RAID question

[MichaelD]

Originally posted by: Confused
I noticed a BIG difference when going from a single, 8MB cache, 7200rpm drive, to a RAID 0 with another 7200rpm, 2mb cache added too (not the same drives, I know, but I had them around).

The loading times for BF1942 dropped from about 30-40 seconds to about 15-25 seconds. That's a HUGE drop in loading times, and gives you an advantage when playing multiplayer because you'll spawn before most others and have a chance at getting the good vehicles.

Of course, as HendrixFan said, you also need enough RAM for all the data to be loaded into, otherwise there's really no point. For BF1942, and especially the Desert Combat mod, you'll need at LEAST 512MB, and preferably 1GB of RAM.


Confused

Exactly. Excellent reply, Confused. I have seen pretty much the same results as well. Load times are dramatically reduced, and that is what I want.

Also, as has been said, if you have RAID0'd two drives and found almost no improvement, you probably didn't install correctly. Stripe size has a huge effect on performance.

 

[Pariah]

I would say based on this snippet along with their other arguments that they are assuming RAID 0 is not worth 2x the cost plus a RAID controller,

You can selectively read to make any argument look good.

and completely ignore the notion that someone may be considering buying a 2nd drive because they need more disk capacity

More like you completely ignored where they did:

"Suffice it to say that with two drives, when properly configured, RAID 0 will offer double capacity and sequential transfer rate offered by the standalone drive"

They also don't mention some of the other benefits and uses of a RAID array and controller other than multimedia editing.

Once again, ignoring what you don't want to see:

"Again, RAID 0 does have its advantages in a handful of key applications and uses where data files are huge and/or data requests are highly sequential in nature."

The rest of your 1st paragraph is accusing SR of assumptions that had nothing to do with what they were trying to show. I have no idea where you are going rest of your post. The rest of us are talking about quarter mile times and you are off arguing about trunk space. Can we stick to the original topic?

 

[chizow]

You can selectively read to make any argument look good.

• RAID will give you practically 0% performance increase in gaming:
  
  Scroll down a bit
  
  SR Gaming DriveMark 2002 519 IO/sec 529 IO/sec
  
  The first number is a single drive the second is RAID. That's nearly a whopping 2% gain in performance. Is RAID really worth the hassle for that kind of improvement?

Yep, and you can selectively quote numbers from flawed and antiquated tests that don't translate into current real world performance, no matter how wrong or inaccurate the point you're trying to make. You're right though, there's no difference when I load up Solitaire on my machine at work compared to my RAID'd rig at home. They're both lightning fast.

More like you completely ignored where they did:

"Suffice it to say that with two drives, when properly configured, RAID 0 will offer double capacity and sequential transfer rate offered by the standalone drive"

No I didn't ignore it, but that statement only gives emphasis to their flawed argument that RAID costs 2x as much for less than 2x performance. Why else would they say it doubles the capacity of a single drive unless you are comparing 2 to 1? Once again, if you had the choice of an 80GB HDD or 2 x 40GB HDDs for the same price, which would you choose? SR's entire write-up gives arguments against RAID based on the assumption that spending 2x as much for a HDD + controller will *NOT* give you 2x the performance across the board.

They also don't mention some of the other benefits and uses of a RAID array and controller other than multimedia editing.

Once again, ignoring what you don't want to see:

"Again, RAID 0 does have its advantages in a handful of key applications and uses where data files are huge and/or data requests are highly sequential in nature."

Yet none of their "benchmarks" specifically address these "key applications", but instead focus on such pedestrian tasks as office apps and bootup times along with 4 year old games.

I was actually referring to the flexibility a RAID card offers in terms of additional IDE drives, which will be increasingly beneficial as optical drive variations, low HDD costs, and SATA becoming mainstream put a premium on available headers.

The rest of your 1st paragraph is accusing SR of assumptions that had nothing to do with what they were trying to show. I have no idea where you are going rest of your post. The rest of us are talking about quarter mile times and you are off arguing about trunk space. Can we stick to the original topic?

What were they trying to show? And why did you even bother to reference it? Aren't we talking about gaming here? Specifically BF1942 and Morrowind as indicated in the OPs post? Do you own either game? RAID lately? Sounds like you need to put that link to SR *back* in the trunk........


Chiz

 

[CZroe]

Originally posted by: Pariah
I still don't see what "expensive RAID" has to do with anything.

Nearly every high-end RAID company now creates ATA RAID cards. Almost invariably, they refer to it as a "Cost-Effective ATA RAID Card." Read Adaptec's box. You can't separate "Cost-Effective" from "ATA RAID" anywhere in their marketing documentation! 15k RPM SCSI and the SATA Raptor are still considered high-end and as such, so are their RAID-3+ capable RAID controllers.

How is buying a RAID controller and second drive for 1.9% performance increase cost effective?

Don't pretend those drives on $400-to-800 controllers are alternatives to ATA RAID. Listing them as the "single" drives which outperformed the DM9 array was a mistake because it totally blew the point of saying that you shouldn't use the DM9 array (As if you should purchase them instead!) Most people with that kind of money are spending even more to RAID those or else they wouldn't bother! IT's an entirely different class and has nothing to do with the cost-effectiveness of ATA RAID.

