• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

velociraptor: to raid or not to raid?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Sorry about the book -- killing time. ;) Hope this helps.
That's a nice post, really, but I don't need to see another HD Tune/Tach graph which doesn't correlate with real application use. :eek: Comparing two PC's with the same hardware isn't an equal comparison and a veteran PC builder should know this. Install the same OS twice on the same computer and tell me if the boot times are within 1 second of each other, which I found to typically be the difference between a single drive and a RAID0 setup. Lots of people on ICHxR chipsets and I'll post my great RAID0 improvements later.........
 
I've read somewhere that seek times increases through raiding so the transfer of small files will increase in time. Perhaps it's tusken's own thread.

"Lots of people on ICHxR chipsets and I'll post my great RAID0 improvements later" what do you mean?
 
Tusken isn't dissing RAID0, he runs it himself, he doesn't like it when short-stroked Matrix (or other) RAID gets hyped because in theory an equal-sized partition would be the same as short-stroking, it just isn't measurable because all drive performance apps can only measure drives not partitions of drives.

Yeah, I totally get this: people who hype things they're either doing wrong or don't understand irk me too, and I won't pretend I fully understand it either, but like anyone who's been around for a long time, I know enough to confidently call "bull****" when I see it. ;)

Enough, in fact, to point out that that the above noted theory is is wrong; just partitioning isn't the same and it is measurable. To Tusken: Yeah, in large part I agree (and already stated up front) synthetics aren't the end-all be-all -- they're just indicators of what you can expect. YMMV, as they say. They're far from yawn-worthy though, and I think you both missed something by dismissing my screeny out-of hand: Both of you take a closer look at the access times on my RAID. If short-stroking didn't work and I hadn't been successful in hiding the rest of the hard drive's space from the OS/benchmark app, it would be in the 14-15ms range, and my transfer diagram would trend down like the Velociraptor's does.

The proof is in the pudding. Consider this: Access time = seek time + rotational latency. For all 3.5" 7200RPM drives, rotational latency is fixed at ~4.16milliseconds (sometimes rounded to 4.2ms). That means the seek time on my RAID volume is a mere 3.64ms. Given that the seek time for each individual drive is about 8.9ms (also a pretty fixed figure in this class, btw), 3.64ms is attributable to RAID overhead and testing just the barest sliver on the outer edge of my drives' platters: 107GB of 1.25TB of potentially usable capacity. I believe the RAID controller is masking the non-RAID-volume-assigned space.

Further, consider that access time is somewhat misleading in that drives have been able to read files out-of order for years now: once the heads are over the right track, they can start reading, picking up the earlier parts of the file on the next pass. The Velociraptor's platters are smaller and spin faster with a rotational latency of 3ms, meaning it's seek is 4ms and thus my RAID's seek time, even with the added burden of RAID calculation overhead and head-synching is faster by ~.46ms. When you factor in how out-of-order reads reduces the relevance of rotational latency, this short-stroked (for real) RAID is probably about equal in effective access time to the Velociraptor. It then proceeds to demolish it in terms of STR. A set of short-stroked Velociraptors in RAID would probably turn in some fairly evil performance...

Anyway, you can poo-poo synthetics all you like, but they're more than just marketing tools -- and in the case of hard drives, a lot closer to painting a true picture of real-world performance than, say, 3dmark is at giving a true picture of how a vid card will perform in games.

Finally, dismissing the comparison just because Anandtech's base hardware was different than mine isn't valid. SATA controllers turn in some fairly standardized performance numbers -- they aren't bottlenecked by systems, they do the bottlenecking. ;) There is more variety amongst the plethora of possible RAID controller and hard drive pairings, but my argument doesn't really live or die by all raid controllers being as good as what ever functionality Intel crammed into my southbridge (though I think it safe to presume better controllers do exist...) either.
 
I have my drives in RAID-0, and I'm pretty sure I didn't just do it for ****s and giggles...

For Windows boot time, I can't "officially" comment because no, I did not use a stop-watch to time the start-up before and after the upgrade. I can however hint that usually I don't have time to take a bathroom break while my OS loads anymore.

