How good is Raid...really?

G

govtcheez

Gawd
Joined
Jul 11, 2002
Messages
660
I've been itching to try a raid setup lately. Is it really worth it? Are there any stability issues? I have a 80gb WD SE HDD, with 8mb buffer. Can I use any HDD with the same specs? or will it have to be the EXACT same drive? Will I get a big performance increase? Will everything boot up, and run that much faster?


Thanks. :)
 
It's best to use identical drives in a RAID array. There's been quite a debate recently, but here are the facts of the most common forms.

RAID0: Striped across several drives. Gives no real performance increase in real world usage, especially not gaming. If one drive dies, your data is lost on all drives in the array. Total space is total of all drives put together.

RAID1: Known as mirroring. Typically two drives, but you only get the total capacity of one drive. When data is written to one drive, it's written to the other as well. If one drive dies, your data is still safe on the second drive. No performance change from a single drive.

RAID5: Add up total capacity of drives, and subtract one drive's amount. This is your total storage space. Built for speed and redundancy. This is by far the best option for most usages, but can be expensive. Minimum of three drives I think.

There are other RAID types, but these are the most common. The RAID0 in a desktop fad has come and gone, since all testing is now proving it does nothing for you, except make you worry more about your data.
 
RAID1 read performance actually goes up, because reads can
be split between the drives. But I can't imagine that would
be a factor for anyone choosing RAID1.
 
govtcheez said:
I've been itching to try a raid setup lately. Is it really worth it?
Click to expand...
Most likely no, but that depends on your specific usage.
govtcheez said:
Are there any stability issues?
Click to expand...
Yes, there are. Running two or more hard drives in Raid-0 will significantly increase your chance of data loss because it only takes 1 of the drives to fail for you to lose all of the data on the array.
govtcheez said:
I have a 80gb WD SE HDD, with 8mb buffer. Can I use any HDD with the same specs? or will it have to be the EXACT same drive?
Click to expand...
It is usually best to use a drive of equal size and performance specifications. Exact model would be the best option.
govtcheez said:
Will I get a big performance increase? Will everything boot up, and run that much faster?
Click to expand...
You will most likely notice little performance increase if one at all. The most common situations where IDE Raid-0 actually helps is when you are working with very large files like video editing or something. I doubt your XP boot-up speed will increase very much either. In fact, the time it takes to run through the IDE Raid controller bios would probably negate any kind of speed increase in your computers startup time and may actually make the entire process longer. In my opinion, IDE raid is used far too often as a performance upgrade without any research or though put into the specific usage.
 
aug1516 said:
In my opinion, IDE raid is used far too often as a performance upgrade without any research or though put into the specific usage.
Click to expand...



that's kind of what I thought....that's why I'm doing my "research". :)
 
aug1516 said:
In my opinion, IDE raid is used far too often as a performance upgrade without any research or though put into the specific usage.
Click to expand...

I agree. I have a single drive running on the onboard RAID controller on my motherboard purely to isolate it from my Linux HD, DVDRW and hotswap tray. It doesn't run any faster than on the regular IDE channels. Having a single regular IDE channel to give to an IDE tray is nice so that you don't have to worry about jumper settings since nearly every new drive is jumpered master or single by default.
 
I would have to say Raid0 is great
I noticed a huge difference in map loading times with my drives in a Raid0 setup compared to individual drives. But do not use a different drive model for an array. A friend of mine has an array of 4 120s 3WD 1seagate and he gets data corrupion due to the fact not all the drives are the same

Now for my shameless plug: Use seaget drives up until 10 months ago i was using Wd and Maxtor drives some little over a year old and I had four drives die in a two month period I have had no problems with my seagate plus they are quieter than me sneaking into the house after curfew
 
OCgamer666 said:
I would have to say Raid0 is great
I noticed a huge difference in map loading times with my drives in a Raid0 setup compared to individual drives. But do not use a different drive model for an array. A friend of mine has an array of 4 120s 3WD 1seagate and he gets data corrupion due to the fact not all the drives are the same

Now for my shameless plug: Use seaget drives up until 10 months ago i was using Wd and Maxtor drives some little over a year old and I had four drives die in a two month period I have had no problems with my seagate plus they are quieter than me sneaking into the house after curfew
Click to expand...

