RAID-0 proved ineffective at boosting desktop application/game performance

D

DougLite

Supreme [H]ardness
Joined
Jan 3, 2005
Messages
4,764
Clicky, check out the bar graph towards the bottom of Eugene's post. SR will be posting a pretty comprehensive article on RAID, hopefully soon. The graph is a selection from the results obtained for that forthcoming article.

Even running _four_ WD740GDs in RAID-0 is not enough to match the single user performance of a single WD1500ADFD. The 4x RAID-0 array manages a tie in the OfficeDriveMark, but falls short in the remaining four of SR's Desktop DriveMarks. Also observe the 4x RAID-0 array delivering _slower_ performance than 2x RAID-0 in FarCry and WoW, and only managing a gain of a few percent in the Sims 2.

You are welcome to say what you want, but the simple fact remains that no RAID setup will deliver improved desktop performance. Power users looking to increase storage performance should look to faster single drives or adding indendent spindle(s) to service additional loads.

RAID-0's shortcomings are well documented:

A) Reduced reliability
B) Increased heat/power draw/noise
C) Increased system complexity
D) Greatly complicated backup and disaster recovery
E) Substantially increased cost

However, many have tried to justify/overlook those shortcomings by simply saying "It's faster." Anyone who does this is wrong, wasting their money, and buying into hype. Nothing more.

Graph posted with permission of Eugene at SR:
RAID0.JPG
 
Why not put two of these fast drives in a RAID 0 together? If you've got the cash make it fast. :D
 
Just figuring this out? RAID 0 was never a good idea. The only time it's useful is in a 0+1 or 10 array.
 
See reasons A-E above, and witness the _reduced_ performance of the WD740GD when run in RAID-0 in FarCry and WoW.

A single WD740GD manages over 800 IO/s in SR's FarCry tests, 833 to be exact, while neither of the RAID-0 setups can even break 800. WoW sees a similar drop, with a single WD740GD earning 671 IO/s, while neither of the RAID-0 setups can break 600 IO/s

Is anyone listening? RAID-0 is hype, nothing more.
 
RAID0 isn't all hype. I have a feeling you are doing this to slap it in my face from the other day.

RAID0 is great if you want to bind a few physical disks into one larger logical array. It's also better if you are hoping to draw lots of data off a network and onto the drive. I'm also curious as to what hardware controller did they use and what software revisions was used. I don't believe everything I see of the internet because 78% of statistics is made up :p
 
The vast majority of [H]'ers that are running RAID-0 aren't using it for disk to disk backup, as a content creation scratch disk, or in other situations where performance is limited by the sustained transfer rate of the disk subsystem. They are using it under the erroneous presumption that it improves desktop application level storage performance, which is simply untrue.

As for the controllers and methods SR used, I do have a little bit of info for you, and it's probably more than you expected:
Eugene Ra said:
we have all raid levels possible with up to 4 drives with arrays featuring the NL35, WD740GD, and WD4000 operating off of the SI3124, LSI SATA 300-8X, and 3ware 9550SX in addition to some less sweeping results with 4-drive arrays of the Atlas 15K II SAS and Cheetah 15K.4 SAS.
Click to expand...
 
DougLite said:
The vast majority of [H]'ers that are running RAID-0 aren't using it for disk to disk backup, as a content creation scratch disk, or in other situations where performance is limited by the sustained transfer rate of the disk subsystem. They are using it under the erroneous presumption that it improves desktop application level storage performance, which is simply untrue.

As for the controllers and methods SR used, I do have a little bit of info for you, and it's probably more than you expected:
Click to expand...


Alrighty, so they did do thier homework a little better.


I raid0 mainly as a staging area of massive amounts of data that needs to be transfered to fast and from fast... the data never really sits on those drives, it's kinda like a temp holding cell. As for my multiple file servers, i just run JBOD since RAID kills off too much space. Last night I had a RAID0 array collapse on me and I was kinda pissy about that, makes you reanalize your stance on it... I had no data loss (new array), but the reliability just isn't there with it.... and it's not always the drives fault.
 
I've used RAID 0 for the combined hard drive space, but now that the new Raptor is out, the pair I have aren't going to be sitting in my box for very long. Couple that with the significant performance improvements, this seems to be a nice upgrade. I'm starting to get into WoW, so it's good to hear that single drives dominate RAID 0 configs there. Can't wait to get my hands on the new one.
 
