drives you're using etc. but yes, you can easily exceed the speed of a single SATA port.

Using multiple channels to simultaneously access multiple HDD is the point of RAID.

It's looking like either I'm misunderstanding your point or you seem to have RAID confused with SATA III. Yes RAIDing two SSDs on two separate SATA III channels will let you exceed the through put of a single SATA III port.

So will RAIDing two mechanical HDDs on two SATA II ports let you exceed the performance of a single SATA II port, but SATA III ports won't improve things any though. Since the SATA II ports aren't the limitation to begin with.

RAIDing the HDD has nothing to do with the kind of port they are on. There were IDE raid cards at one point.

 

[thelostswede]

Yes, I'm aware of that, I used to use IDE RAID controllers back in the days...

Anyhow, I think we're misunderstanding each other here somewhere.
Sadly I don't have a really fast drive to do a test with (as they're in my system), so I had to make do with a WD Blue 500GB drive which doesn't really point out the benefits of SATA 6Gbps over SATA 3Gbps to any major extent, but the first screenshot is of it attached to a SATA 3Gbps port and the second on a SATA 6Gbps port.
With such a slow drive, you only gain about 3MB/s on average for read operations and about 1MB/s for write operations, but interestingly you're also seeing over 2MB/s random writes, which is something hard drives such at and for some this will at least be a tangible benefit.
Using a faster drive, you'd see an increased benefit over this move, but yes, it's not the same as using an SSD and I never suggested this.

 

[tot31]

What ggeorge said is true, RAIDing 2 HDDs using different channels on any SATA either SATA 2 or SATA 3 won't have much impact as the bottleneck is on the HDDs themselves, unless you are using 3-6 HDD then you can see the performance on par or achieving the same speed of a single SSD.

 

[ggeorge]

Yes, I'm aware of that, I used to use IDE RAID controllers back in the days...

Anyhow, I think we're misunderstanding each other here somewhere.
Sadly I don't have a really fast drive to do a test with (as they're in my system), so I had to make do with a WD Blue 500GB drive which doesn't really point out the benefits of SATA 6Gbps over SATA 3Gbps to any major extent, but the first screenshot is of it attached to a SATA 3Gbps port and the second on a SATA 6Gbps port.
With such a slow drive, you only gain about 3MB/s on average for read operations and about 1MB/s for write operations, but interestingly you're also seeing over 2MB/s random writes, which is something hard drives such at and for some this will at least be a tangible benefit.
Using a faster drive, you'd see an increased benefit over this move, but yes, it's not the same as using an SSD and I never suggested this.

I guess I would simply attribute those differences to the vagaries of disk testing, more than anything else and wouldn't consider them a significant enough difference to worry about. After all there's all sorts of subtle issues that can arise when trying to benchmark HDDs. You get caching effects, fragmentation, inner track vs outer track... the list goes on and on.

I doubt it's going to hurt things to put a HDD on a SATA III port vs a SATA II, but I wouldn't go out of my way to do it.

 

[thelostswede]

I guess I would simply attribute those differences to the vagaries of disk testing, more than anything else and wouldn't consider them a significant enough difference to worry about. After all there's all sorts of subtle issues that can arise when trying to benchmark HDDs. You get caching effects, fragmentation, inner track vs outer track... the list goes on and on.

I doubt it's going to hurt things to put a HDD on a SATA III port vs a SATA II, but I wouldn't go out of my way to do it.

3MB/s isn't not a variance in testing and if you look closely at the random write tests, you'll see that in some of the tests there's over 10MB/s difference, but clearly this is also just a variance?

But whatever, if you want to believe whatever ever you want to believe go ahead. There's a measurable difference, but as I said, it's not huge, but for certain programs it'll make a difference.

Can I ask why Intel is bothering with more SATA 6Gbps ports on next years platform, if it is as you say pointless? As I can't see a lot of people slapping four SSD's in their system, even with the current prices...

 

[ggeorge]

3MB/s isn't not a variance in testing and if you look closely at the random write tests, you'll see that in some of the tests there's over 10MB/s difference, but clearly this is also just a variance?

Yes, actually I would expect to see differences much greater than a couple of Mb/s from variations in testing with something like this. Every drive has it's quirks as to how it behaves at different block sizes, que depths, file sizes, sequential vs random, due to the firmware, head design and architectural issues.

