WD VelociRaptor

[fadetoblack]

i was wondering if it's worth going from 2 150GB raptors in RAID 0 to a single velociraptor? would there be a noticeable improvement?

[jas420221]

Ummm no...b/c you arent RAIDing the drive. It would be significantly slower.

[tiro_uspsss]

u'd lose outright MB/s, but gain access time..

[NeedMoMegaHurtZ]

u'd lose outright MB/s, but gain access time..

I think you mean lower access time :D .... as in improvement

So it depends on the application and what you are doing.

[tiro_uspsss]

I think you mean lower access time :D .... as in improvement

So it depends on the application and what you are doing.

yes ur right, sorry, been awake since 3am :p: its 2245 now :D

[jas420221]

Why would access time be faster? Its the same drives going across the same data....?

[tiro_uspsss]

Why would access time be faster? Its the same drives going across the same data....?

the VR has faster access time than RX.. plus RAID always adds a wee bit to access time

[Speederlander]

Why would access time be faster? Its the same drives going across the same data....?

RAID has worse access time, including RAID 0. And the more drives you add the more it chips away.

[NapalmV5]

raid does not increase access time :)

[fadetoblack]

thanks guys thats what i thought.. now if i could get two velociraptors that would be sweet.. but i guess im not really in need of more space yet. ill prob wait until ssd prices drop and capacities get larger.

[jas420221]

raid does not increase access time :)see...thats what i though....:up:

[m^2]

raid does not increase access time :)

It does. Because instead of rotational latency of single drive, you got max of all drives in the array. If I calculate correctly, with 2 drives you get 18 1/3% higher latency.

[NapalmV5]

lol just what do you guys base your raid statements on..

anyways.. like i said.. raid does not increase access time.. lousy controllers may do


as for 2x 150GB vs 1x velorap

2x 150GB = about 160MB/s
1x 300GB = 120MB/s by spec

depends on what you do..

[jcool]

What do YOU base your raid statements on Napalm? lol...
Of course raid increases access times. In a perfect, purely theoretical scenario, when using the exact same disks in a raid configuration, you get exactly the same access time with x drives as with a single one. Right?
Well forget about that, that's why it's called theory. There are no 2 drives truely identical, same as 2 CPUs are ever identical.
Of course it doesn't really matter with only 2 drives, the increase in access time will be 0,1ms at most, so this is not an issue.

What's an issue is the generally lower access time and higher IOP yield of the velociraptor. Being a 2,5" drive it has some advantages, and considering a transfer rate of over 100MB/s will make no difference in 95% of all usage situations, a Velociraptor will most likely be faster than two old Raptors.

I ordered one by the way, I'm interested to see how it fares against my Savvio 15k.1
But I'm expecting similar performance, since the Savvios aren't really tuned for single user performance, whereas the Velociraptor is aimed at the desktop enthusiast market. We'll see.

[NapalmV5]

i dont base my statements on theories or calculations per m^2

everything i say on raid is from my own testing/use of raid

once again.. raid does not increase access time :)

[m^2]

http://www.tomshardware.com/reviews/RAID-SCALING-CHARTS,1635-9.html
http://www.tomshardware.com/reviews/RAID-SCALING-CHARTS,1662-5.html

[jas420221]

It does. Because instead of rotational latency of single drive, you got max of all drives in the array. If I calculate correctly, with 2 drives you get 18 1/3% higher latency.YOu should read your own links... (or is it too early in the mornign and I am misreading them?)

While a single HD321KJ has an average access time of over 14 ms, we achieved shorter access times due to the fact the only 80 GB out of the available 320 GB was used. However, protocol overhead and head positioning time were clearly significant, as the access time increases with the drive count. A maximum of 30 ms cannot be called quick anymore,but I/O performance isn’t affected by the poor average access time


.....the average access time clearly suffered from the high drive count

According to that first link m^2, RAID0 DECREASED it with 2 drives, and was the same with 3 as compared to one drive.

Your second link shows arrays that likely nobody here uses. WTF is raid6? Raid5 are fro servers. So you may be correct overall, but for the majority of people here who have 2 drive raid0 or 2 drive raid 1, it DECREASES from teh original 14ms.

THanks for posting that up!

