RAID 0 worth it or not

[tweeve2002]

What is faster RAID 0 or single drive? Im not talking just about the high end 76GB Raptor, what about the every day 7200 RPM drives.

In Junes Issue of Maximum PC they have some speed test of with and without RAID, what RAID stripe size is the fastest. Now Im saying that the scores arnt perfect or reflect real world performance. I just know that RAID 0 is faster if it is set up correct.

I've been running RAID 0 for close to 4 years now. I have loaded into games faster than computers that are faster than mine and have a single drive. I have also seen computers slower than mine with their RAID set up just about perfect and finish loading levels faster than me and my RAID 0 set up. I have never seen a single drive beat me in loading a level.

For most people I think that a 16k or a 32k stripe is the fastest ( info out of Maximum PC) I was running a 1MB stripe on my Highpoint Rocket 133 RAID card and a 64k strip on my onboard Via RAID with my P4P800 Deluxe. When I switched both RAID over to a 16k stripe I did notice a speed boost. On the via RAID games load faster and Windows did Install in a about 30 min compaired to 45 - 60 min before. I highpoint RAID is just use as a RAID 0 backup drive, since I was able to pick up two WD 80GB drives drit cheap.

I belave RAID 0 is faster than a single drive, but you need to have is set up correct and do some research as to have chip set is faster, if you are going to use an add-in card what one is faster, and what stripe size is fastest for you.

 

[dullard]

Copy and paste from the other thread:

Lets look at Anandtech's latest article in more detail:

Benchmark / RAID 0 36 GB Raptors / single 36 GB Raptor / Speed boost using RAID 0
UT2004 level loading / 29.4 / 34.1 / 13.8% faster with RAID 0
Far Cry level loading / 40.6 / 41.1 / 1.2% faster with RAID 0
Sysmark overall / 211 / 204 / 3.4% faster with RAID 0
Sysmark overall internet / 235 / 230 / 2.2% faster with RAID 0
Sysmark overall office / 189 / 180 / 5.0% faster with RAID 0
Sysmark communciation office / 196 / 171 / 14.7% faster with RAID 0

I got tired typing, but you can see the pattern. RAID 0 with two 36 GB drives is faster in every case than the single 36 GB drive. You'll get similar results with any hard drive - benchmarks are abundantly available to show that. But it is only slightly faster (see Far Cry results above). Also you have to pay more to buy two drives and you can lose data more easilly.

Whether it is worth it or not is up to you. I personally don't think hard drive speed is very important and thus I won't buy two drives for a 0%-15% boost in programs.

 

[Markfw]

I don;t have any numbers to back it up, but in my only non-enterprise raid0 array, I swear it loads faster also. When my new system gets built (1 week or so ??) I can provide real enterprise level raid0 benchmarks. Right now, the SCSI card is choked, so my benches stink....

 

[EeyoreX]

I thought we had it straight already. Sometimes it is worth it, sometimes it is not. And it largely depends on the person. I personally saw no benefit when I had a RAID0 array. Was the RAID0 faster? Undoubtedly yes. Was it fast enough for me to notice in real world usage? Absolutetly not. Why should I bother spending the extra money? Maybe becasue I don't came and therefore don't have any levels to load, but even if I did, I doubt the added expense would be worth it to me. If my experience is skewed that much by a few seconds, I need help.

\Dan

 

[Optimummind]

I guess if you're a gamer, then RAID0 is not needed. I own some games that take a pretty long time to load like Far Cry and Deus Ex 2. Shaving 2 or 3 seconds will not be worth the cost of the upgrade for me since it's cost me money to buy 2 HDD's plus a RAID controller card and the reliability of data gets halved.

More importantly, it won't increase my in-game FPS so why bother??

 

[LordPhoenix]

good point. i agree, optim

 

[DJFuji]

i wanna see RAID1 benchmarks. Because that's something i'd actually USE, regardless of the speed increase.

 

[alexgontijo]

I agree with Tweeve2002 when he says: a RAID 0 setup will ALWAYS be faster than 1 HD only.

I don't believe in those Anandtech benchmarks, mainly in game loading time. I am on the 2nd pair of HDDs using RAID from 6 months, 2 x SATA Maxtor 160 GB, and they are definitively blazing fast! With one drive only, I set up Windows + Service Pack in around 17 minutes. With the RAID 0, I made the same installation in just 10 minutes! This is real world, not just an benchmark.

But speaking in benchmark, using SANDRA 2004, with one drive the performance is 52 mbps, and with RAID 78 mbps!

Loading games in the same time?! Are they kidding?!! My Far Cry loads at least 5 seconds faster, Battlefield 1942 and Desert Combat I'm always the first to load the maps in the servers... and I don't believe in Santa Clara...