As for games, again, I did not measure them with a stop-watch. I used my own common sense and logic to realize that (when I used to play) Star Wars Galaxies seemed to load twice as fast after I got the second drive. Was it actually twice as fast? Probably not, but it felt like a bigger jump than going from 512MB of RAM to 1GB.
 
Last edited:
Well I'm not going to make it all pretty as I did before on my NF4/AMD platform:

http://www.wideopenwest.com/~vacationdave/raiddata.JPG
http://www.wideopenwest.com/~vacationdave/aa.JPG
http://www.wideopenwest.com/~vacationdave/boot.JPG
http://www.wideopenwest.com/~vacationdave/obliv.JPG
http://www.wideopenwest.com/~vacationdave/photoshop.JPG

Same OS image, each test run once, system rebooted to clear drive cache. Time is an average of four tests timed with stopwatch. Drives are 36GB Raptors:

Boot:

RAID0: 20.45 sec
Single: 24.69 sec

America's Army Launch to Main Screen:

RAID0: 41.29 sec
Single: 41.78 sec

Oblivion Save Level Load:

RAID0: 7.88 sec
Single: 7.79 sec

Zip 105MB file(used timer within program):

RAID0: 31 sec
Single: 31 sec

Move Zipped file from HDD to thumb drive:

RAID0: 19.05 sec
Single: 19.09 sec

Launch Adobe Photoshop Elements Editor:

RAID0: 7.16 sec
Single: 8.01 sec

Not great overall results for RAID0 for me as I expected, except for boot up which was a pleasant suprise. I have no problem eating crow on at, but I'm not certain I was able to accurately begin timing in the same place with the difference of the controller running in RAID and then AHCI mode. As noted, I use RAID0 now and have used it many times before and never took the word of others to predict how it will work out for me. For those that seek advice, I offer my truthful experience backed by testing. The subject of RAID performance is too contentious to speak blindly and I'm sure there are plenty of apps I don't use that benefit from RAID0. And remember, theory doesn't trump real experience. :beer:
 
Right, and those are the kinds of numbers people use to say RAID is over-hyped and/or doesn't do anything.

The problem being it's an incomplete experiment that ignores a lot of factors like not all controllers/drive pairings being equal (YMMV), and intentionally optimizing file location on disk for sequential reading based on usage patterns -- this is something Vista is supposed to do over time and should work heavily in an array's favor. To detect that effect, you'd have to compare a system that's been used for a while and a fresh install...it wouldn't be detectable by just imaging over. When have we seen that tested for? Never. I'd guess because Vista was plauged by bug-induced crap file transfer speeds until just recently, it takes a long time and is a pain in the *** to separate the differences made by hard drive optimizations from whatever they're calling "superfetch" (aka, why Vista always seems to be using 50% of your RAM at startup) now.

About your file transfer tests -- unless you're moving between RAIDs, no, of course you're not going to see an increase...the RAID is going to saturate the STR of the lesser drive. But, try doing some power-user **** like watching an HD .mkv file off your RAID drive, while someone on your LAN watches another vid, and then backup your music collection from the RAID to a lone drive...then try to do that juggling act with a lone drive -- Your vids will stutter, and the transfer will slow to a crawl. Now I'm not doing that kind of crap ALL THE TIME (but a lot more than I would now that I CAN), but then people don't need dual video cards with an average of 200fps in UT3 all the time either...they only need (ok...want) it that .05% of the time things are really busy and that "average" fps hit's the "low" and they're the only guy in the frag-fest not watching a slide-show. Pwnage and gleeful laughter follow.

And that's the kind of thing not related to directly measurable numbers your average RAID user (who, right away, you have to remember is far from average) will forget to cite when he asserts his RAID "feels faster" than a single-drive -- because he's used to it and doesn't even think about such things. I know there are other examples I can't think of... It's a lot like asking someone why they believe in gravity...it's such a given, that even if you were inclined to think about it, you would forget half the reasons you do, and most people are content to supply something like: "..Well, I pushed my brother at the top of the stairs one time, and he, you know...fell."

