raid 0 10k raptors

markish

Distinguished
Oct 10, 2006
38
0
18,530
0
would having two 10k raptor 150gb hard drives in raid 0 (300gb) run faster and have better performance then just having one without raid. i was reading around and i dont know much about raid but from what i was understanding in most of the posts is that having them in raid 0 makes them run faster then just having one?
 

nh484000

Distinguished
Dec 24, 2006
164
0
18,680
0
Yes, for most things. Nothing will run slower, but some things will run faster. In raid 0 there are two harddrive to write the data to instead of one, this gets rid of that bottleneck.
 

zenmaster

Splendid
Feb 21, 2006
3,867
0
22,790
3
See .................................

http://www.tomshardware.com/2007/03/12/cheap_raid_ravages_wd_raptor/

This shows the boost going from 1 7200 to (2) Raid-0 drives will give you.
Expect something similar from the raptors, but perhaps some less gains since the raptors will be less of a bottleneck to start.
 

russki

Distinguished
Feb 1, 2006
548
0
18,980
0
OMFG you morons...

I am tired of this argument.

Pray tell me, what bottleneck?! And what, in Raid 1 there is fewer than 2 harddrives? What is the meaning of the sentence in Raid 0 there is two harddrive to write to, then? How about Raid 5? There are at least 3 then, is it even faster than Raid 0, by your logic?

Ignorance is contagious.

To the original poster - the answer is it depends on your use. But if you're planning to RAID Raptors for a home PC, you're probably somebody with too much money to burn without actually knowing what you're getting into.

Here, read this at your leisure. It's old, but still true.

http://www.anandtech.com/showdoc.aspx?i=2101&p=1
 

cb62fcni

Distinguished
Jul 15, 2006
921
0
18,980
0
Unless you do video editing or encoding on your machine you won't see any major performance increase, and in some cases your machine will be less responsive because RAID0 increases seek times. It would be a waste of money. Upgrade your CPU or GPU instead, it will net you a much higher performance gain.
 

zenmaster

Splendid
Feb 21, 2006
3,867
0
22,790
3
Read the conclusions and test results for both your link and my link.
Both articles confirm an increase in performance.

It is up to the user to decide if the performance increase is sufficient to offset the risks.

RAID-5? I never mentioned RAID-5.
But yes, in some cases RAID-5 can be faster than RAID-0.

The key is to have a sufficiently large enough number of drives and a distributed parity bit and a controller that can handle.

Your failure is that you are making value judgements where others are not. Is Raid-0 faster? Yes. That is what we claimed. That is the case.

We made zero value judgements in regard to reliability or the best way to spend money for an upgrade.

Rather we ponited the poster to a comprehensive article that would discuss and show these things.
 

russki

Distinguished
Feb 1, 2006
548
0
18,980
0
As has been beaten to death in the appropriate thread, Tom's article is anything but comprehensive. All it does is show peak performance. Theoretical peak performance. The value of that to the average user is precisely 0.

I do not think, all things equal, RAID 5 is faster than 0. Parity takes overhead over striping. Of course, the comparison is difficult because no one in the right mind would take 0 over 5 if they have a controller that can handle 5 in an efficient manner. And I imagine controllers optimized to run 5 are not really optimized to run 0. (although who knows what latter means. And I'm guessing here). I was simply referring to your point about multiple drives, which is not what makes 0 faster - it is the way the drives are used. Heck, 1+0 or 10 requires at least four drives but is but it is in no way faster than 0 with 2.

The point is, even if you understand all these issues, it did not come through in your post. In fact, you just confused a person seeking advice into thinking that spending some $250 (for the sake of the conversation) or $1.70 per Gb is going to give him screamer of a system. Which is a fallacy. Like someone pointed out, that money may be better spent elsewhere. Now, if he's doing tasks that require fast sequential transfers, all the power to him, go RAID0 (hopefully with back up, and not for the system partition, etc.). Something makes me doubt that, though...
 

markish

Distinguished
Oct 10, 2006
38
0
18,530
0
im looking to get a high end gaming machine soon when ati comes out with their dx10 card thats why i was asking.
 

cb62fcni

Distinguished
Jul 15, 2006
921
0
18,980
0
You may want to consider Hitachi's new line of Deskstars. 1TB for only $399, and the performance in phenomenal, they're very close to the Raptor 150 in real-world and application performance.
 

markish

Distinguished
Oct 10, 2006
38
0
18,530
0
Alright, here's what you're going to do, you are going to forget this raid performance thing as we are too tired of answering to the same question over and over again, if you want, look it up in the archives or check tom's review on raid for performance, but what you are going to do is buy a single hitachi 1tb drive, it performs as fast or faster in some tests as a single raptor, but only costs $400, so for the price of two raptors, you get a almost 4 times the storage, gee, I wonder, what's the better deal? Two seconds off your boot time, or tons and tons of storage and stability, yes, stability, something you will never achieve with raid 0 raptors unless you do raid 1+0, which requires 4 drives and is a waste of money when you could just get a single tb drive and be happy
thank you and thats what im going to get. im glad i got a straight answer from someone without all the confusing stuff :D
 

mrwoo

Distinguished
Mar 23, 2007
10
0
18,510
0
Markish..