Have you actually tested your map loading or is it just a feel thing? :rolleyes: General consensus seems to be it doesn't really do jack squat for most games level loading, in fact level loading is usually so CPU intensive as it decompresses textures that the HD speed is pretty much a non factor I hear. /shrug

If it works for you and you like it, fine, don't recommend setups w/o empyrical data tho.
 
I recently setup a RAID 0 array using two SATA 120GB Seagate Barracuda 7.200s.

While the performance increase is not quite what I hoped for, I managed to come up with these results.

A stop watch was used for all tests; (all times are actually a rounded average of three tests each way, with a reboot between each test, fresh install of WinXP Pro SP2)

Doom 3 maps:

Alpha Labs 3 (without raid 0) : 45 secs
Alpha Labs 3 (with RAID 0) : 30 secs

Winrar:

(I do a lot of encoding and decoding of movies and such, and use winrar to compress shit)

decompression
1.1 GB .tar file (without raid 0) : 65 secs
1.1 GB .tar file (with RAID 0) : 49 secs



As you can see, in hardrive intensive tests, RAID 0 is slightly better performing.
However, this depends on many different factors. I use the Intel RAID controller on the I875 chipset. I have a promise controller but don't like it much.

I also noticed moving files around seems faster.

Warning: with a RAID 0, you need to defrag more often.

It is ultimately up to you. I will probably keep this RAID 0 array and throw in an old WD 40gb ata 133 drive for backup.


Hope this helps.
 
Warmonkey said:
However, this depends on many different factors. I use the Intel RAID controller on the I875 chipset. I have a promise controller but don't like it much.
Click to expand...

That's probably why you have better read times. I wish I could remember the link I found it at but there was a review of the various south-bridge RAID setups in 0, 1, 0+1 and 5 and their performance. SIS had better throughput hands down but drew a lot of CPU and had slow read times. Intel's was the clear winner with the lowest CPU usage and best to second best read/write times. Nvidia's used as much as 20% CPU in RAID 0 and 1.
 
Warmonkey said:
You doubt my claims?

Just so you know, why would I lie about load times? That is idiotic.
Click to expand...

No I do not doubt them Mr. Paranoia, I was honestly saying they were interesting because I had seen results from level loads on other games that showcased a much smaller gap when run under RAID 0 arrays. :rolleyes: Doom 3's textures and levels may very well be more demanding/bigger than anything else out there tho so I could see how results would be different, ergo, interesting.
 
Impulse said:
No I do not doubt them Mr. Paranoia, I was honestly saying they were interesting because I had seen results from level loads on other games that showcased a much smaller gap when run under RAID 0 arrays. :rolleyes: Doom 3's textures and levels may very well be more demanding/bigger than anything else out there tho so I could see how results would be different, ergo, interesting.
Click to expand...

Sorry then. Just seems some of the guys on this forum are a bit self righteous.

:cool:
 
Is the claim of little or no performance gained with IDE raid setups due to the fact that most (errr, all of the ones I am looking at) are software raid setups? My sources are
http://linux.yyz.us/sata/faq-sata-raid.html#ich (extrapolate the raid device for your mobo)

Would a true host adapter for SATA (do they even exist?) make a difference?

I've seen a Raid-0 setup, scsi work quite well in Windows. Opening an MP3 directory in windoze explorer takes a noticeable time without the raid setup, with raid, its seemingly instantaneous.
 
software raid is lame. Do not do it. Hardware raid is the only thing we are talking about. Yes they do exist, I am using an Intel SATA RAID controller.
 
I've used the RAID controllers on my 865 and 875 boards. I, along with others, have done extensive testing on various setups, using various apps and games. Our results have always concurred with consensus, that RAID0 is not worth it.

Many people can argue about most of it, but we all agreed gaming gets absolutely no benefit from it. I'm raising an eyebrow reading the claims on Doom3, because that was one game I tested. Along with several maps on BF9142 and UT2004. I ran the game "modified" to only use the hard drive...no CDs were involved. Not one of the games, on any of the 4 systems I test showed any real increase using RAID0. My images were kept clean using Norton Ghost. It was as scientific as I could get, and I do have a science degree that involved extensive research tactics. We could argue for years about this, but the fact is, more and more people and sites are doing tests for themselves, and finding out the truth. Disagree if you want, but the facts and just that...facts.
 