Also, not all games respond the same way -- and programs are doing other other crap (initializing this and that...) not necessarily related to, nor dependent on HD performance while loading. Consequently, you only see an increase in performance proportional to a.) how much better a RAID is over a given single drive in a given area (STR, access time) and then only b.) to the extent that factor "a.)" is even involved in whatever is going on.

To belabor the point and continue the amusing (to me) analogy, gravity might not play the central role in the holding together an atomic nucleus, or in attracting chicks to my beer belly, but it sure is important when I'm pushing my brother down the stairs...
 
Also, Tusken,

If you want to continue playing devils advocate, try a couple of the things I mentioned earlier in the thread: regular every-day little things like opening a directory with 12GB of 630MB vid files and seeing how long it takes to generate thumbnails for them. Try opening a 6GB photoshop file...while running bit torrent, running a virus scan, and creating a drive image.

Something I should have driven home earlier is that there is this overly prevalent notion of posting numbers from "real world apps benchmarks" which is flawed in that the testers so often try really, really hard to make these measurements "pure"; to isolate them from other influences. That's fine and dandy if it's PART of a larger data set, and presumably more science-y (science-ish?) sounding, but you should take the words "real world" right the **** out of the description because I don't know anyone not responsible for predicting the weather who computes in a bubble running just one app. How are such "real world" app benchmarking suits so different from any other "synthetic" measurement when you're running it and ONLY it alongside the (maybe 5?) other processes Windows absolutely needs to run? -- I've got four cores and four hard drives and consider it a moral imperative that they be occupied...if only to justify the time, effort and money I've lavished on my box without feeling entirely embarrassed by the UV-lit, LED-encrusted, black monolith sitting proudly atop my home theater equipment tower...<sigh> at least I retain enough self-respect and perspective to have avoided giving it a girls name or something...not one I'd share with anyone...I digress...

And, before it's said, no, RAID, even properly short-stroked RAID is not THE killer mod for every game which was the emphasis the OP had. But, we're not console 'tards...we do other things with our computers where a nice RAID-0 can matter a lot more. Let's be honest here: you're not wanking it to the poorly-rendered orc avatar controlled by your World-of-Warcraft girlfriend; hell no! The WoW client's in the background while you're watching 20-second WMV clips of naked Japanese school-girls beating each other with wiffle-bats, hoping the questionable website(s?) you're on isn't introducing new viruses your AV scanner can't detect through as-yet-un-patched holes in your firewall program the new version of which your ftp client is still downloading, because your bit-torrent client finally found some high-bandwidth peers for the full 5-seasons of <insert random pirated anime> which is causing the naked japanese girl clips to stutter, and then a all-finished DING!!! from either dvd shrink or autogk goes off really loud because you turned off dynamic range compression in your audio-synth console making it frustraingly difficult to cum before the suspicion that your iGF whom you suspect won't get on a web cam because her voice on teamspeak sounds suspiciously like an obese 45-year-old house-wife or a dude with a voice modulator >>that sick ******* would be who I play in this long-winded scenario, btw<< recording your voice with in-game footage via Frapzbecause it'd be just so-damned funny to convert in Premiere and post to YouTube...causes your "run on sentence" to go limp. I think I'm thoroughly done amazing myself with the cleverawesomtude of my writing...if anyone's actually read this, I suspect they get the point: it's a multitasking world and RAID STR bandwidth is your friend. :)

Finally, try short-stroking your RAID from within the controller BIOS where you assign drives and create the volume, NOT through the OS and I bet you'll see an effect better than just doing RAID...Not Jesus giving you a back-rub better, but...better.
 
Last edited:
Finally, try short-stroking your RAID from within the controller BIOS where you assign drives and create the volume, NOT through the OS and I bet you'll see an effect better than just doing RAID...Not Jesus giving you a back-rub better, but...better.
I've done the short-stroking thing and have tested it. I makes not one bit of difference for me because I use a third party disk defragmenter that puts all the files in order from the outside of the platters in. Whether I have a 1GB or 100GB before I hit a brick wall behind the data doesn't matter. You post lots of good information and there are variables for each user, but if someone says I've got a gaming machine and will RAID0 make things faster the answer is still "not much" based on my experience. Yes, one second here or there is what can mean the difference between the best and worst in many things and ultimately is up to the user whether the cost is worth it.
 