I have no idea what you know about raid 0. The posters here prolly know more than I, but I will offer my layman's terms & info for you.

Nutshell.. Raid 0 basically takes a given portion of data, and splits it into pieces, part onto each drive in the raid array. It give a performance increase for sure, but not in effect of twice as fast.

Most boards I have used or setup with Raid 0 have integrated chips, which typically do not offer the performance of a dedicated card.

It is also worth noting that early SATA drives were sometimes P-ATA drives with a converter for serial, meaning you may not be any faster than standard ATA. It would seem that a true (or have heard called Native) SATA drives are SCSI drives with SATA interface. This is just what I have read from different places, so don't flame me for repeating.

My experience with Raid 0 has been this..

A new, 8mb cache 7200 rpm ATA-100 hdd, can sustain on a new mobo with new cpu, writes of around 85, reads around 75. A new SATA 150 hdd, same board/cpu, can sustain writes around 100, reads around 85. New SATA 150 hdds, RAID 0, typically sustains 125 writes & 100 reads. This is largely dependent of course on the drive & the Raid controller. The VIA 8237 chip has offered my pretty good performance in the past, usually better than the Sil or Nvidia chips. However, the Nforce4 chip I am happy to say gives me slightly faster r/w's, with a little more cpu use though.

I have been building some Core2 Duo's on the ICH7 (intel) chip controllers, and have seen some really really great results using some Raptor 150's in Raid 0. They definately put the 7200 models to a lower level. It is also interesting to note, that from my experiences, the 3gb type SATA drives do not necessarily perform better than 150's. Using Raptor 75's is still better than SATA 7200rpm, but the Raptor 150's are better.

My take on the whole thing, having played with servers & desktops using both ATA, SCSI & SATA, having both integrated controllers as well as dedicated cards?

You gain approximately 25% more by using Raid 0. I care not to comment about whether Raid 0 is the fastest or not, as that seems to have been covered already :)

Of course, after you digest what Raid 0 does, and figure out after some testing how much it give YOU in performace, you can then dive into the whole Stripe Size to Cluster Size debate. Let me just say, that with enough reformatting and reinstalling, you CAN see some very noticable differences between them.

later.
 

markish

Distinguished
Oct 10, 2006
38
0
18,530
0
ok in a nutshell which will give faster performance raid 0 150gb 10k western digitals or 1tb hitachi. im not looking for which is the better for the money im wondering which is the best for performance. im looking to spend around $4000-5000 on this system. also i was planning on putting in a 750gb storage drive anyway.
 

mrwoo

Distinguished
Mar 23, 2007
10
0
18,510
0
Hmm. I have read lots on the very topic of what will be the fastest. Most credible reports or tests show a Raid 0 array, consisting of Raptor 150gb drives, as just edging out a SCSI Ultra 320 Raid setup. So it would seem that at the moment, the raptors rule, in terms of pure adrenaline.

It is also true that looking at $$ per gb, they do not rule.

It is also very true that many times having a dedicated Raid card can improve things even more. I have used some 3ware & adaptec, both of which I have had good luck with.

I have no data on the Hitachi.

Myself, I am hoping that the SAS drives will become mature and offer the best performance overall. But that may be a bit yet.

My opinion, should I be lucky enough to build a system for 5000. I would say that right now the ICH8 controller is the best integrated chip, and obviosuly the Core2 duo cpu.

However, the ICH9 is out or soon to be, and it may be interesting to see some reviews. My testing and tinkering have led me to belive that for performance, that is, loading up your game as fast as possible :) that Raid 0 using Raptor 10k 150gb, NOT the 3gb or as called SATAII, but what is called SATA 1.5

That being said, you just asked for performance with a high pricetag, and that is what I would buy. Of course I have built 5 systems now with just that for peeps, so I could be biased a little. I have not tried any nforce 5 or 6 chips yet.

later.
 

choirbass

Distinguished
Dec 14, 2005
1,586
0
19,780
0
agreed. if you want performance raid, ditch the onboard (its software based as most things are done by the cpu, resulting in typically poor I/O management and raid 0, 5 performance), and consider a dedicated add on hardware based raid controller (hardware based due to an onboard logic chip, dedicated memory, etc)

heres an example of a hardware based raid controller for pcie, its also the cheapest areca newegg has, but offers far more performance than onboard does
http://www.newegg.com/Product/Product.aspx?Item=N82E16816131003
 

choirbass

Distinguished
Dec 14, 2005
1,586
0
19,780
0
thats why im suggesting just going for a single fast hdd instead (the 1TB hdd you suggested for example)... youll get close to the same realworld performance as onboard raid offers (within reason), for possibly less cost even... and for redundancy, just use a few independant (internal/external) hdds containing replicate data, making regular backups as well...

