4x Raptor 150s in RAID 0

J

jebo_4jc

[H]ard|DCer of the Month - April 2011
2FA
Joined
Apr 8, 2005
Messages
14,572
at some site called gamepc

(insert lame benchmark joke here)

*ahem*

"These things totally rock in HDTACH!!!!! YESS! I use that program all the time!!! I'm totally getting 4 of these!!!!!111one1!1"
 
provoko said:
GamePC is such a great site.
Click to expand...
Are you being serious? I've never heard of them.
It looks like a mix between a tech review site and hardware sales site?
 
Looks to me if you're doing any type of High I/O operations, the 74 gig raptor is still the way to go. It's lower latency just allows it to pump ou tmore I/O's per second
 
And when one of those drives fails--Pooof! Kiss all that information goodbye! Not worth the risk and you gain nothing in performance in Games/Apps. Yeah, you're machine will bootup quicker, umm whoopie do? Yeah you can transfer data faster/stream video smoother --whoopie do again. Other than that, where's the benefit? I would Raid 0+1 them or Raid 5 those 4 Raptor 150s.
 
Nn'theraq'pss said:
Why would you want to do that when RAID10 is so much better in terms of speed and recoverability?
Click to expand...


Ah yes, you are correct. My bad.
 
jebo_4jc said:
Are you being serious? I've never heard of them.
It looks like a mix between a tech review site and hardware sales site?
Click to expand...

Uhh are you serious? Read their articles, they go into detail about everything and are THOROUGH! Hehe.

They actually know and explain how raid and raid controllers work. They have tested multiple software and hardware raid devices and versed them together.

They have great articles on video cards. They're so thorough they dissected and went through every test you can imagine for creative's x-fi.

Reviewed a few dual cpu motherboards.

Just because you never heard doesn't make them a bad site:
http://www.gamepc.com/labs/

The proof is in their articles. Oh and I don't know anyone else thats taken 8x WD hard drives (2 terabytes!) and raided them. But damn 4x 150gb raptors. Come on, thats just interesting.
 
provoko said:
Uhh are you serious? Read their articles, they go into detail about everything and are THOROUGH! Hehe.

They actually know and explain how raid and raid controllers work. They have tested multiple software and hardware raid devices and versed them together.

They have great articles on video cards. They're so thorough they dissected and went through every test you can imagine for creative's x-fi.

Reviewed a few dual cpu motherboards.

Just because you never heard doesn't make them a bad site:
http://www.gamepc.com/labs/

The proof is in their articles. Oh and I don't know anyone else thats taken 8x WD hard drives (2 terabytes!) and raided them. But damn 4x 150gb raptors. Come on, thats just interesting.
Click to expand...

/me hands over sarcasm oscar.. wow, you are good!
 
I wish all the raid0 and raptor threads would just die... no one can even post on here anymore without someone trashing it.
 
It is truly unfortunate that such synthetic diagnostic and/or server benchmarks are used to justify running RAID-0. None of those benchmarks account for locality. None of those benchmarks stress the buffer strategy of the disks. Also, the article makes no mention of the drawbacks of running RAID-0.
 
Ockie said:
I wish all the raid0 and raptor threads would just die... no one can even post on here anymore without someone trashing it.
Click to expand...
Maybe we need to introduce the sectioning as it is in the Video Card forum:

RAID-0 belivers
RAID-0 trashers

and put the threads in the respective subforums so that nobody needs to listen to anything that they do not believe in.
 
Ockie said:
I wish all the raid0 and raptor threads would just die... no one can even post on here anymore without someone trashing it.
Click to expand...

thats because they're too [H] and most likely fell into the raid 0 trap themselves at one point and realized that its not even worth it first hand.

I run raid 0 raptors and I could care less what anyone else feels about it. Not their money, we share the same hobby, not interests.

That said I can say myself that its amazing anyone put out the coin for 4 150's thats 1400 dollars for a C drive.
 
provoko said:
They actually know and explain how raid and raid controllers work. They have tested multiple software and hardware raid devices and versed them together.

Oh and I don't know anyone else thats taken 8x WD hard drives (2 terabytes!) and raided them.
Click to expand...
This was, and still is, the worst review of a "serious" raid card (by which I mean hardware raid 5 on pci-x or pci-e) that I've ever read. And knowing how raid and raid controllers work is no great miracle - read through this, it's a lot more understandable than you'd think.

See also: [H]ardforum Terabyte club discussion.

 
provoko said:
Uhh are you serious? Read their articles, they go into detail about everything and are THOROUGH! Hehe.

They actually know and explain how raid and raid controllers work. They have tested multiple software and hardware raid devices and versed them together.

They have great articles on video cards. They're so thorough they dissected and went through every test you can imagine for creative's x-fi.

Reviewed a few dual cpu motherboards.

Just because you never heard doesn't make them a bad site:
http://www.gamepc.com/labs/

The proof is in their articles. Oh and I don't know anyone else thats taken 8x WD hard drives (2 terabytes!) and raided them. But damn 4x 150gb raptors. Come on, thats just interesting.
Click to expand...

Uh, Why didn't they test it on a board that is more IO optimized? How about with a decent raid controller (areca ? adaptec ? someone we've heard of ?)

If the raid controller they used is tied to the pci bus, that is a HUGE bottlenech.
I can get 125MB/second sustained read (winblowz software raid5) on large files on 4 x 400GB 7200rpm drives, but I am using pci-x (NOT Pci-e) with an 8 port controller.