So you compared the two with the exact same OS image to make a proper comparison? What testing process did you use? Did you document your times to share? Please, please tell me it's not just your "impression".

Uhmm, easy there slugger. Yes, I used the same OS image, it was just a fresh install. My testing process was counting the number of times the little green bar went across the screen during XP's boot process. It went across 5-6 times on a single drive, 3-4 times on the RAID array.

Games I can tell load faster because when I switched to RAID, I started being among the first people in the round to load new maps where before I was somewhere in the middle of the pack.

Scientific? No. But it's faster, and if you insist on evidence, go test it yourself ;)
 
just going back to the OP, and a few posts after. weren't we comparing 2 raptors in RAID0 vs. ONE VELOCIRAPTOR?

(I don't know much about the topic myself, but judging solely on responses), we have determined that 2 drives in RAID0 are faster (to SOME extent), than a single drive (same manufacturer and model ofcourse), but what I'd like to know is if 2x WD raptors are faster than one single velociraptor.

If the difference between 2 vs 1 drive isn't "much", and the velociraptor is certainly faster than the previous raptor, wouldn't the velociraptor be a better choice for the OP? (wouldn't it be cheaper as well)

edit: faster applies to the OP's computer uses, i.e. windows boot, gaming, and whatever else it is that he does
 
If the difference between 2 vs 1 drive isn't "much", and the velociraptor is certainly faster than the previous raptor, wouldn't the velociraptor be a better choice for the OP? (wouldn't it be cheaper as well)

edit: faster applies to the OP's computer uses, i.e. windows boot, gaming, and whatever else it is that he does

The problem with wanting a one-line response is that, because it's a complex topic, you have to ask very specific questions; this is hard to do if you don't know all the crap Tusken and I were going on about. Otherwise, you have to deal with a half-assed answer that doesn't really answer the question you think you want a simple answer to. :eh?: The ***** of it is that with almost everything else computer related every answer to a performance-related question that starts with "yes" or "no" can be followed by "if" or "but." I guess I'm sort of an old-school elitist in that I don't think you should be dicking around with things you don't care to understand.

Knowing, though, that sometimes n00bs just want a simple answer to complex questions and they'll learn the rest as they go because they don't want to spend 3-months in "forum-school" before building their new rig, I'll try to give short responses to what I think are the right n00b questions to ask...

Would it be ****ing awesome to RAID-0 a pair of Velociraptors?

To the extent that anything you can do to/with your PC is "awesome," yes. In addition to automatic bragging rights, it will add length and girth... if only to your 3p33n.

Is RAID a good thing for a first-time PC builder to mess around with?

Probably not, but it depends on your capacity to absorb frustration. RAID adds another layer of potential troubleshooting complexity. It is, simply put, another thing that can go wrong. If you're not easily frightened (but especially if you are) google "trouble with RAID."

Is RAID worthwhile if I'm interested in playing with something new?

Absolutely. I wish I'd messed around with it sooner. It's almost a PC enthusiast right of passage. Hell, Tusken doesn't think it works and he uses it. :D

Is RAID-0 good for gaming?

Yes, but not especially great, and occasionally, slightly bad. One thing Tusken's (and many review sites') numbers are good for is indicating that a lot of the wait time related to level and game loading has to do with crap other than your hard drive. It's counter-intuitive, but true.

Will RAID-0 Make My Computer "Faster?"

Generally yes. Keep your expectations reasonable, though. As I said so very eloquently above: the performance improvement provided by using RAID-0 for the hard drive subsystem of your PC is proportional to a.) how much better the RAID's various performance attributes (STR, access time, etc.) are than the single drive you'd otherwise use, and b.) how dependent the thing you want your PC to do faster is dependent on those hard drive attributes.

What it will pretty much always do is make your computer more responsive in multi-tasking situations. Command Queuing + big transfer "bandwidth." If you're doing several things simultaneously that are HD-intensive, RAID will help you out big-time.

Is a dual Velociraptor RAID-0 "worth it?"

Beauty is in the eye of the beholder.
 