[Serra]

^ Crash & Burn. Nothings worse than trying to throw something back in someones face and getting called on it, but I'll give it a shot and hope no-one does the same to me (in fairness to me, I'm just reading your quote, not the article, so it may happen) :)

The link you quoted shows that they short-stroked the drives. So yeah, average access time will decrease, but not because it was RAIDed... I could do the same with partitioning on a single drive.


But anyway, yes - RAID-0 *does* offer higher access times *on average*, regardless of controller. It must by all the laws of physics. Lets say you have drive A and drive B. If they were completely identical, there would be change... but realistically, there is no such thing as two fully identical hard drives. Thus, if one hard drive takes 0.000001ms longer to access a file or 0.1ms longer (or more), the total access time will be the time of the fastest drive PLUS the differential between the fastest drive and the slowest. Controllers have nothing to do with it. Is it likely to be by much? Not likely. But the fact is your access times will be bottlenecked by your slowest drive, end of story.

[jas420221]

I only want the facts out...and that guy's links essentially proved the opposite for MOST consumers. Obviously on a commercial level with different RAID types, and a lot more drives, he is correct. But in the context of this thread, his links proved himself wrong.

Looking forward to your tests Serra!

[m^2]

YOu should read your own links... (or is it too early in the mornign and I am misreading them?)




According to that first link m^2, RAID0 DECREASED it with 2 drives, and was the same with 3 as compared to one drive.

Your second link shows arrays that likely nobody here uses. WTF is raid6? Raid5 are fro servers. So you may be correct overall, but for the majority of people here who have 2 drive raid0 or 2 drive raid 1, it DECREASES from teh original 14ms.

THanks for posting that up!

As Serra pointed, this particular gain was because of short-stroking drives. Means nothing. Graphs show clearly that more drives=higher access time.

[jas420221]

As Serra pointed, this particular gain was because of short-stroking drives. Means nothing. Graphs show clearly that more drives=higher access time.Whats the point in this thread? It doesnt answer the guys question...The OP was swtiching from 2 RAID0 drives, to 1 drive. If anyone notices the negligable increase in access time from 2 drives in RAID0 to one drive NOT in RAID, I would be impressed and also call shens. Mostly what people see are faster load times in OS/apps/games/large files etc in a raid0 config.

This OT digression may be best in its own thread. :up:

Also, may I mention that though it *may* scale linearly, the drives tested in those links were not 10k raptors, so those numbers just dont apply (but I see the theory does).

[NapalmV5]

^ Crash & Burn. Nothings worse than trying to throw something back in someones face and getting called on it, but I'll give it a shot and hope no-one does the same to me (in fairness to me, I'm just reading your quote, not the article, so it may happen) :)

The link you quoted shows that they short-stroked the drives. So yeah, average access time will decrease, but not because it was RAIDed... I could do the same with partitioning on a single drive.


But anyway, yes - RAID-0 *does* offer higher access times *on average*, regardless of controller. It must by all the laws of physics. Lets say you have drive A and drive B. If they were completely identical, there would be change... but realistically, there is no such thing as two fully identical hard drives. Thus, if one hard drive takes 0.000001ms longer to access a file or 0.1ms longer (or more), the total access time will be the time of the fastest drive PLUS the differential between the fastest drive and the slowest. Controllers have nothing to do with it. Is it likely to be by much? Not likely. But the fact is your access times will be bottlenecked by your slowest drive, end of story.

we may not agree on everything.. i dont expect you/dont have to be on my side.. but at least be consistent


As Serra pointed, this particular gain was because of short-stroking drives. Means nothing. Graphs show clearly that more drives=higher access time.

by raid does not increase access time i do not mean "short-stroking"


Whats the point in this thread? It doesnt answer the guys question...The OP was swtiching from 2 RAID0 drives, to 1 drive. If anyone notices the negligable increase in access time from 2 drives in RAID0 to one drive NOT in RAID, I would be impressed and also call shens. Mostly what people see are faster load times in OS/apps/games/large files etc in a raid0 config.