The only thing missing from those graphs is a single wd740gd. It doesn't make sense to me to have left it out; sure a single new drive is faster, but what about a single old drive? And it's not like you don't have the drive right there...

Wow, I just called the 74g raptor old.

Oh, I guess the other one would be two of the new raptors in raid 0. Right now it looks like apples to oranges. I'm not saying I support raid 0 on the desktop, but this post does anything but kill the myth for me.

 
unhappy_mage said:
The only thing missing from those graphs is a single wd740gd. It doesn't make sense to me to have left it out; sure a single new drive is faster, but what about a single old drive? And it's not like you don't have the drive right there...

Wow, I just called the 74g raptor old.

Oh, I guess the other one would be two of the new raptors in raid 0. Right now it looks like apples to oranges. I'm not saying I support raid 0 on the desktop, but this post does anything but kill the myth for me.

Click to expand...
I made light of the fact that both RAID-0 arrays of WD740GD are significantly slower than a single WD740GD in both WoW and FarCry, see post #4 in this thread.
 
Nice post DougLite. It's a shame some people still have to argue, despite the fact this has been proven over and over. :rolleyes:

Oh well...I'll be linking to this sticky for quite some time. Thanks.
 
My favorite line:

...and that RAID's applications arise more in the server world than in the desktop, despite what every major taiwanese manufacturer would have you believe.
Click to expand...

I fell into the trap, enjoyed my time, then stepped out and still couldn't be happier with my experience. I ran a striped array for a year with no tangible benefit for it other than a $20 RAID card that now has one disk connected to it.

Then the douchebag udaman stepping in with some tripe, which it looks like his forte. I bet he misses the point quite a bit. ;)
 
There are better and safer ways to increase disk space from 2 disks opposed to using RAID 0 to do so...
 
There's two problems with this whole thread here. One is you can't compare different configurations of different drives to measure the differences in the configurations. If you want to compare the differences in the configurations, put the same drives in the configurations and compare them. Usually the issues with not seeing raid performanice is both bus limitations and driver dependant raid. There have been way too many times I've seen people compare single drive performance to a striped array when the drives were striped by windows as dynamic disks. If you use a real hardware haid solution, you often will see the difference as long as the system isn't bus limited. I ran some quickie tests with a Promise Vtrak 15100 over a few days recently, which is an external IDE raid chassis that is seen by the system as a single SCSI drive. The striped arrays were simply faster than single disks. More drives in a sriped array would boost speed untill the bus was saturated. Is this practial, no, but that leads me to:

Two, Speed and safety are different things. If you're measuring speed, you can't say one isn't faster because it's not safe. Sure, say it's not safe as a caveat, warning, disclaimer, but not to discredit the speed. I'd rather get in a wreck in a NASCAR stock car than an Indy Cart, but does that make the cart slow? no.

So, in reality (which is where I live) not only do apples to apples measurements hold water in a competition, but factors that are not directly related to the measurement do not disqualify the results. Use those other factors in decision making, please do, but don't come to a conclusion like this from this kind of info.
 
matguy said:
There's two problems with this whole thread here. One is you can't compare different configurations of different drives to measure the differences in the configurations. If you want to compare the differences in the configurations, put the same drives in the configurations and compare them. Usually the issues with not seeing raid performanice is both bus limitations and driver dependant raid. There have been way too many times I've seen people compare single drive performance to a striped array when the drives were striped by windows as dynamic disks. If you use a real hardware haid solution, you often will see the difference as long as the system isn't bus limited. I ran some quickie tests with a Promise Vtrak 15100 over a few days recently, which is an external IDE raid chassis that is seen by the system as a single SCSI drive. The striped arrays were simply faster than single disks. More drives in a sriped array would boost speed untill the bus was saturated. Is this practial, no, but that leads me to:

Two, Speed and safety are different things. If you're measuring speed, you can't say one isn't faster because it's not safe. Sure, say it's not safe as a caveat, warning, disclaimer, but not to discredit the speed. I'd rather get in a wreck in a NASCAR stock car than an Indy Cart, but does that make the cart slow? no.