It boggles me that people will go out and spend $400+ on a cpu, then $500+ on disk drives to stick on a $100 motherboard.
Rob
 
i currently have 4 of the 36gb raptors in RAID0. it kicks ass for compression, and data verification. i think it gets around 190MB/s in HDTach, not sure though.
 
provoko said:
Uhh are you serious? Read their articles, they go into detail about everything and are THOROUGH! Hehe.

They actually know and explain how raid and raid controllers work. They have tested multiple software and hardware raid devices and versed them together.

They have great articles on video cards. They're so thorough they dissected and went through every test you can imagine for creative's x-fi.

Reviewed a few dual cpu motherboards.

Just because you never heard doesn't make them a bad site:
http://www.gamepc.com/labs/

The proof is in their articles. Oh and I don't know anyone else thats taken 8x WD hard drives (2 terabytes!) and raided them. But damn 4x 150gb raptors. Come on, thats just interesting.
Click to expand...
I wasn't trying to be critical. I've just never heard of the site.
 
provoko said:
Oh and I don't know anyone else thats taken 8x WD hard drives (2 terabytes!) and raided them. But damn 4x 150gb raptors. Come on, thats just interesting.
Click to expand...

This close enough? http://www.hardforum.com/showthread.php?t=1000787

30x 250GB Maxtor Maxline II's 6.36TB Formatted. Was hooked to a slow machine (PIII) so I didn't run any benchmarks, would have been very pci bus limited.
 
Robstar said:
Uh, Why didn't they test it on a board that is more IO optimized? How about with a decent raid controller (areca ? adaptec ? someone we've heard of ?)

If the raid controller they used is tied to the pci bus, that is a HUGE bottlenech.
I can get 125MB/second sustained read (winblowz software raid5) on large files on 4 x 400GB 7200rpm drives, but I am using pci-x (NOT Pci-e) with an 8 port controller.

It boggles me that people will go out and spend $400+ on a cpu, then $500+ on disk drives to stick on a $100 motherboard.
Rob
Click to expand...

Did you read the review? This is taken from the first page, end of the last paragraph: "Not only did we test with the fastest Serial ATA drives on the market, we also utilized the fastest RAID card on the market today as well, Areca's ARC-1220 PCI Express x8 controller, to grab every bit of performance possible from these new drives. Let's get those platters spinning and those bits flying."
 
Does RAID-0 always have such a marked improvement in workstation Iometer results?
 
IOMeter is a fantastic server benchmark. The ability to adjust queue depths, read/write percentages, transfer sizes, etc means that it can simulate just about any server access pattern imaginable. However, IOMeter simply cannot account for locality. An IOMeter test works across the entire physical area of the disk, while desktop applications work within a very small area of the disk. Even an enormous UT2004 installation that takes up 6GB is a scant 4% of a WD1500ADFD, and seeks will be made over roughly 5%, maybe as much as 10% if fragmentation is servere, of the disk's area. All of the HDD makers tune the buffers and firmware on their drives to make the most of this phenomenon, which is one of the reasons why the WD1500ADFD trails SCSI drives in server performance, even at the lower queue depths where command queuing is less of a factor. However, in single user tests, these firmware tweaks make an enormous difference. The WD1500ADFD turns the tables on the desktop, dominating 10K, and even edging out 15K, SCSI drives in desktop performance. The vast majority of desktop access has 80% of the seeks over 20% or less of the disk. The bigger the drive, the more intense the effects of locality. Simply doubling the capacity of the drive and changing nothing else boosts throughput in SR's tests around 7-8% - ie, moving from a 120GB drive with 3 heads to a 250GB drive from the same family with six heads, as the larger drive means seeks are made over an even smaller area and the buffer becomes even more important.

There is not a single synthetic benchmark out there that accounts for locality. HDTach, Sandra, ATTO, IOMeter; they all have no provision for locality in their measurements, thusly they do not reflect the work that HDD makers put into making their buffers and firmware as smart as possible.

Making a desktop hard drive purchasing decision based on HDTach, IOMeter, or Sandra results is no better than buying a video card based on 3DMark or fill-rate benchmark performance. None of these tests actually measure application level performance, they are mere diagnostic benchmarks.
 
So what benchmark should someone rely on? SR's "DriveMark?" All I see when looking at these drive reviews are benchmarks.
 
There is no good desktop HDD benchmark that is available to the public :(

All of the synthetic ones don't account for locality, and the only one (that I'm aware of) that allows sophisticated modeling of actual desktop access patterns, the IPEAK SPT, has been discontinued. Anandtech, Tweakers.net, and StorageReview are the only sites that I'm aware of that have access to it.

IPEAK allows a capture to be made of all of the disk accesses for a given application, then it can replay those accesses on whatever drive you like. Since it only replays the disk calls from the storage driver down, the hardware the test is run on only influences the recording of the trace. When an IPEAK trace is played back, the CPU/RAM/Video card don't affect the results, as only the requests to the disk are played back, not the calculations that are used to formulate those requests.

HDTach is a great diagnostic tool, and can alert you to mechanical problems with the drive or issues with your chipset drivers. However, it doesn't measure actual desktop application performance :(

Really, it's a lot like 3DMark with Video Cards. Your score in 3DMark doesn't tell you much about how your system will perform in games, but it can tell you if your score is unusally low that you may have a driver problem. To measure actual game performance, you have to play the games and decide what is playable, just like the [H] does in their video card reviews.
 
You must log in or register to reply here.
Back
Top