To RAID-0 or not to RAID-0 ... answers inside

[DocSilly]

Ah yes, there're still plenty of people telling you "RAID-0 is better than sex experience when it comes to games, a real must have" ... without providing any real proof.
What can I say ... well, I can say RAID-0 isn't worth the risk, it has never been worth it but people still say it is.
Now here's the proof in this review > http://www.anandtech.com/storage/showdoc.aspx?i=2101 (which has been on storagereview with comparable negative results before).

Final Words
If you haven't gotten the hint by now, we'll spell it out for you: there is no place, and no need for a RAID-0 array on a desktop computer. The real world performance increases are negligible at best and the reduction in reliability, thanks to a halving of the mean time between failure, makes RAID-0 far from worth it on the desktop.
...
Bottom line: RAID-0 arrays will win you just about any benchmark, but they'll deliver virtually nothing more than that for real world desktop performance. That's just the cold hard truth.

You can also look at the latest post on storagereviews newspage here > http://www.storagereview.com/
I hope this finally answers the question if RAID-0 is really worth it (and it looks like it ain't at all).

edit: fixed link to Anandtechs review

[buddhazen]

nice link
i guess i can use the extra money for more food at the taste of chicago

[nexus_7]

I personally thing that is utter BS!

Greg

[DoPeY5007]

eh, I like my RAID-0

[FlyingPenguin]

I used RAID 0 for a year and came to the same conclusion. Performance increase was negligable (I did notice HL maps loading slightly faster, but nothing to write home about) and the risk factor made me nervous.

Granted I ghost my drive regularly, but not the whole drive just the boot, apps, and data partitions. It still would have been a hassle to suffer a crash, and statistically you greatly increase your chances with RAID 0.

Not worth it for the average desktop. It was worth it once for video editing, but with modern ATA133 drives with 8Mb cache, that's not a good arguement either.

[VidmanII]

I'll take a read burst speed average of 100MB/s on an integrated SATA RAID with 10,000RPM HDD's w/ 8MB cache <b>ANYDAY</b> over a single 7200RPM ata133 HDD w/ 8MB. There's no comparison. Best upgrade you can do in my opinion, in that the HDD is the biggest bottleneck in your system. If you're dilligent about backups, it doesn't matter if you run RAID 0 or a single IDE.

[DocSilly]

VidmanII

I guess it is a given that "an integrated SATA RAID with 10,000RPM HDD's w/ 8MB cache" is faster than "a single 7200RPM ata133 HDD w/ 8MB" lol

Wouldn't it be more interesting to compare it the other way around? Or comparing "an integrated SATA RAID with 10,000RPM HDD's w/ 8MB cache" with "a single 10,000RPM HDD's w/ 8MB cache" ... but wait, that's exactly what they did in Anandtechs review

The review clearly shows that there is little to no performance increase by going to RAID-0 ... just look at the game loadtimes (yes, look at those since almost everyone defending RAID-0 will tell ya it reduced their levelloadtimes greatly) http://www.anandtech.com/storage/showdo ... =2101&p=10
Or look at the other benchmarks .... real world performance improvement of Raptor II RAID-0 over single Raptor II is 2-4% average.

[VidmanII]

To be honest, I didn't even read the article you mention above. I'm not much of a gamer so those numbers don't matter much to me. I use my rig mostly for music and movies etc which entails sliding BIG files around. I notice the difference using raptors RAID 0 in things such as doing a winXP OS install in less than 16 minutes, much faster boot times, copying huge folders to and from other drives etc, etc.

I used the comparison I mentioned because I think it's fair to say that most desktops still use 7200rpm HDD's.

Another reason I went with the da raptors is that they have a 5 year warranty. I don't have 3-4 PCs like alot of people here. Just my rig and my wife/kidz' rig. So over the last 6 months I chose to completely "ghetto" mine out with some high end stuff ( high end relative to my budget) so I could put off the upgrade cycle for a good while. What with Ghost, an additional Samsung Spinpoint 120GB 7200/ ata133 / 8MB, and a Plextor DVDRW for storage, my redundancy/backup needs are all taken care of.

It's fast, it's stable and it works for me. Around here, that's all that matters.

One other note. I'm not convinced that all reviewers take the time to find the best stripe/cluster when setting up a RAID 0. THAT can make a BIG difference in performance.

Here's a 7200RPM Samsung SpinPoint 120GB ata133 / 8MB cache....