Generally, the claims that RAID0 isn't fast actually have
nothing whatsoever to do with RAID. The claim is that you
don't need faster disk IO.
 
I think one common misconception in this thread by people who are posting things about RAID0 not helping at all is that the recient publicity/outcry has been based on using very fast WD Raptors in RAID0. If you take a couple of older, crappy 40gb, 5400RPM IDE drives and make an array with them they *will* be faster than the single drive. It's Redundant Array of Inexpensive Disks!
 
Vertigo Acid said:
I think one common misconception in this thread by people who are posting things about RAID0 not helping at all is that the recient publicity/outcry has been based on using very fast WD Raptors in RAID0. If you take a couple of older, crappy 40gb, 5400RPM IDE drives and make an array with them they *will* be faster than the single drive. It's Redundant Array of Inexpensive Disks!
Click to expand...


Tested already...using 2 older Maxtor 5400 rpm 2 MB cache 20 GB drives. I was using an IT7-MAX2 board because of it's PATA raid controller onboard. Same results apply. I see this comment posted often, but I'm not really sure why some think it matters which drives. It really doesn't matter whether you use Raptors or older drives. As long as the single drive tests are done with one of the drives used in the RAID array, it isn't going to matter. This also keeps the validity of the experiment. These tests are designed to test RAID performance...not old drives versus new.
 
Vertigo Acid said:
I think one common misconception in this thread by people who are posting things about RAID0 not helping at all is that the recient publicity/outcry has been based on using very fast WD Raptors in RAID0. If you take a couple of older, crappy 40gb, 5400RPM IDE drives and make an array with them they *will* be faster than the single drive. It's Redundant Array of Inexpensive Disks!
Click to expand...

There's tests with older drives and current non-Raptor drives that don't indicate that... 'Sides what's the point of that, most enthusiasts out there aren't building RAID array as a way of cost cutting... they're doing it with the hopes of performance gains beyond what the newest drives offer.

At the end of the day it really comes down to the specific application and the controller I guess much more than the drives used.
 
The other thing I should mention is that all of the tests were done using onboard controllers that were NOT on the PCI bus. The PCI bus (such as using an add-in PCI RAID controller) will not offer the same performance as the onboard controllers sitting on the southbridge bus. This wouldn't be a fair comparison, so we elected to only use onboard, native controllers. My SATA tests done with my two Raptors were done on an IC7 and an AI7 motherboards.
 
wallijonn said:
Why not just use one Raptor instead of 2 drives in a Raid0?
Click to expand...

That's what I chose to do. After my testing, and after seeing others had the dame results, I was gaining nothing by using my Raptors in RAID0. It's no question the Raptors are the fastest drives though, so I split them up into two computers. I get much more benefit out of them this way than in a RAID0 setup.
 
Warmonkey said:
I recently setup a RAID 0 array using two SATA 120GB Seagate Barracuda 7.200s.

While the performance increase is not quite what I hoped for, I managed to come up with these results.

A stop watch was used for all tests; (all times are actually a rounded average of three tests each way, with a reboot between each test, fresh install of WinXP Pro SP2)

Doom 3 maps:

Alpha Labs 3 (without raid 0) : 45 secs
Alpha Labs 3 (with RAID 0) : 30 secs

Winrar:

(I do a lot of encoding and decoding of movies and such, and use winrar to compress shit)

decompression
1.1 GB .tar file (without raid 0) : 65 secs
1.1 GB .tar file (with RAID 0) : 49 secs



As you can see, in hardrive intensive tests, RAID 0 is slightly better performing.
However, this depends on many different factors. I use the Intel RAID controller on the I875 chipset. I have a promise controller but don't like it much.
Click to expand...

One could perform a t-test to determine whether these results are significant. You will need to collect more than two samples. At face value, it does appear to be significant.
 
The argument is not against using Raid-0 at all, but rather using it as common practice to increase system performance. Under the right type of usage Raid-0 can provide a worthwhile performance increase and would be well worth the investment. The recent articles written about Raid-0 have shown that gaming is not one of those situations that routinely increase performance enough to make Raid-0 worth implementing.
 
I would also like to point out that Raid arrays arnt always faster but a lot of times allow a sustained throughput at the speeds where other drives just peak
 
You must log in or register to reply here.
Back
Top