So, in reality (which is where I live) not only do apples to apples measurements hold water in a competition, but factors that are not directly related to the measurement do not disqualify the results. Use those other factors in decision making, please do, but don't come to a conclusion like this from this kind of info.
Click to expand...
The speed is not there either.
DougLite said:
A single WD740GD manages over 800 IO/s in SR's FarCry tests, 833 to be exact, while neither of the RAID-0 setups can even break 800. WoW sees a similar drop, with a single WD740GD earning 671 IO/s, while neither of the RAID-0 setups can break 600 IO/s
Click to expand...
In the case of 4X WD740GD in RAID-0 compared to a signle WD740GD, the RAID-0 setup falls short of the single drive by over 100 IO/s, a deficit of 25% in the WoW test and about 20% in the FarCry test.

I don't see where you get an arugment about these tests being bus limited...the SI3124, LSI SATA 300-8X, and the 3ware 9550SX (the cards SR is using) are all PCI-X, with a minimum of 533MB/sec of bus bandwidth available. Also, the LSI and 3Ware are both full blown hardware RAID cards. Once again, I don't see where you get an argument about SR's results being limited by a poor RAID implementation. Admittedly, they aren't using Areca boards.

However, this does not change the overriding factor here: RAID-0 does not improve buffer implementation or localized seek performance, the dominant factors in desktop productivity and gaming storage performance. Yes running RAID-0 will substantially improve scores in synthetic benchmarks, and is useful for content creation scratch arrays and in disk to disk backup, but running RAID-0 will not significantly improve desktop storage performance. For gamers (remember that this is primarily a gamer board) RAID-0 actually can negatively impact system responsiveness.
 
All I wanna say is...here we go again! More crap studies from SR to fuel DL's hatred of RAID0. Up next will be clips from Anandtechs article "see we told you so",j/k of course. IMO DL, the bashing is getting a little old. Everytime this comes up from you, dj or someone else is going to jump in with all kind of studies that are refuted by yet more studies...aka Tweakers.net among others. Either you believe it or you dont. Nothing is going to change when new hardware arrives. RAID0 is RAID0 nomatter what drives it consists of. Your graph in the first post proves this. Most times, WD740GD 4xRAID0 is faster than WD740GD 2xRAID0, which if course is faster than a single WD740GD. Put two WD1500ADFD in RAID0 and it will be faster overall than a single WD1500ADFD. A loss in two games does not make a conclusion in which RAID0 is dead. As I once said before DL, I respect your knowledge. But I think it foolish to keep on a subject like this. One in which you cannot and will not prove.
 
Is it hatred? Or is it me just wishing that enthusiasts that often have very tight budgets shouldn't get taken by something that is nothing more than hype? For double your money, you get a performance boost of maybe 5%, maybe even a drop in performance, while accepting increased system/driver complexity, reduced reliability, increased heat/noise and greatly complicated disaster recovery. On the other side of the ledger, upgrading to a faster single drive may very well still double your money- but you will still be able to use generic ATA drivers and install Windows without a floppy, you will hold the line on power consumption and reliability, and achieve better performance in the applications you use everyday. Why run RAID-0 on a gaming rig? Show me a good reason why. Please. I'd love to hear one.

If a 6800 GT was only 5% faster than a 6600GT in everything except 3DMark, there would be people all over the Video Cards forums telling you not to buy a 6800GT, and justifiably so. So why should we accept people getting taken with hype and synthetic benchmarks for their storage dollars?
 
Dont get me wrong, most of what your saying I understand. But im telling you that your generalizing everyone as a gamer. You must realize that gaming isnt the only(I dont game anymore-the kids do) thing people in these forums do. Step back and look at the big picture. Look how many time poeple Ghost there HD's, take a look at our Digital Artwork section, might be a couple people there using RAID0, alot of people backing up there CD's by way of Nlite, alot of HD activity there wouldnt you say? You need to look at the grand scheme of things. We arent all gamers, and you dont always take a huge hit in access times.

What im saying it sounds like you want your stance the be all end all of RAID0. And it cant be. You know that. RAID0 isnt "dead", matter of fact it will continue to thrive. Theres a time and a place for RAID0, so whats the big deal if someone wants it all the time "just in case"?
 
We agree. Really. It just doesn't look like it.