You can?t go wrong with a 50%+ performance boost, and it?s not as insecure as warned. I had a first couple of Western Digital 120 GB SE, and now the Maxtors SATA 160 GB, running into a MSI 865PE Neo2 with the Promise 378 RAID controller.

Guys who use a RAID 0 setup knows: it's one of the best?s upgrades that you can do.

Regards,

AleX

 

[GoSharks]

Originally posted by: alexgontijo
I don't believe in those Anandtech benchmarks, mainly in game loading time. I am on the 2nd pair of HDDs using RAID from 6 months, 2 x SATA Maxtor 160 GB, and they are definitively blazing fast! With one drive only, I set up Windows + Service Pack in around 17 minutes. With the RAID 0, I made the same installation in just 10 minutes! This is real world, not just an benchmark.

But speaking in benchmark, using SANDRA 2004, with one drive the performance is 52 mbps, and with RAID 78 mbps!

Loading games in the same time?! Are they kidding?!! My Far Cry loads at least 5 seconds faster, Battlefield 1942 and Desert Combat I'm always the first to load the maps in the servers... and I don't believe in Santa Clara...

You can?t go wrong with a 50%+ performance boost, and it?s not as insecure as warned. I had a first couple of Western Digital 120 GB SE, and now the Maxtors SATA 160 GB, running into a MSI 865PE Neo2 with the Promise 378 RAID controller.

1. the anandtech results support storage review's results for a while back. storage review is the #1 site for storage related things, ask anybody in the know. http://www.storagereview.com/
1a. how many times do you install windows per day?

2. sandra is one of the worse benchmarks for hard drives. sustained transfer rates do not mean very much.

3. so you load a few seconds faster - its still taking you probaly 30-50seconds to load up? (far cry) 5 seconds isnt a very large difference.

4. 50% performance boost? maybe in the sustained data rate benchmarks i'll give you that. but we all know they mean next to nothing.

 

[FlameDeer]

Agree with Optimummind.

Originally posted by: DJFuji
i wanna see RAID1 benchmarks. Because that's something i'd actually USE, regardless of the speed increase.

RAID 1 practically is only a little bit slower than single HDD solution. But the main thing is RAID 1 help to keep your important data safe and secure. Data lost only occur if that is so unfortunate both drives in RAID 1 array die at the same time.

 

[Matthias99]

Originally posted by: tweeve2002
For most people I think that a 16k or a 32k stripe is the fastest ( info out of Maximum PC) I was running a 1MB stripe on my Highpoint Rocket 133 RAID card and a 64k strip on my onboard Via RAID with my P4P800 Deluxe. When I switched both RAID over to a 16k stripe I did notice a speed boost. On the via RAID games load faster and Windows did Install in a about 30 min compaired to 45 - 60 min before. I highpoint RAID is just use as a RAID 0 backup drive, since I was able to pick up two WD 80GB drives drit cheap.

A 1MB stripe size? Jeez, that would be really slow for anything except working with HUGE files (100MB+) that are being written or read sequentially (for instance, if you loaded up some super-high-res Photoshop file, or were editing large high-res videos). Who the heck told you to do that? Or did you just set it to the biggest size it supported? The problem with a stripe size that big is that almost all 'normal' file accesses are going to be less than 1MB -- so you're not getting the benefit of striping your data, since only one disk will handle each transfer.

I definitely recommend no more than a 64kB stripe size for 'normal' desktop computing. Smaller is probably *better* for loading games -- while game levels may be *big* (100MB+), they are generally assembled from lots of small textures, models, etc., rather than being loaded in one fell swoop from a single big file.\ (although this varies from game to game; often all the level geometry and lighting data -- which is not THAT big -- is loaded from one file, and then the character models and textures are loaded individually).

 

[Cerb]

Dullard, the comparison should be from 2x36GB to 1x73GB.

 

[Tostada]

AnandTech and StorageReview both seem to agree. Yes, RAID 0 is a little bit faster sometimes, but not really enough faster to make it worth it.

 

[alexgontijo]

Originally posted by: GOSHARKS

1. the anandtech results support storage review's results for a while back. storage review is the #1 site for storage related things, ask anybody in the know. http://www.storagereview.com/
1a. how many times do you install windows per day?

2. sandra is one of the worse benchmarks for hard drives. sustained transfer rates do not mean very much.

3. so you load a few seconds faster - its still taking you probaly 30-50seconds to load up? (far cry) 5 seconds isnt a very large difference.

4. 50% performance boost? maybe in the sustained data rate benchmarks i'll give you that. but we all know they mean next to nothing.

