Any Benefits From Putting 2 Samsung 830s in RAID0?

[thm1223]

I have a 256GB Samsung 830 currently installed, and just bought another one. I am unsure if it is worth putting it in RAID0. Although I am not worried about losing data in case of a failure (I habitually back up everything and host important documents/apps from my network attached storage device), I am curious if there are any significant performance gains, particularly given my mobo (Asrock Z77 Extreme4).

Keep in mind that these SSDs are primarily used to store games, so if there is a reasonable performance gain I'd probably be willing to deal with the increased risk of a failed array.

Any thoughts my fellow geniuses!?

 

[Zap]

Yes, if your Steam collection won't fit on one 256GB volume, then you benefit from having a single 512GB volume.

 

[thm1223]

Yes, if your Steam collection won't fit on one 256GB volume, then you benefit from having a single 512GB volume.

Cool cool, although I guess I'm curious what kind of performance increases I can expect. Also, as a bit of a hardware nub, will I lose all data when I add the 2nd hard drive and create the array?

 

[tweakboy]

Is your motherboard SATA 2.0 or SATA 3.0 ? If its 2.0 then RAID will make it 500mbps. If its sata 3.0 then it will be 1020mbps

Yes you can RAID SSD fine with your mobo. 520mbps will be
1020mbps read. That is the advantage. Your not gonna get a faster system because its RAID0 SSD , its only good for bulk transfer performance. As for SSD performance raid or non raid the speed is same as far as booting, launching apps reading files fast saving fast,,, instant launch of every app...... Problem is its only 256GB your gonna be limited and soon run out of space and what not. You go single you get same speed as RAID , except you read will be 500mbps instead of double that,, that is all you will gain out of doing raid with SSD . With hard drives RAID,, they become snappier and better access times.

Well people dont essentially use SSD from 2008 to now not for games but for OS , boot drive, apps and data. Why didn't you grab a 512GB instead. Should be about 550 to 600 now. over 300 more then Crucial 512GB for no noticeable diff, plus Im sata 2.0 ,,,,, Im glad with 400 bones for my M4 and I endorse 010G firmware. Boot time decreased by 5 to 10 seconds. PS CS6 launches in 2 seconds on first launch. everything else is instant. :whiste:

 

[Coup27]

I ran 2 x 128GB 830's in RAID0 for a month and although it looked good in benchmarks, there was no overall appreciable difference in system speed. Overall boot time actually went up because the RAID has to load when in AHCI it doesn't.

I went back to a single drive and use the other elsewhere.

 

[Insert_Nickname]

Unless you benchmark, no...

 

[sm625]

90% of enthusiasts will tell you that there is no appreciable difference between

A) A smallish SSD OS drive with a large HDD for games folder
B) A giant SSD that includes all game folders

So I wouldnt expect anyone to tell you that RAID is a logical solution here.

 

[kleinkinstein]

All you naysayers make me laugh and question your credibility. It makes a huge difference! Here is a single 128 Vertex 4

And here are two in Raid0. Insane speeds!!

I need sata IV!

No better time than right this instant to snag another 256 for $155 @ Amazon
http://forums.anandtech.com/showthread.php?t=2277393

 

[GlacierFreeze]

All you naysayers make me laugh and question your credibility. It makes a huge difference! Here is a single 128 Vertex 4

Awesome synthetic benchmarks. You missed the point of the "naysayers". Of course it puts up nice and pretty numbers in nearly useless benchmarks, but does the perceivable gain in real world usage in the OP's specific scenario (games) justify the money spent on them? Not really.

 

[hhhd1]

You might loose some performance by putting them in raid0 due to:

1. slower boot times due to more drivers need to load.
2. slightly slower access time due to the added layer of raid.
3. the array could get slower over time IF your raid doesn't pass TRIM to the SSDs.

 

[Zxian]

1. slower boot times due to more drivers need to load.

Only a concern at boot time, and we're talking about an additional 2 seconds.

2. slightly slower access time due to the added layer of raid.

Slightly slower random access times, sure, but that's only a concern if you're looking at a single access. Sequential access and random access with high QDs are going to improve overall.

3. the array could get slower over time IF your raid doesn't pass TRIM to the SSDs.

Only a concern if you're doing many writes and the drive has little-to-no internal garbage collection. In every day use, you won't see any performance degredation from the lack of TRIM.

I've got two SSDs in my system in RAID0. Can I notice a difference in terms of overall system performance compared to a single drive? Not really. You gain the simplicity of having a single volume and some e-peen status when it comes to your benchmarks.

 

[thm1223]

Sounds like the primary advantage of RAID0 then in my case is the fact that I will have a single volume. Boot up/shutdown times I could care less about because the computer is always on.

Now, one of the specific questions (I think) I asked revolved around the actual installation of the 2nd drive. Do I just plug it into one of the available SATA ports and that's it? Or do I need to format it/configure the array through the BIOS? AND MOST IMPORTANTLY, will the drives get wiped during this process?

 

[WilliamM2]

You need enable RAID in the bios, and then configure it in the RAID configuration utility (press ctrl+I on boot with Intel raid). Yes, the data will be wiped. No different than RAID 0 with standard platter drives..

 

[jhansman]

Awesome synthetic benchmarks. You missed the point of the "naysayers". Of course it puts up nice and pretty numbers in nearly useless benchmarks, but does the perceivable gain in real world usage in the OP's specific scenario (games) justify the money spent on them? Not really.

