RAIDing SSDs in a SATAII enviornment.

[darkspartenwarrior]

I know the general consensus on RAID SSD is, bad, too little gain for too much risk. But if you were limited to SATAII wouldn't putting 2 SSD in RAID0 theroretically give you double the bandwidth?

 

[DarkSable]

... Not even close.

The reason RAID 0 of hard drives work is because they don't come anywhere close to saturating the SATA controller.

A SSD, however, will easily provide more information than a SATA II controller can handle, and thus you'd see very little to no benefit unless you had VERY slow SSDs.

 

[Idonno]

... Not even close.

The reason RAID 0 of hard drives work is because they don't come anywhere close to saturating the SATA controller.

A SSD, however, will easily provide more information than a SATA II controller can handle, and thus you'd see very little to no benefit unless you had VERY slow SSDs.

I think you missed the point: "limited to SATA II"

You would indeed get close to twice the bandwidth with 2 SSD in RAID0 on the SATA II ports. To be clear that is twice what you would get on a single Sata II port not twice what the SSD is capable of in it's optimum environment like Sata III for instance.

A perfect example is my 2 Sata II SSD's in raid 0 on my Sata II ports here:

With one SSD the speeds are roughly cut in half.

 

[apache_lives]

just remember you loose TRIM, performance over time will decrease

 

[blazorthon]

You don't lose TRIM in RAID 0 if you have an Intel board with the right storage drivers installed.

 

[Deleted member 217926]

Only Z77 supports TRIM on SSD RAID I believe. At least that's the last I heard. If he has a Z77 board he's not limited to SATA II.

 

[blazorthon]

I thought that it was all LGA 1155 chipsets (some boards of which do lack SATA III) and IDK if AMD has any TRIM through RAID supporting drivers too.

Regardless, there are many SSDs such as current SandForce SSDs that handle operation without TRIM very well. So long as OP doesn't let capacity usage get out of control, they should last plenty long.

 

[Idonno]

just remember you loose TRIM, performance over time will decrease

This is an old wives tale that really needs to be dispelled. In 2010 when I built this PC and implemented this SSD raid 0 array everyone was saying what a terrible idea it was because of the loss of trim. Yet 3 years later without trim the entire time my SSD's aren't even starting to slow down.

Trim is a good thing and I would never suggest otherwise but, the need for trim is vastly overrated. The fact is you don't need trim if your SSD's firmware has good garbage collection which all SSD's with sandforce firmware have and even then if you buy a non-sandforce ssd with less than stellar garbage collection there are many utility's that will clean your free space aimed directly at keeping your SSD's running fast and smooth.

If you use an SSD without trim it helps to: Buy one with good garbage collection, leave a little more free space about 35% (even with a regular hard drive the suggested is 30%), If you don't have adequate garbage collection you should occasionally use an SSD free space cleaning utility.

Using one of these utility's is no more difficult and allot faster than defragging a regular hard drive to keep it running fast.
(Edit: I used the term "defragging" as a speed reference only, Don't Defragment an SSD!)

So with just a little (un)common sense using an SSD without TRIM is not an issue what so ever.

Regardless, there are many SSDs such as current SandForce SSDs that handle operation without TRIM very well. So long as OP doesn't let capacity usage get out of control, they should last plenty long.

+1

 

[darkspartenwarrior]

Well I've used the 2 SATAIII on my mobo and I have one more question.

Is it better to use the Marvell controlled addon SATAIII(individual JBOD) or RAID the 2 SSD in native SATAII

SSD=520 Cherryville series.

 

[blazorthon]

nvm

 

[Idonno]

Is it better to use the Marvell controlled addon SATAIII(individual JBOD)

That depends on what you mean by "Marvell controlled addon SATAIII" If you mean the ones on your motherboard, Then yes that would be better but, If you mean on an add-on-card then that would depend on the card.

The reasons I use Raid 0 on the Intel Sata II ports on my motherboard instead of one good Sata III SSD the Marvell Sata III ports are two-fold.

#1 at the time of this (My) build there were no Sata III SSD's with good write speeds.

#2 With the exception of the Rampage III Extreme Black Edition and the GIGABYTE G1.Assassin that both came out at a later date, all x58 (LGA 1366) motherboards used the Marvell 88SE9128 SATA III Controller which was (IMO) pure trash. It uses only one PCIe lane shared with both Sata III ports theoretically giving the combined Sata III ports a total possible throughput of 5G. While this by itself doesn't sound too bad if you only were to hook up one SSD, the reality was much worse.