It is one of the reasons good testing involves multiple run so that such variations will average out. Doing good testing is time consuming, tedious and requires a lot of attention to detail, but without it you're really only able to make rough comparisons. A 10 Mbps difference on 130 Mbps total throughput is simply not that significant, it's roughly 7% greater. 2-3MB/s is a rounding error in most cases.

If you want to look at cherry picked results, on the 64kb sequential writes, the 3 Gbps throughput was 12mb/s, greater than the 6Gbps throughput. (139 Mb/s vs 127 Mb/s)

So does that now mean that 6Gbps is worse?

The lesson I take away from your two tables is that the vast majority of the results are very close between the two.

But whatever, if you want to believe whatever ever you want to believe go ahead. There's a measurable difference, but as I said, it's not huge, but for certain programs it'll make a difference.

If you need a stop watch to detect the difference, it's not worth worrying about.

Can I ask why Intel is bothering with more SATA 6Gbps ports on next years platform, if it is as you say pointless? As I can't see a lot of people slapping four SSD's in their system, even with the current prices...

1) I don't think it's pointless putting more 6Gbps ports on a motherboard, with the way prices are dropping on SSDs. 120 Gig SSDs are getting close to $50 a pop on sale. Also it's an expected feature and really it isn't costing them anything to increase the number of ports.

2) They're doing it for the same reason companies put more ram on graphics cards. People often don't know the difference between a GK104 chip and a GK107, but do understand that 4 > 2.

I know if I was looking at two motherboards that were identical other than the # of SATA III ports, I would pick the one with more SATA III ports, simply because it might be handy, even if I was never going to put 10 drives in my computer.

 

[tot31]

Yes, actually I would expect to see differences much greater than a couple of Mb/s from variations in testing with something like this. Every drive has it's quirks as to how it behaves at different block sizes, que depths, file sizes, sequential vs random, due to the firmware, head design and architectural issues.

It is one of the reasons good testing involves multiple run so that such variations will average out. Doing good testing is time consuming, tedious and requires a lot of attention to detail, but without it you're really only able to make rough comparisons. A 10 Mbps difference on 130 Mbps total throughput is simply not that significant, it's roughly 7% greater. 2-3MB/s is a rounding error in most cases.

If you want to look at cherry picked results, on the 64kb sequential writes, the 3 Gbps throughput was 12mb/s, greater than the 6Gbps throughput. (139 Mb/s vs 127 Mb/s)

So does that now mean that 6Gbps is worse?

The lesson I take away from your two tables is that the vast majority of the results are very close between the two.



If you need a stop watch to detect the difference, it's not worth worrying about.



1) I don't think it's pointless putting more 6Gbps ports on a motherboard, with the way prices are dropping on SSDs. 120 Gig SSDs are getting close to $50 a pop on sale. Also it's an expected feature and really it isn't costing them anything to increase the number of ports.

2) They're doing it for the same reason companies put more ram on graphics cards. People often don't know the difference between a GK104 chip and a GK107, but do understand that 4 > 2.

I know if I was looking at two motherboards that were identical other than the # of SATA III ports, I would pick the one with more SATA III ports, simply because it might be handy, even if I was never going to put 10 drives in my computer.

Very well pointed out. I'm totally agree with you.

 

[totenkopf4]

Yes, I'm aware of that, I used to use IDE RAID controllers back in the days...

\]

thelostswede

You made me chuckle when I read the line above. ...

I wonder if any one remembers some of the other lost Drive technologies ....

like SCSI which was an Apple standard a many years back.
This includes the versions of SCSi like - Wide SCSI, Narrow SCSI, Fast and Wide SCSI, SCSi 2 and 3.

Or ESDI the red headed step child / grandfather of IDE ... and yes even before IDE we had an earlier version of IDE w control and data cables separate along w power like MFM.

Or the original MFM drive writing technologies when Hard Drives were only 5 MB and were 8 inch form factor. aka "early Western Digital - Winchester drives"

Anyone remember in a Galaxy far far away.... the IBM 14 inch Removable Multilevel Disk Packs....came in 75 or 288 MB - and the drive was the size of a washing machine whoopee....

But back to the question - SSD will out perform any Mechanical Hard drive weather its on SATA 2 or SATA 3 - due to the data bandwidth inside the box itself
choke points are the issues that any hardware or electrical designer has to deal with as the data moves a different speeds from place of origin to where it needed --- the only time that Mechanical HD ever getting close would to SSD speed be w custom SAS or Fiber Channel hard drive's and controllers w 15K hard drives like the main frame and mini mainframe / servers such as HP / IBM use over a 10 GB Ethernet or 16GB Fiber channel links using raid 10 /50. But we wont see that in anyone's hackintosh soon...... wink wink!