Throughout this thread, I've acknowledged the utility of RAID-0 for stuff like disk-to-disk backup and content creation scratch arrays, although admittedly I have to edit that in some times ;)

Yes the thread has a generalized title, and I may yet change it. However, it is my firm belief that I should do anything I can to get the word out that RAID-0 is very much overrated in the enthusiast community, and that an erroneous belief persists among many gamers that running RAID-0 will get them into the BF2 game faster, etc.

My views on this are no more extreme and out of touch than the people who say "RAID-0 FTW!!!1111!!!one!!!eleven" and automatically presume that RAID-0 will improve storage performance across the board. You are quite right that this is not a matter of black and white, and that deploying RAID is a decision that needs ot be carefully evaluated before being committed to. However, after such an objective analysis, most users will find that running RAID-0 doesn't meet their goals.
 
How in the hell did this get stickied. Its like one guy standing with his tongue sticking out razzing the world for no good reason.
 
Im pretty certain we agree. And this is why when I seen the thread I had to ask myself "should I respond or shouldnt I?"..lol. I wanted , for my own sanity(lmao) to see if you really were this fanatical about the end of RAID0. I truthfully believe that most of the arguements concerning this subject, in here at least, would never get to the extent that they do if you, or some of the others who down RAID0, would recognize the benefits of it instead of saying RAID0 is dead, or just hype. There are times im with you wholeheartedly, for example once someone said RAID0 made there downloads faster...lol. If this forum was made up of "just a bunch of kids" id be backing you 100%. But some of these kids know a helluva lot more than I do. And their uses range farther than gaming. And I think we need to recognize that when we explain such things as the subject at hand. Again, your right, we agree ;) Take care.
 
I stuck it. If you disagree with what I have presented, you are welcome to present data to the contrary. Not once in this thread have I used mod powers to silence someone who disagrees with me, or otherwise affect the debate.

So far, noone has presented any data to refute the conclusions that Eugene and I have reached, let alone justify running RAID-0 in light of its drawbacks. Once again, I will not sit idly by and allow the myth that RAID-0 offers an across the board desktop storage performance boost to be perpetuated.
 
daemionsos said:
How in the hell did this get stickied.
Click to expand...
It got stickied because it's a fact, and far too many people rely on a placebo effect to continue to argue that RAID0 is more than hype. What you meant to say it was one guy standing up and holding proof, and expecting others to finally give up the damn argument. This is the modern day version of the argument whether the earth is flat or round. It's been proven round, so why argue differently?
 
Dont sound so hostile dj..lol. ;) But as far as placebo, you gotta realize thats not exactly true. Just take DL's chart, besides the two games, we see a benefit. Might not be worth it to you to see that little of a benefit. But theres always someone out there looking for that last little bit of speed, whether its a CPU or a HD. And from that chart alone you see its not placebo. Just as I said to DL, I say to you...we agree----for the most part :D .
 
I'll agree there are some unique cases where it actually helps out and makes a difference. I just think overall, the difference is very small, the risk of data loss is high, that it doesn't justify doing it. Especially when you consider the most popular thing to do was RAID two Raptors together. When you compare that to a system with one Raptor and one much larger storage drive, you end up with two computers of nearly identical performance, but one has a ton more storage space. I just get very frustrated when I see people still spouting how much faster their games are (loading and playing) because of their shiny new RAID0 array. I have an offline friend (I know, weird) that built a RAID0 array for his file server. When his power went out, he lost thousands of mp3 files. All I could say to him was..."no shit, dumbass".
 
Dumbass really F'ed up didnt he? Not only is he not bright, he must be deaf to, since im almost positive you have been preaching to him how valueable a good backup plan is...right? ;) Putting two Raptors in RAID0 wasnt nearly as popular as grabbing two cheap 80's and running them in RAID0 to try to get Raptor like transfer rates. Actually I still see alot of people doing that. I think thats the point, most people running RAID0 with no "real" use try to get better transfer rates on the cheap. Whether its worth it is up to them.
 
I'm setting up RAID 0 soon and I'm a gamer but I know it won't increase my game performance. Only reason why I am getting it is because I got a free harddrive (WD 80GB SATA150) over the holidays from "older" PC's and I noticed a big difference in apps and windows loading. So i just did it b/c it was a good time :) Besides I think it will kick my current OS HDD ass! I currently use a WD 80GB ATA100 2MB CACHE 7200RPM HDD.