as far as performance boosts, a good hardware controller will offer pretty close 100% improvement for raid 5 even, but, theyre expensive nonetheless, whereas onboard is anywhere from 0% to maybe 30% or so for raid 0 (greatly depending on the specific use too), and raid 5 is typically very poor (i guess due to cpu overhead dealing with parity *shrug*), and, theres no real I/O management, so games dont really benefit much either typically (again, enough threads on that already here)
 

choirbass

Distinguished
Dec 14, 2005
1,586
0
19,780
0
i agree. im not too sure either about it being 100%, but i do remember reading someone posting about their raid 5 performance even pretty close to that (it would also depend greatly on the controller too, again)

but for the price, you simply cant beat onboard, seeing as how its free, lol

as far as the raptors, they would mainly just benefit for random access times and responsiveness as an OS hdd im sure (unless the 1TB hitachi has near 7-8ms access times too)
 

choirbass

Distinguished
Dec 14, 2005
1,586
0
19,780
0
well... free in the sense that its not an additional seperate purchase from the motherboard when you buy it. then again, yeah, i guess it really isnt free either
 

choirbass

Distinguished
Dec 14, 2005
1,586
0
19,780
0
i guess no taco for me, lol... i googled and referenced numerous online language translators and different sites that contained parts of it... but, to no avail for an actual translation... i was assuming that greek might even work too, even if loosely... oh well, lol
 

xaat_kil

Distinguished
Jan 24, 2007
155
0
18,680
0
them are ducks.
they are not!
see them wings?
why, I be!! Them are ducks!
I get a taco! Make it trimmed up nicely...unless that's not the kind of taco you be talkin' about.

P.S.-RAID 0 is a groovy experiment for desktops, but not really worth the effort, expense, and security problems, unless you backup alot, at which point if you're spending 5% of your time doing backups, there goes your speed advantage.
 

cb62fcni

Distinguished
Jul 15, 2006
921
0
18,980
0
Look man, in a nutshell, as I said earlier: Unless you plan on doing A/V editing, you won't notice a difference. RAID0 doesn't magically make everything faster. There is NOT a bottleneck on your HDDs when it comes to gaming and general purpose use, load times are primarily limited by CPU and RAM. The Raptors bench a little faster, but two of them aren't even halfway to 1TB. I'd get the 1TB drive.

http://www.anandtech.com/storage/showdoc.aspx?i=2949

And whoever said that Raptors in RAID are faster than SCSI was high.
 

mrwoo

Distinguished
Mar 23, 2007
10
0
18,510
0
Hmm. I must be high then, because I have read I believe 2 articles on the fight between sas, sci, & sata, with sas 'theoretically' creaming both of them, but also showing a slight lead to sata 2.5 over u320. I am sorry but Idid not bookmark it, nor do I recall what controllers or drives were used, other that there was an lsi &adaptec sas controller being tested for the sas, and that the raptor 10k was one of the sata tested. Granted, I could have misread. Or the review could have been bunk.

lol

As to whether or not Raid 0 makes any difference... well, I would say it does make a difference. Like I stated earlier, the average difference that I see is only at max 25% faster. IMO, it does make a difference loading a game. It does make a difference transferring files. Pretty much anything and everything disk related, I have found Raid 0 to be faster... depending.

Being your average everyday geek, I wanted to find out. Ah, but you are already saying "yeah, whatever. and just how do you know?"

lol. that is what I said too until I investigated.

Without all the droll details, I wanted to know does Raid 0 improve performance? And how to tweak it.

I took 2 sata drives, installed XP on them @ stripe sizes of 8,16,32 & 64, and each of those stripe sizes was installed with cluster sizes of 4,8,16,32 & 64. That makes for a lot of install lol.

Then I proceeded after each install to run a series of tests, in the exact same order int he exact same method, starting on stripe 8 cluster 4, working my way up the clusters until 64, then back to stripe 16 cluster 4.

And what I found to be true is that

A. drivers affect performance
B. cluster to stripe size affect performance
C. performance is indeed greater that either single SATA or ATA hdd.
E. performance is measurable in many different ways, but is measurable
F. that all of this testing can be recorded and show that while Raid 0 can be considered 'faster' than not Raid 0 (being whatever it is, say a single ATA hdd), that it is not particuarily that MUCH faster.

But mostly, those of us who do not 'know' much, can seem to 'feel' the difference. Maybe that is our imagination. As for me, whether my methodology of testing is 'crap' or not, I use the Raid.

And, because there have been a lot of good points here from you who seem to know what you are talking about, the number one thing on my list for tommorrow is to check out that Hitachi 1tb drive. Sounds interesting.

later.
 

ASK THE COMMUNITY