This OT digression may be best in its own thread. :up:

Also, may I mention that though it *may* scale linearly, the drives tested in those links were not 10k raptors, so those numbers just dont apply (but I see the theory does).


http://www.xtremesystems.org/forums/showthread.php?t=185677

check results posted by buckeye

post # 1 - 2x ssd raid 0
post # 31 - 8x ssd raid 0

if we go by the statements/calculations/theories/laws of physics in this thread.. there should be crystal clear access time increment even @ ssd

while thats ssd raid.. i was referring to hdd raid aswell

hdd raid does depend on the controller for no access increase.. depending on the power/limitations of the controller after a number of drives access time does increase.. yet that increase is very negligible

why do hdd manufacturers incorporate raid specific tech @ enterprise level hdds? to combat access/latency increment

yet again.. raid does not increase access time

:)

[tiro_uspsss]

[B]why do hdd manufacturers incorporate raid specific tech @ enterprise level hdds? to combat access/latency increment

total utter BS

its to get higher IOPS :stick: :slapass: :down:

[Serra]

we may not agree on everything.. i dont expect you/dont have to be on my side.. but at least be consistent


Where was I inconsistent? :confused:

[Talonman]

i was wondering if it's worth going from 2 150GB raptors in RAID 0 to a single velociraptor? would there be a noticeable improvement?

No ...

The speed of a single VelociRapto:
http://www.desktopreview.com/shared/picture.asp?f=321
Max transfer rate: 123.9 MB/sec

The entire review:
http://www.desktopreview.com/default.asp?newsID=334

The speed of a two disk RAID 0 array using normal Raptors.
Max transfer rate: 158.8 MB/sec
http://img360.imageshack.us/img360/2301/run23fx9.jpg (http://imageshack.us)

I am trying to cutt a deal with Western Digital right now, and I would be going to 4 VelociRapto's in RAID 0.
Wish me luck! ;)

[Buckeye]

http://www.xtremesystems.org/forums/showthread.php?t=185677

check results posted by buckeye

post # 1 - 2x ssd raid 0
post # 31 - 8x ssd raid 0

if we go by the statements/calculations/theories/laws of physics in this thread.. there should be crystal clear access time increment even @ ssd

while thats ssd raid.. i was referring to hdd raid aswell


:)

I wish it lowered access times, my Raid would be something like 0.0125ms which would be Smoking Hot !!!

But it did not change it one bit, no matter how many drives I put into the Raid.



I am trying to cutt a deal with Western Digital right now, and I would be going to 4 VelociRapto's in RAID 0.
Wish me luck! ;)

Good Luck !!!
I was planning on doing that with my left over 4 SATA connectors :)

I am still trying to figure out how to tweak my setup for max bandwidth. I should be getting something like 960mb/s but I am only getting ~874mb/s. Something about Vista 64 and my hardware. Oh well I guess I can live with it but I would prefer to push it. Going to have to talk with Areca.

[Talonman]

Thanks...

The only thing that bothers me at all, is that the RAID 0 graph's seem to have allot of spikes:

http://www.maxishine.com.au/documents/wd_velocitaptors.html

Still, the data is moving faster than standard Raptors. I would like to see a better driver made for the Velociraptors.

[specofdust]

Whats the point in this thread? It doesnt answer the guys question...The OP was swtiching from 2 RAID0 drives, to 1 drive. If anyone notices the negligable increase in access time from 2 drives in RAID0 to one drive NOT in RAID, I would be impressed and also call shens. Mostly what people see are faster load times in OS/apps/games/large files etc in a raid0 config.

Actually if you refer to these two articles from anandtech:

http://www.anandtech.com/storage/showdoc.aspx?i=2101&p=10

http://www.anandtech.com/storage/showdoc.aspx?i=2974&p=5

You can see that RAID 0 makes a negligable difference to games. I would also doubt OS load times increasing, given that you've got the RAID BIOS to get through. Large files and some apps though? For sure.

[jas420221]

Actually if you refer to these two articles from anandtech:

http://www.anandtech.com/storage/showdoc.aspx?i=2101&p=10

http://www.anandtech.com/storage/showdoc.aspx?i=2974&p=5

You can see that RAID 0 makes a negligable difference to games. I would also doubt OS load times increasing, given that you've got the RAID BIOS to get through. Large files and some apps though? For sure.
My OS load times DECREASED significantly (i think I said/inferred that) as well as some of my games (BF2, FEAR to name 2 I played). I wonder if its gaem specific, b/c I know for a fact that it decreased low times.