If i had the money I would get one of those new raptors though hehe, 295 @ the egg? :(

I wouldn't say RAID 0 is dead though. How do the new drives perform in raid 0?
 
You're the exact point of this thread. You are a gamer, you experienced the placebo effect in windows, so somehow you think that two slow drives in RAID-0 = one fast drive. Sorry
 
Vertigo Acid said:
You're the exact point of this thread. You are a gamer, you experienced the placebo effect in windows, so somehow you think that two slow drives in RAID-0 = one fast drive. Sorry
Click to expand...
QFT.
 
Vertigo Acid said:
You're the exact point of this thread. You are a gamer, you experienced the placebo effect in windows, so somehow you think that two slow drives in RAID-0 = one fast drive. Sorry
Click to expand...

Let me ask you this...how do YOU know what he is experiencing is placebo? Quick answer you dont. Period. Did you personally test his system? No, you didnt. Maybe he didnt either, but either way you dont know, yet it must be so because?? This is what bugs me in these types of posts. Just because you tested a system doesnt mean someone will see it the same way. Just take a look at SR's
FAQ on RAID0 and scroll down to the chart. In that example, SR shows there is a gain in speed. Is it worth it? Only the end user can say for sure if it is worth it. Is there a placebo? No, there is definately an increase in most cases. It all comes down to what you feel is an acceptable increase. But placebo? Hardly.
 
The gains SR noted were less than 10% in everything except their Bootup DriveMark, and any gains that were made i nthe Bootup DriveMArk were probably erased by having to allow a RAID BIOS to run before the Windows XP Boot loader is even called.

Once again, let's take a look at value for money. If 7800GT SLI was only 10% faster than a single 7800GT...for example, let's say going from 20fps to 22fps at a given IQ level in a game like FEAR, then you can bet there would be people shouting everywhere "SLI is a joke." However, there's a problem here. Running RAID-0 does greatly impact scores in synthetic benchmarks (HDTach, Sandra, etc), so going back to the SLI example - let's say the SLI setup gets 12,000 3DMarks while the single card setup gets 7,000 3DMarks - the SLI setup is better right? Never mind that the return on investment in our example is actually dismal. Of course, we know to measure performance in actual applications (in game frame rates for video cards) like Kyle and Brent do i ntheir video card reviews, yet many [H]'ers buy into synthetic benchmark results that do not account for locality as justifcation for running RAID-0.

If running SLI only offered a 5% performance boost in games, but nearly doubled 3DMark scores, we'd be making fun of anyone who was running SLI "You n00b!! lolz, you spent an extra $50 on your power supply, another $80 on your mobo, and bought another $300 video card for 5,000 3DMarks and a bigger e-penis, what a n00b :rolleyes:"
 
PaHick said:
Let me ask you this...how do YOU know what he is experiencing is placebo? Quick answer you dont. Period. Did you personally test his system? No, you didnt. Maybe he didnt either, but either way you dont know, yet it must be so because?? This is what bugs me in these types of posts. Just because you tested a system doesnt mean someone will see it the same way.
Click to expand...
There is *no* subjectivity in performance differences. Either there is or there isn't a difference. No room for discussion. Your perception of a situation has no effect on reality for the rest of us.
And I'll gladly admit I make sweeping generalization, that apply to 90% of the people who are considering RAID-0 for two slow drives to "make their computer faster". He plainly stated that he was a gamer, not a video-editing professional.
 
My point is dj and Vertigo say theres placebo when clearly there is not. Ultimately it is not up to you or I to mislead end users, only educate them...correct? If there is a gain let it be known, however little. Whether it be 2% or 15%, its not up to you or I to say what is acceptable, only if there is a gain or not. And in all studies/tests RAID0 shows gains. Again clearly not placebo. As SR put it ..

"So what's a "real world" speed increase of typical Windows and Linux applications? It is difficult to put a solid number on this figure, because of the diversity of software out there, but it is reasonable to assume a 0% - 15% overall disk performance increase moving from a single disk drive to two in RAID 0, with rapidly diminishing returns as you add more drives."