Or, let's see that RAID configuration load, say, 64-bit Photoshop with less than 16GB of RAM. I'll wager you'll see little or no improvement. Benchmarks, schmenchmarks...

 

[Synthet1X]

I am looking into raiding my OS drive, but if there is no noticeable performance increase, what is the point?

 

[Insomniator]

I am looking into raiding my OS drive, but if there is no noticeable performance increase, what is the point?

Check your user name.


Its useful in specific rare situations... but you have half the reliability.

 

[GlacierFreeze]

Or, let's see that RAID configuration load, say, 64-bit Photoshop with less than 16GB of RAM. I'll wager you'll see little or no improvement. Benchmarks, schmenchmarks...

Photoshop? No one is talking about Photoshop. The OP is talking about *games*, then some ...guy... tried counterpointing it with benchmarks that have no bearing on *his* scenario. So yeah, benchmarks, schmenchmarks. The conclusion is that there isn't much advantage/point in doing RAID SSD for games, especially if features such as TRIM are lost. No one said RAID SSD was completely pointless. Of course Photoshop and other disk heavy programs will love RAID SSD setups. But that's not what the thread is about.

 

[Coup27]

All you naysayers make me laugh and question your credibility. It makes a huge difference!

My single 830:

Two in RAID0

If you think you need SATA "4" then you clearly do not have an actual understanding of the numbers themselves. The most relevant number in those benchmarks which is responsible for overall system speed is the 4k read. As you can see from my RAID0, mine doesn't change at all (outside normal tolerances). Yours has gone up by about 16% which is actually the best return I have seen on RAID0. I have seen 3x m4's in RAID0 and the 4k read was the same as a single drive, much like my experience. The reason for this is as follows:

Relatively little change in 4k read results is exactly what you should expect. SSDs rely on parallelism to achieve their high performance: Either through servicing multiple requests at once (like in the 4k-64Thrd tests in AS SSD and all the ATTO tests), or by splitting up large IOs (like in the large tests in ATTO or the 512k tests in CDM), or by read ahead, where it notices you are accessing LBA sequentially and starts reading following LBAs ahead of the requests (as happens in every sequential test put there).

The 4k read tests fall in non of these categories. 4k is so small an IO that it isn't split up across multiple dies and doesn't get split over the array. It also doesn't have any queue depth. The next 4k request isn't issued until the previous one has been finished. So the SSD controller only gets one request at a time. And it is random, so you don't get any read ahead.

What this means is that there is nothing for the SSD controller to do but pass on the read request to the appropriate die. So the 4k test is basically a read latency test and a number of drives in Raid 0 doesn't make much difference.

Regds, JR

It doesn't matter how many you RAID, you will not get an overall appreciable speed notice from them. Due to the already high numbers a modern single SSD can do on sequential reading and writing, RAID0'ing more into the mix does not actually result in a real world speed difference and without benchmarks you would not even know it.

 

[kleinkinstein]

Without benchmarks all we have is hollow claims, self inflicted infocest, unfounded opinion, and half-wit perception...all which seem to fester here. I see the light, I'll take my 830 raid and go home...allowing you to play amongst yourselves.

 

[thm1223]

Without benchmarks all we have is hollow claims, self inflicted infocest, unfounded opinion, and half-wit perception...all which seem to fester here. I see the light, I'll take my 830 raid and go home...allowing you to play amongst yourselves.

Your avatar is appropriate given this comment.

 

[BFG10K]

Yours has gone up by about 16% which is actually the best return I have seen on RAID0.

Not just that, but look his sequential write and 4K write tests - 2.35x and 2.32x faster, respectively. That's significantly more than the theoretical maximum increase of 2x.

If I had to guess, I'd say something else has changed between those runs, not just adding another drive.

 

[Phynaz]

Without benchmarks all we have is hollow claims, self inflicted infocest, unfounded opinion, and half-wit perception...all which seem to fester here. I see the light, I'll take my 830 raid and go home...allowing you to play amongst yourselves.

Considering your posting history, that may be your best mode of action.

 

[blastingcap]

Considering your posting history, that may be your best mode of action.

Ok I have bashed Klein myself for saying outlandish things or giving wrong advice, but let's try to keep it constructive. His heart is in the right place, he just needs to do a little more research and be more careful in how he words things. Like in this case there needs to be a sanity check as you shouldn't be able to double up unless something else changed, as BFG noted.

 

[Old Hippie]

Yes I can tell the difference.

No it's not astounding but it is noticable.

As usual in horsepower, audio, video, and many other things it'll take 90% of the origional price to achieve the last 10% of performance.

Whether it's worth it to you is the ultimate test.

It's worth it to me and many others.

 

[Rakewell]

I'm in the same boat as the OP: I bought 2x256 830s I was planning on putting in RAID0 strictly for games. (I've got another SATA II 128 for the OS.)

From what I'm reading here,

Would it be best to just leave the 2x256 drives plugged into a SATA III non-raid PCI-E controller without raiding (leaving the drives as two volumes with the STEAM game mover app), keeping TRIM, or raiding them in SATA II and losing TRIM? (I've got Intel, I believe, in my EVGA 4-way SLI.)

I'm just concerned I'd lose TRIM just to have one volume.

Thanks-