1 - I know the sites and I believe they are very serious, be the point I was talking is that you can't just see benchmarks, mainly synthetic ones, to judge the real performance. "You have to see it for yourself" by Morpheus.

1a - Hehehehe. Not quite often, just once in each 6 months I guess.

2 - I know SANDRA isn't much precise, the results aren't all the same, but was just to illustrate. Running other benchs like HD Tach and AIDA32 I got the same levels of boost, and even higher.

3 - Far Cry loads like in 25 seconds... Should be certainly more than 30 with a single drive.

4 - Installing big programs, like Far Cry from a Virtual device (DVD image from 3.5 GB) you will see that a RAID 0 is like a Ferrari against a Honda Civic... Personally I find out at least 25% faster installations.

But its up to you... I think the performance pays the price! No boring waits for slow and big loadings!

Regards!

AleX

 

[TerryMathews]

OK, I'm going to start this post off by stating that I use RAID 0 (4x WD400BB on a FastTrak TX4000) for my OS and apps and RAID 5 (3x WD2000JB on a SuperTrak SX6000 w/ 128MB RAM) for things I don't want to lose - downloads, documents, backups, etc.

Now for the dispensing of the knowledge.

16kb is the ideal general purpose stripe size. If you don't know enough to determine the optimal stripe size for your application, put in 16kb.

The only purpose of RAID 0 (and to a lesser extent RAID 5) is to increase the sustained read and write performance of your storage. If you don't need to sustain a transfer rate of 80MB/sec, you're probably going to feel that RAID is worthless (for you) and rightly so - it's overkill for a lot of applications.

I've got the RAID stuff from some previous dealings, but I don't really need it. I'm just too lazy to get rid of it. *Most people don't need to push 80MB/sec around at will*

RAID0 is not really any more risky that using a single drive *IMO*. The math tells us that RAID0 lowers your total MTBF since any one drive failure would destroy the whole array. That being said, it's just as likely that the one drive in the array that died would be the one you would use as a stand-alone drive so it's a wash. The only thing that's truly more secure is not RAIDing the drives at all/using RAID 1. Then if one drive dies, you don't lose everything.

That's why I have my OS and apps on the RAID 0 and save all of my documents to the RAID 5.

Another thing: Any modern drives on a good controller can saturate a desktop PC's PCI bus in a 2 drive or 4 drive RAID 0. When I mentioned 80MB/sec earlier, there was a reason. It's not the upper limit of the array - it's the upper limit of my Abit IC7-G's PCI bus after overhead and other peripherals eating away at the bandwidth.

 

[Sunner]

dd

Originally posted by: alexgontijo

Originally posted by: GOSHARKS

1. the anandtech results support storage review's results for a while back. storage review is the #1 site for storage related things, ask anybody in the know. http://www.storagereview.com/
1a. how many times do you install windows per day?

2. sandra is one of the worse benchmarks for hard drives. sustained transfer rates do not mean very much.

3. so you load a few seconds faster - its still taking you probaly 30-50seconds to load up? (far cry) 5 seconds isnt a very large difference.

4. 50% performance boost? maybe in the sustained data rate benchmarks i'll give you that. but we all know they mean next to nothing.

1 - I know the sites and I believe they are very serious, be the point I was talking is that you can't just see benchmarks, mainly synthetic ones, to judge the real performance. "You have to see it for yourself" by Morpheus.

1a - Hehehehe. Not quite often, just once in each 6 months I guess.

2 - I know SANDRA isn't much precise, the results aren't all the same, but was just to illustrate. Running other benchs like HD Tach and AIDA32 I got the same levels of boost, and even higher.

3 - Far Cry loads like in 25 seconds... Should be certainly more than 30 with a single drive.

4 - Installing big programs, like Far Cry from a Virtual device (DVD image from 3.5 GB) you will see that a RAID 0 is like a Ferrari against a Honda Civic... Personally I find out at least 25% faster installations.

But its up to you... I think the performance pays the price! No boring waits for slow and big loadings!

Regards!

AleX

Explanation.

 

[ArmchairAthlete]

Definitely not worth it on my budget =/

 

[GoSharks]

Originally posted by: alexgontijo
4 - Installing big programs, like Far Cry from a Virtual device (DVD image from 3.5 GB) you will see that a RAID 0 is like a Ferrari against a Honda Civic... Personally I find out at least 25% faster installations.

installing from a totally separate drive (ie a seagate to a wd, names used to clarify) should be just as fast, or faster than installing from raid0 to the same raid0.

for the record i run a windows 2000 software stripe of two 9gb seagate cheetah 36xl, ultra 160. i did not see very much difference in everyday use compared to the single drive solution i was using before. only reason i still use it is becuase i have not needed to reformat for a very long time now it was nice when i was video editing though.