Here's a RAID 0 16K stripe/ 16K cluster w/ 2 x 10K rpm 37GB Raptors / 8MB cache

[The_Frapster]

I don't think you can argue with that there at all. Nice benchmarks damnit.

BTW, how does the drive I'm getting from you fair with these benchs?

[VidmanII]

Originally posted by The_Frapster
I don't think you can argue with that there at all. Nice benchmarks damnit.

BTW, how does the drive I'm getting from you fair with these benchs?

Yours will look similar to the Samsung I would think. Keep in mind that the one your getting is a SATA drive and not an IDE. Excellent drive by the way. Extrememly quiet and very good all around performance.

[VidmanII]

Here's why Banand's article is unmitigated BS.........

c/p

We have seen in the past that for most desktop applications, the largest stripe size that a desktop RAID controller will offer is usually the best choice for performance. With Intel's ICH5/6, that translates into a 128KB stripe size, which for our comparison is what we decided to go with. The other stripe size options didn't offer any better performance for our desktop test suite.

end c/p

The fact that he's using a 128K stripe and makes no mention of trying different cluster sizes tells me the test isn't valid. Especially since he's probably using winXP's default NTFS 4K cluster size. I've done extensive testing of stripe/cluster combinations with both FAT32 and NTFS and there are HUGE differences with the various stripe/cluster configs.

The best RAID 0 configs generally are ones where the stripe/ cluster are equal ( e.g. FAT32 32K/32K and NTFS 16K/16K) The fact that he's using a 128K stripe and what appears to be the default NTFS 4K cluster size, removes any credibility as far as I'm concerned. To even talk about testing RAID 0 and not mention cluster sizes is ludicrous. What's sad is people take this kind of crap for gospel instead of testing on their own and coming to their own conclusions.

If you don't want to use RAID 0, fine. But I certainly wouldn't make my decision based on that lame test. That's for sure.

[Toad21]

Very nice Link. I was just thinking about this earlier today. So basically, yes, RAID0 any 2 drives and you will see a perfomance increase. But as far as it being worth it, it is like buying another 200$ video card to get 500 3d marks. Get it?

[DocSilly]

Why did Anandtech test only the default settings?
The Joe Sixpack users have no idea about stripe/cluster size and whatnot when they setup RAID-0 ... they plugin the drives, format them and go with leaving everything on default ... so those are the results they would see. I doubt many would know info like http://faq.storagereview.com/StripeSize ... or know how to format their HDDs with a different cluster size. I also doubt they'll sit down a couple days and recreate/reformat their RAID-0 and benchmark their main applications to see which cluster/stripe size combo would be the best match.

Your benches are nice and show the STR and it is no surprise that the RAID-0 wins there by a big margin. STR is important when you move many big files all the time or work with huge files like with Photoshop. The pure STR increase could be also seen in this older RAID-0 vs single drive test http://faq.storagereview.com/SingleDriveVsRaid0 (2 driver RAID-0 is twice as fast as single drive) though it is the only thing that greatly benefits from going to RAID-0.

The Anand review also didn't really say "RAID-0 suxs by default":

There are some exceptions, especially if you are running a particular application that itself benefits considerably from a striped array, and obviously, our comments do not apply to server-class IO of any sort. But for the vast majority of desktop users and gamers alike, save your money and stay away from RAID-0.

Or Storagereview:

Again, RAID 0 does have its advantages in a handful of key applications and uses where data files are huge and/or data requests are highly sequential in nature. Data requests are not highly sequential, however, in typical desktop productivity and most gaming usage patterns, the most often cited in "Help me build my RAID 0!" posts.

Finetuning stripe and cluster size would improve the results for some benches but it would on the other hand also decrease the results for other benches, depending on wether they're STR dependant or for non-sequential I/O applications as explained in the stripe faq linked above.
I dunno if yer can get a lot more with any finetuning regarding non-sequential I/O applications (normal desktop usage), they're already using 128KB stripe size which is targeted at that usage and I don't think you can gain much using a different cluster size.
Your array seems to be more finetuned towards STR and benchmarks like HD Tach / WinBench 99.

[VidmanII]

Msg to Joe 6-pak

I think Anand knows all and we should all do what he tells us !



I'll be keeping my RAID

[smb]

joe sixpack probably has no business running raid of any kind. None of use here has the last name of sixpack. Most of us here are looking for that extra oomph in our computers that's why we use raid, or scsi for that matter.