If you have drives in RAID 0 on what every interface you want the data is written to both drives and when needed it is read from both drives - so this is the best example of improving the data channel bandwith - the theory is that from a single drive to two drives in RAID 0 you are able to read the data from the disk to the CPU at a 40 % improvement vs the single drive - now if this is done w TWO SSD's then shazam its as fast as greased lightning.... and your too rich! Now this improvement is not universal for all data - your packet size is also important the sector size - 4k 8k 16k etc. This also has an effect on how fast the data is moved from HD to CPU.

But what we are also missing in this Discussion is that Hybrid Drives which have both SSD and fast mechanical drive assemblies can be used in RAID to get to ALMOST the levels of an single SSD only drive and would most likely be a good alternative to SSD only for RAID - but as the Cache Ram which is the SSD side of a Hybrid is rather small IMHO and will be increased to improve performance once acceptance is established is what is needed - so if you are looking at Hybrid Drives get the biggest or largest amount of ram (cache) that you can for this technology. Where I see this as a player is in laptops that have the need for speed and capacity - as having two of these Hybrid drives in Raid 0 would be assume if your laptop has two drive bays.

What is going to be interesting is when we really get what we all want is LARGE CAPACITY SSD- say 1-2 TB or more. As the current design specs for either memory types used in today's SSD's will have issues w performance due to the large array of ram that the drive contains. The snappy speed we see today in the current SSD's will be more like the fastest SATA mechanical Hard Drive's of today.

And yes I do think that SATA will also go the way of the DODO bird and we will be using something else to increase data transfer cause its all about getting to the data in today's information ( its all about me) hungry world.

Its all about the bandwidth Baby...

Yes I am older than dirt..... Peace Out..

 

[ggeorge]

thelostswede

You made me chuckle when I read the line above. ...

I wonder if any one remembers some of the other lost Drive technologies ....

like SCSI which was an Apple standard a many years back.
Thin includes the versions of Wide SCSI, Narrow SCSI, SCSi 2 and 3.

Ah yes, at my old job we had an external SCSI raid box from dell that every couple of months would suddenly "forget" the raid configuration and go "Drives what drives?". It was simple to fix when you knew what to do, but it lead to some "interesting" times.

Or ESDI the red headed step child of IDE ... and yes even before IDE we had an earlier version of IDE w control and data cables separate along w power like MFM.

Ya got me on this one. Managed to miss this somehow.

Or the original MFM drive writing technologies when Hard Drives were only 5 MB and were 8 inch form factor. aka "early Western Digital - Winchester drives"

These are what they used with the Lisa and Apple III weren't they?

Anyone remember the IBM 14 inch Removable Multilevel Disk Packs....came in 75 or 288 MB - and the drive was the size of a washing machine whoopee....

I've seen pictures, but never had to tangle with one IRL.

But back to the question - SSD will out perform any Mechanical Hard drive weather its on SATA 2 or SATA 3 - due to the data bandwidth inside the box itself
choke points are the issues that any hardware or electrical designer has to deal with as the data moves a different speeds from place of origin to where it needed --- the only time that Mechanical HD ever getting close would to SSD speed be w custom SAS or Fiber Channel hard drive's and controllers w 15K hard drives like the main frame and mini mainframe / servers such as HP / IBM use over a 10 GB Ethernet or 16GB Fiber channel links using raid 10 /50. But we wont see that in anyone's hackintosh soon...... wink wink!

What is going to be interesting is when we really get what we all want is LARGE CAPACITY SSD- say 1-2 TB or more. As the current design specs for either memory types used in today's SSD's will have issues w performance due to the large array of ram that the drive contains. The snappy speed we see today in the current SSD's will be more like the fastest SATA mechanical Hard Drive's of today.

The next step looks like it is going to be some sort of PCIe connection and they've just announced a new standard that is essentially that. The current PCIe SSDs look like they are just going to be an interim step, since they are largely just SATA drives on a board with a bridge chip.

And yes I do think that SATA will also go the way of the DODO bird and we will be using something else to increase data transfer cause its all about getting to the data in today's information ( its all about me) hungry world.

Its all about the bandwidth Baby...

Yes I am older than dirt..... Peace Out..

I'm not quite as old, not a spring chicken either though.