This crappy Marvell 88SE9128 SATA III Controller was also used in many (if not most) LGA 1155 motherboards but, as far as I know All x79 (lga 2011) use a much better Marvell SATA III Controller that uses two PCIe lanes. While I wasn't able to find any info on what Sata III controller your motherboard uses, it's clear from reviews like this one: http://www.bit-tech.net/hardware/motherboards/2012/01/06/intel-dx79si-review/7 that states:

At stock speeds, the DX79SI proved just as capable as the other LGA2011 boards we've tested. Its score of 1,735 in the image editing test was top of the class, as was its SATA 6Gbps read speed of 553MB/sec. It also managed a SATA 6Gbps write speed of 521MB/sec, essentially maxing out our OCZ Vertex 3 SSD.

that your motherboard uses a much better SATA III controller.

So one Sata III SSD on your motherboards Sata III ports would be about the same as 2-SSD's in raid 0 on the Sata II ports and if you can get the same speed without using raid that is always the better option.

If you're talking about a Marvell SATA III Add-on-card, all the ones that are PCIe x1 only use 1 PCIe lane and they suck. You would need one that uses at least 2 lanes like one of these: http://www.newegg.com/Product/Productcompare.aspx?Submit=ENE&N=100007531&IsNodeId=1&page=5&bop=And&ActiveSearchResult=True&CompareItemList=73|16-115-114^16-115-114-TS%2C16-115-118^16-115-118-TS%2C16-115-077^16-115-077-TS%2C16-115-130^16-115-130-TS Right now I'm also trying to find out how good these are for my own use. The reviews are very mixed, however it is my feeling that they should be OK for one good Sata III SSD.

 

[darkspartenwarrior]

Oh ok so an external controller on the motherboard(not card) is still faster than 2 SSD in SATAII.

 

[Idonno]

Oh ok so an external controller on the motherboard(not card) is still faster than 2 SSD in SATAII.

That of course depends on the Sata III controller that the motherboard has, But on your motherboard with the right SATA III (not Sata II) SSD you should be able to double the speed of the SSD by having it on the Sata III ports.

 

[blazorthon]

I think that it'd be a little better to go dual SATA3Gb/s. SATA 6Gb/s rarely truly doubles performance over SATA 3Gb/s because the drives usually can't sustain quite that much performance whereas two decent SSDs can easily saturate two SATA 3Gb/s ports in normal usage.

 

[Idonno]

Where the Sata III used to have the biggest issue is that it only used one pcie lane theoretically limiting 2-6G-Sata III ports to a combined total of 5G with Marvell's less than stellar efficiency that translated into combined total of about 400MB/s

The new controllers on the x79 boards use two PCIe lanes for a combined total of 10G with Marvell's less than stellar efficiency this is still no where near enough to run both Sata III ports @ full 6G potential but, since it's a shared 10G and not a distributed evenly 10G it's plenty to run 1-Sata III port at (or close to) it's full potential.

The newer Marvell Sata III ports are still less than perfect and are known to have issues in raid as well as the inability to achieve full (or close to full) 6G potential on more than one port at the same time but using only one port isn't an issue.

It's also important to note that I could find no documentation suggesting that this Intel DX79SI had Marvell sata III ports and not Intel sata III ports

But regardless in the review quote I posted the SATA III port got a read speed of 553MB/s and a write speed of 513MB/s essentially maxing out the OCZ Vertex 3

The OP's Sata III Intel 180GB 520 Series SSD has a 550MB/s Read and 520MB/s write so he should have no problem getting close to those speeds.

These same SSD's on the Sata II ports only get 280MB/s Read and 260MB/s write so, to put two of these on the Sata II ports in raid 0 would amount to nothing more than a big waste of money for a (theoretical) improvement of 10MB/s read.

 

[blazorthon]

I never said that it would be a big difference. My point was just that going by a purely what is best stance, it would be best. That's why I said a little better isntead of a lot better or anything like that.

Also, it's been my experience (especially with SandForce) that maximum theoretical performance is rarely achieved except in rare perfect scenarios and some benchmarks. Tom's real-world tests show that they'll almost never really hit up performance with a single drive, especially once it has settled in, unless you compare to to a hard drive where the difference is obviously huge.

 

[Idonno]

Don't take offense, I know you know your stuff. My post wasn't a critique of yours. I was just trying to make sure the OP understood all the facts, what was best for him and why.

 

[pyite]

I know the general consensus on RAID SSD is, bad, too little gain for too much risk

Not true, unless you are using MLC drives. I stick to SLC, which doesn't require TRIM and has better write performance. They can get expensive though.

 

[blazorthon]

I don't know of many SLC SSDs being up to date. Most of them have been surpasses by MLC through sheer modernization overcoming the cost issue of SLC. Even the endurance advantage of SLC has been mostly overcome through eMLC.