 

[zlooop]

A LOT MORE MONEY... NOT SO MUCH PERFORMANCE GAIN

Save the money for a faster processor...

 

[dguy6789]

I dont think it is worth it. Raid 0 may be slightly faster in loading screen, but it wont improve your fps no matter what, and you will have 2X the harddrives in the pc, meaning 2X the posibility of one breaking, and losing all the data on both harddrives. Not to mention more heat, and more noise.

 

[dullard]

Originally posted by: Cerb
Dullard, the comparison should be from 2x36GB to 1x73GB.

That is what the Anandtech article was about (1x73 GB Raptor was the better option than 2x36 GB Raptor). But that article's focus does not answer tweeve2002's question. So I pulled out the 1x36 GB Raptor vs 2x36 GB Raptor data to show exactly how RAID 0 helps. And RAID 0 helped in all tests.

 

[KF]

Since all the points have already been made, my comments will not add anything, but that won't stop me.

My latest mobo has onboard RAID (DFI NFII Lanparty.) I wanted to see for myself about this "blazing speed" that I hear touted often, but which shows up minimally in real-life benchmarks. So RAID was on my list of required features. What was it really? To sum up, it may be faster, but not enough so I can be sure. This is unlike speed increases in video, CPU and memory bus, and even fast HDs, where when you first see it, it does seem blazing. I believe that is because those relativerly short times leave a feeling in your brain which you get conditioned to. When the wait is always less than expected, you feel it as great speed. The feeling of extra speed goes away after a short aclimatization period. Then if you go to a slower system, it seems sluggish, but that goes away too. IAC, I didn't get it with the RAID 0 vs plain. Nothing blazed. So as a user-experience, RAID was a 0.

In price per MB, the extra cost of RAID 0 is not large, and may even be less. The thing is, most users probably don't need or use the MB on the small HDs appliance computers com with, so the extra drive would be extra cost.

If you run HDs until they die in use, or flog it day in, day out, then the MBTF number may be a cause of concern. Otherwise, it is too low to consider HD failure to be anything more than an "accident."

I tried RAID 1.5 (so-called), which could be as fast as RAID 0 for reads. But when I tried out breaking the RAID 1 to see what I would do in a failure, I inadvertantly did something wrong and wiped both drives unrecoverably. I thought I was following instructions, but I guess not. Losing data due to "user error" is far more likely than a hardware failure, I think, especially if this is your first try. Therefore, I think RAID 1 is not such a hot idea for ordinary users, and not for accident-prone klutzes like me.

RAID 1 is not the equal of a backup. Deleting/changing files, does it on both drives. A worm works on both drives.

 

[douglite]

Is RAID 0 worth it?

When I can do a complete backup of my system from three different source volumes (my WD800JB drive, my 3x WD1200JB RAID 0, and my 200GB Maxtor backup drive) to my 2x Samsung SP1213C 120GB SATA RAID 0, and copy from of those three sources simultaneously, RAID 0 is most difenitely worth it. I copied 4000 MP3s from the Maxtor and 15 GB of CD Images from the 3x WD RAID - neither copy process took a performance hit because of the other. That's some serious write speed on the part of the Samsung drives. When I'm LANning with my buddies, I'm always waiting for them in game while they're still loading the levels. Granted my box is heavily optimized (see sig) but hard drive performance is important, and squeezing everything you can out of your HDDs is one of the biggest performance gains available. Those Samsung drives get 72.6 MB/sec average on random read test and an incredible 146MB/sec average/139MB/sec minimum buffered read. :Q

As for the more likely drive failure when using RAID 0, my new setup was to slim down my box from seven HDDs to three, with Samsung SATA RAID backed up on the 200GB Maxtor. My OS is on a 10GB partition on the RAID, with all of my data on the other ~210 GB, which I manually copy over to the ~190GB Maxtor. Poor man's RAID5, but my data is still ok if one of my drives fail. As for the Windows installation, ask yourself how long one of those lasts anyway?

As another point, my SATA RAID is on 32KB block size. Kicks *$$ on sequential transfers (admittedly higher CPU usage - but who cares when you got 3 GHz+?), does at least somewhat better on my gaming tasks, including the random read benchmark in Aida32.

You've got to be smart with RAID 0, and it's not perfect. But I got a huge fault tolerant performance boost out of a pair of $76 Samsung drives on Neweg refurb and a $80 rebate special at CompUSA in the 200GB Maxtor. Sure, you can do a similar setup with 74GB Raptors and 160GB PATA too, then you'd really be cooking...

 

[addragyn]

RAID 0 = fart mufflers for computers

To double the chance of data loss and get absolutely no benefit in access times is not a performance increase for a desktop machine. Very likely if you need high STR you know it.