Mostly what people see are faster load times in OS/apps/games/large files etc in a raid0 config.

[cantankerous]

Thanks...

The only thing that bothers me at all, is that the RAID 0 graph's seem to have allot of spikes:

http://www.maxishine.com.au/documents/wd_velocitaptors.html

Still, the data is moving faster than standard Raptors. I would like to see a better driver made for the Velociraptors.

I had these spikes real bad when working on my dads new system with two of these in RAID 0 last week. Things were so bad that I actually couldn't use the drives in RAID and ultimately had to run them each singly. Whenever I put the drives under heavy stress such as benchmarking, defragging etc the entire system would lock, drives would drop from the array etc. I noticed when benching with HD Tune that this would happen whenever those crazy drops appeared. Running each drive through the WD diagnostics twice each had them come up error free. Running the drives on their own and out of RAID also has them running just fine. No clue what was going on but it nearly brought me to tears in frustration over trying to solve the problems over 3 days. I evidently gave up for now. I understand these drives were delayed from release because of the firmware issues originally in the engineering samples. I really don't think the firmware is quite there yet still. I am going to wait for the 150GB models, hopefully with a more mature and fine tuned firmware before I pick up these drives for myself.

I saw that disabling NCQ helped the drive immensely, however, how does one go about disabling NCQ on an Intel chipset through Matrix Manager as the option isn't listed like it is on the Nvidia based chipset. Perhaps this could help sort my issue.

[Buckeye]

Actually if you refer to these two articles from anandtech:

http://www.anandtech.com/storage/showdoc.aspx?i=2101&p=10

http://www.anandtech.com/storage/showdoc.aspx?i=2974&p=5

You can see that RAID 0 makes a negligable difference to games. I would also doubt OS load times increasing, given that you've got the RAID BIOS to get through. Large files and some apps though? For sure.

hmm... well for what it is worth.

Perhaps all Raids are not created equal, but with my Raid load times are almost instant, perhaps a few seconds, and about the only thing that slows it down is network traffic in MMORPG’s, the only one so far that I have tried is AoC and it loads very, very fast.

You may not notice huge performance increase with high bandwidth Raid 0’s when using HD’s in gaming and I believe that is mainly due to access time. With SSD’s you take that out of the picture with 0.1 m/s access times, then the bandwidth pumps the data. But the access time makes a Huge dif.

However, all that said, if you don’t have the CPU/GPU’s to back up the Raid then game performance will still stink, it will just stink faster.

VRaptors are very nice drives and will make a very nice Raid. But compare that to even 2x SSD’s in Raid 0 and access times will blow them out of the water.

[specofdust]

Well sorry to say it guys, but I've learned a lot about RAID 0 and frankly I just don't believe anecdotal claims of "way faster game loading times" and similar (Happy to look at hard figures always). At a guess I'd say the impression of faster load times is down to a mixture of people often buying new drives to compose their RAID 0 (Which are generally faster than older drives due to increased platter density and general tech improvements), the RAID often being empty so the data goes on a faster part of the disk than data on an old half full disk would, and the rest? Placebo.

As game sizes increase you may see slight benefits in game loading times, but game loads are generally not large contiguous files, normally it consists of finding a huge number of smaller files. Smaller files need great access times (so you're looking at a Raptor or an SSD if you can afford it). But think about it ,discounting propriatory archives which are generally composed of numerous smaller files incorporated into a single archive, e.g. .pak, individual game files tend not to be very large. Bandwidth just isn't relevent in most game loading times (there are proven exceptions, baulders gate is one).

OS's tend to be similar, OS's don't tend to be composed of a small number of large files, but rather a very high number of small files (average size for files in my windows directory is 220KB, there are 12,000 files). Again here what you want is low access times. RAID can actually increase overall boot times simply because the RAID BIOS takes a few seconds to get going, so even if you save a second or two thanks to the RAID you're seeing little net difference either way. Like I said though, certain apps, and large file transfer, sure it makes a huge difference.

[jas420221]

Well sorry to say it guys, but I've learned a lot about RAID 0 and frankly I just don't believe anecdotal claims of "way faster game loading times" and similar (Happy to look at hard figures always). At a guess I'd say the impression of faster load times is down to a mixture of people often buying new drives to compose their RAID 0 (Which are generally faster than older drives due to increased platter density and general tech improvements), the RAID often being empty so the data goes on a faster part of the disk than data on an old half full disk would, and the rest? Placebo.