There are just too many variables to generalize performance on every machine. And that is when these arguements go haywire. Things such as "RAID0 is dead", and "placebo effect" have no place here. Enough studies have been done, and you can read for a month, but still come up with the same outcome. RAID0 produces gains, mostly across the board. How much of a gain is worth it is up to the end user. Again not you or I.
 
Vertigo Acid said:
Your perception of a situation has no effect on reality for the rest of us.
Click to expand...

And your perception is better because? To make a blanket statement without knowing all variables quite frankly is foolish. Unless you personally know what he uses his pc for OVERALL, you can not suggest anything. Boot time can be faster under RAID0, normal windows apps. can be enhanced also. Again, too many variables to say one way or another.
 
unhappy_mage said:
The only thing missing from those graphs is a single wd740gd. It doesn't make sense to me to have left it out; sure a single new drive is faster, but what about a single old drive? And it's not like you don't have the drive right there...

Wow, I just called the 74g raptor old.

Oh, I guess the other one would be two of the new raptors in raid 0. Right now it looks like apples to oranges. I'm not saying I support raid 0 on the desktop, but this post does anything but kill the myth for me.

Click to expand...

http://www.storagereview.com/articles/200601/WD1500ADFD_9.html

they have some single drive tests there, between all 4 (3.5??) generations of the drive.
 
LoL! All I'm saying is that I'm not setting up raid 0 to get a performace increase in games, I know this already and I'm not expecting anything. I'm only doing it for experience with raid and that it was a good time to do it since I got a free HDD and it is about time to get rid of my old ATA100 HDD that I have had for 4 years or so. I'm just expecting faster boot times and faster app loading times, that is it, that's all I'm expecting from this setup.

So please stop attacking me. I'm not here to argue with anyone I just wanted to say what/why I'm doing raid. Mainly for experience (to learn etc) and I know it will be faster than my current ATA100 HDD.
 
Well, I am only using a couple Hitachi deskstars in my RAID 0 and back when I got it I was on nf2+barton. I thought there was a nice improvement. But, because RAID 0 IS DEAD, I am going to seperate them on my next reformat and see what you are all on about. What is the best configuration for 2 sata drives. Where to put the OS and where to put the page file?

By the way, those words haunted me all night- "RAID 0 IS DEAD". So I thought I better find out for myself when I wake up.
 
And your perception is better because? To make a blanket statement without knowing all variables quite frankly is foolish. Unless you personally know what he uses his pc for OVERALL, you can not suggest anything. Boot time can be faster under RAID0, normal windows apps. can be enhanced also. Again, too many variables to say one way or another.
Click to expand...
I'm not basing it on my personal opinions and/or experiences. I'm basing it on the numbers.
 
PaHick said:
My point is dj and Vertigo say theres placebo when clearly there is not. Ultimately it is not up to you or I to mislead end users, only educate them...correct? If there is a gain let it be known, however little. Whether it be 2% or 15%, its not up to you or I to say what is acceptable, only if there is a gain or not. And in all studies/tests RAID0 shows gains. Again clearly not placebo. As SR put it ..

"So what's a "real world" speed increase of typical Windows and Linux applications? It is difficult to put a solid number on this figure, because of the diversity of software out there, but it is reasonable to assume a 0% - 15% overall disk performance increase moving from a single disk drive to two in RAID 0, with rapidly diminishing returns as you add more drives."

There are just too many variables to generalize performance on every machine. And that is when these arguements go haywire. Things such as "RAID0 is dead", and "placebo effect" have no place here. Enough studies have been done, and you can read for a month, but still come up with the same outcome. RAID0 produces gains, mostly across the board. How much of a gain is worth it is up to the end user. Again not you or I.
Click to expand...
No offense, but this post is a bunch of senseless banter. You're hiding behind some veil of non-information (not a real world, but deal with it) instead of ponying up some reality centered results.

Yes, there are tons of variables, too many to list. Why don't you benchmark your own system with a RAID0 array, since you're so knowledgeable.

For the record, most people would actually evaluate that kind of performance with draconian methods (stopwatch, lol) rather than just come out with some blanket statement of "Oh, there's tons of variables and because neither one of us can predict all the outcome, I'm right and you'r wrong."

Please. Just stop.
 
You must log in or register to reply here.
Back
Top