Last edited:
These are really interesting posts. However, the OP indicated that he was looking at RAID0 for gaming purposes, which is generally one of the last reasons to consider RAID0 as it doesn't have enough positive gaming effect to be worth the cost of a second $300.00 drive. And yes I've used RAID0 before.

Honestly, for the level of improvement it would make, I would literally put that $300.00 in savings instead and use it for a new video card next year (or when your current one no longer maxes what you want it to).
 
Codhein, love the posts man! :thup: Great reading.

If it hasn't already been said, :welcome:

I've just set up a RAID10 (ICH9R) on 4 of the 640 AAKS drives. 200GB XP-32 partition, 200GB Vista-64 partition, 820GB data partition. Running great so far!

I was considering doing Matrix RAID0/5 for a while, but I've just read of too many RAID5 problems on ICH9R.

If I had unlimited funds I'd RAID0 6 v-raptors!

:beer:
 
I saw tusken's replies coming from a mile away ;) in fact I checked this thread just to read them.

These are really interesting posts. However, the OP indicated that he was looking at RAID0 for gaming purposes, which is generally one of the last reasons to consider RAID0 as it doesn't have enough positive gaming effect to be worth the cost of a second $300.00 drive. And yes I've used RAID0 before.

Honestly, for the level of improvement it would make, I would literally put that $300.00 in savings instead and use it for a new video card next year (or when your current one no longer maxes what you want it to).

Damit Shadin! You read my mind.:D
 
The problem with wanting a one-line response is that, because it's a complex topic, you have to ask very specific questions; this is hard to do if you don't know all the crap Tusken and I were going on about. Otherwise, you have to deal with a half-assed answer that doesn't really answer the question you think you want a simple answer to. :eh?: The ***** of it is that with almost everything else computer related every answer to a performance-related question that starts with "yes" or "no" can be followed by "if" or "but." I guess I'm sort of an old-school elitist in that I don't think you should be dicking around with things you don't care to understand.

Knowing, though, that sometimes n00bs just want a simple answer to complex questions and they'll learn the rest as they go because they don't want to spend 3-months in "forum-school" before building their new rig, I'll try to give short responses to what I think are the right n00b questions to ask...

I appreciate your response, but I felt this part of your post wasn't so great...

I'm not sure how much more specific I could get than that, I was simply going back to what I thought the original poster was asking about. So rather than ask a completely new question myself, I asked what I thought the OP wanted out of this thread with his provided details.

Secondly, the conversation between you and tuskenraider was about 1 drive vs. 2 of the same drive, (different from what I was asking 2 "traditional raptors" vs. 1 velociraptor), So I don't understand where the "Knowing, though, that sometimes n00bs just want a simple answer to complex questions and they'll learn the rest as they go because they don't want to spend 3-months in "forum-school" before building their new rig" came from

Lastly, I did not take offense to what you wrote, but it came across unnecessarily condescending.
 
I will be purchasing all new parts very soon and I am reading on alot of sites that raid 0 isn't all that much faster than a single drive for gaming purposes. I don't do any kind of video editting or whatnot but I do want the fastest performance I can get. Should I go with a single 300GB drive or should I get a second one and raid 0 them?

I appreciate your response, but I felt this part of your post wasn't so great...

I'm not sure how much more specific I could get than that, I was simply going back to what I thought the original poster was asking about. So rather than ask a completely new question myself, I asked what I thought the OP wanted out of this thread with his provided details.

Secondly, the conversation between you and tuskenraider was about 1 drive vs. 2 of the same drive, (different from what I was asking 2 "traditional raptors" vs. 1 velociraptor), So I don't understand where the "Knowing, though, that sometimes n00bs just want a simple answer to complex questions and they'll learn the rest as they go because they don't want to spend 3-months in "forum-school" before building their new rig" came from

Lastly, I did not take offense to what you wrote, but it came across unnecessarily condescending.

It's pretty clear the original post is asking whether it's worth it to buy 2 of the same drive and RAID them compared to just 1 of that same drive.
 
It's pretty clear the original post is asking whether it's worth it to buy 2 of the same drive and RAID them compared to just 1 of that same drive.

Fine then, it's my question with the OPs details/circumstances. Didn't think i had to repeat them when i posted my second post in this thread
 
Back