As game sizes increase you may see slight benefits in game loading times, but game loads are generally not large contiguous files, normally it consists of finding a huge number of smaller files. Smaller files need great access times (so you're looking at a Raptor or an SSD if you can afford it). But think about it ,discounting propriatory archives which are generally composed of numerous smaller files incorporated into a single archive, e.g. .pak, individual game files tend not to be very large. Bandwidth just isn't relevent in most game loading times (there are proven exceptions, baulders gate is one).

OS's tend to be similar, OS's don't tend to be composed of a small number of large files, but rather a very high number of small files (average size for files in my windows directory is 220KB, there are 12,000 files). Again here what you want is low access times. RAID can actually increase overall boot times simply because the RAID BIOS takes a few seconds to get going, so even if you save a second or two thanks to the RAID you're seeing little net difference either way. Like I said though, certain apps, and large file transfer, sure it makes a huge difference.Well, if you like reading facts and numbers, then you should ONLY be saying that for THOSE GAMES (1 or 2) in the article you linked, the load times were similar.

What im telling you happened. Happened. In fact I went back from my raid and went to one Raptor and the difference in what I stated is astounding (game loads, and OS load). I do have some hard numbers, though they are in my head...Windows XP load time went from 59 seconds, to 42 (non raid to raid). I do remember that number.

[specofdust]

Well if you don't think games like: WoW, BF2, CoH, SupCom, Farcry, and UT2k4 are representative of games in general then I'd welcome any further direct comparisons with numbers. I've read several other articles on the matter which have me more than convinced, I can't remember where though so obviously that's fair enough if you don't want to take my word for it on all games. It should be fairly obvious from the range of games tested by Anandtech though that the majority of games aren't going to see any difference whatsoever.

Huge volumes of bandwidth just aren't particularly useful for increased game loads - this I know from personal experience because I run a large RAID 5 which has had high performance levels over the last few years, and I'm as good gaming off a modern high capacity fast disk as I am off it.

[jas420221]

I dont have a raid array currently to test and prove/disprove. But I can tell you beyond a shadow of a doubt that BF2 loaded MUCH MUCH quicker. Oh well. Thanks for the info I suppose.

[Anemic]

If you guys havent read this you might find it interesting.
http://www.overclockers.com/articles1297/

[specofdust]

Well, as someone with a RAID - that article seems like a load of crap to me. In some of the tests he did he was getting more than a 100% increase in performance. That's not possible simply because of the drives. At a guess I'd say he must have been doing something very very wrong with his single Raptor. Sure, some of the performance increases he reported make sense (large installations, large file transfers or processes) - but some just don't - and given that and Anand (not to mention my own experience) I'm inclined to maintain my assertion that RAID 0 simply isn't worth it for games, improved OS performance, and a good deal of apps. The more than doubled chance of data loss doesn't make up for the generally small or negligable gains most gamers will experience.

I realise that there are plenty of people who won't care though, who refuse to believe that access time is what's important in most games and apps, and that will absolutely refuse to believe that the extra raptor they bought so they could build a RAID 0 was a waste of money.

[jcool]

This is my new VelociRaptor on the Intel ICH9R if anyone still cares:

http://database.he-computer.de/Bilder/VR.jpg

Quite impressive. Idle noise was subjectively louder than my 15k Savvio (which sits in a Scythe Quiet drive however), access was rather smooth.
Sadly the VR died after just 30 minutes of operation, think I got a prototype lol. First WD ever to RMA for me :down:

[cantankerous]

Weird, im hearing a lot about these dying drives. Newegg has a good few complains about receiving defective/DOA drives.

[Budwise]

cant wait for a 150Gb version.

[Soulburner]

cant wait for a 150Gb version.
I agree, however price/gb will be worse.

I'd expect $209-$220 at intro and eventually dropping to $199 in a couple months.

[fadetoblack]

here's some benches i found on hardforum in case ppl are still interested.

http://www.hardforum.com/showthread.php?t=1313337&highlight=velociraptor

[Eugene08]

Head on over to my posts in Benchmarking

http://www.xtremesystems.org/forums/showthread.php?t=190665