Originally posted by: Boogak

I do alot of RAR uncompression with large (700+ MB) size files and I could not tell a difference between the RAID config and the single drive setup I had previously.

Are we forgetting the real bottleneck here? Compression takes CPU cycles, LOTS of 'em. The larger the file, the more numbers it has to crunch (It has to go through every bit of it). Regardless of file-size, your CPU should be the bottleneck. Check Task Manager's Performance tab.

 

[Pariah]

Don't pretend those drives on $400-to-800 controllers are alternatives to ATA RAID. Listing them as the "single" drives which outperformed the DM9 array was a mistake because it totally blew the point of saying that you shouldn't use the DM9 array (As if you should purchase them instead!) Most people with that kind of money are spending even more to RAID those or else they wouldn't bother! IT's an entirely different class and has nothing to do with the cost-effectiveness of ATA RAID.

I don't understand why you keep going back to highend RAID. Who is talking about SCSI RAID? Assuming that every person that buys SCSI for a home system puts them in RAID arrays is completely inaccurate. The point of quoting the other drives performance had nothing to do with actually comparing them, but just to show that the benchmark SR was using was a HD benchmark in a gaming context, not a gaming benchmark, and gives widely varying results based on the drive, unlike a QuakeIII fps benchmark which would give 0 difference regardless of the drive. You could buy a new IBM 180GXP and get a bigger performance improvement than buying a second Maxtor and RAID card. If you already own all the necessary components, by all means use it, there's no reason not to, but specifically going out to buy a drive and RAID card for the intentions of improved gaming is ridiculous.

 

[AgaBoogaBoo]

Listen benchmarks don't say everything because they will show which one can perform the large tasks the fastest. What about small everyday things? Loading Winamp3 has a hard drive bottleneck for me, and it takes about 8 or 9 seconds to load, which is more than a media player should take.

Also, you'll will get faster bootup speeds because lots of the botting process is reading from hard drives, if you monitor what the bottleneck in your boot up is, very likely it will be the hard drive.

A Raid 0 setup will increase performance, as stated earlier 40-50%, it may not be proven in every benchmark, but in real life it is faster.

 

[VTrider]

I know this topic is covered all the time, but here is my 'real world' .02c.......

I just installed my first RAID 0 array (2x120gb WD 8mb cache) for the primary purpose of making my video editing process more efficient. Not only did it help significantly, I have noticed a postive performance increase in my general computing. Now forget about benchmarks and all that hoopla, i'm talking about 'noticeable' difference. Many programs load quicker, the OS loads quicker and in general my whole system seems more vibrant. I say forget what I just said, and forget was everybody else says and go try it for yourself. I got all caught up in the pros/cons of RAID 0 and it drove me nuts. I decided to try for myself and so far it has been the right choice and works for me, i'm better off than I was before.

-VTrider

 

[Pariah]

Listen benchmarks don't say everything because they will show which one can perform the large tasks the fastest.

The benchmarks that SR runs are not standalone benchmarks like Winbench. The tests that SR use record actual computer use.

if you monitor what the bottleneck in your boot up is, very likely it will be the hard drive.

And you would be absolutely correct. SR runs a bootup test which records exactly what happens when a system boots.

SR Bootup DriveMark 2002 288 IO/sec 474 IO/sec

1st is non-RAID, 2nd is RAID. 66% performance increase bears out exactly what you said that the HD is bottleneck when booting.

A Raid 0 setup will increase performance, as stated earlier 40-50%, it may not be proven in every benchmark, but in real life it is faster.

This assumption is not correct, and the other DriveMarks bear that out as well. Yes, there are applications where RAID will give big performance boosts like media editing. But anyone doing that already knows that and doesn't need a benchmark to prove it. For typical computing, you will see no where near 40-50% performance increase once you are in Windows.

 

[MichaelD]

Originally posted by: VTrider
I know this topic is covered all the time, but here is my 'real world' .02c.......

I just installed my first RAID 0 array (2x120gb WD 8mb cache) for the primary purpose of making my video editing process more efficient. Not only did it help significantly, I have noticed a postive performance increase in my general computing. Now forget about benchmarks and all that hoopla, i'm talking about 'noticeable' difference. Many programs load quicker, the OS loads quicker and in general my whole system seems more vibrant. I say forget what I just said, and forget was everybody else says and go try it for yourself. I got all caught up in the pros/cons of RAID 0 and it drove me nuts. I decided to try for myself and so far it has been the right choice and works for me, i'm better off than I was before.

-VTrider

*ding* Give the man a cigar. A real-world testimonial...I wanted to write ^^that^^ but I was at work earlier and couldn't.

I won't get involved in that debate, above...too much statistical BS for me. RAID0, on a controller card, whether it's software (Highpoint RocketRaid) or hardware (3Ware 7000-xx) is going to stomp all over a single IDE drive in read times...and usually write times as well. I say this from seat of the pants experience, like VTrider, above. *stamp* MichaelD Approved!

 

[Boogak]

Originally posted by: Boogak

I do alot of RAR uncompression with large (700+ MB) size files and I could not tell a difference between the RAID config and the single drive setup I had previously.

Are we forgetting the real bottleneck here? Compression takes CPU cycles, LOTS of 'em. The larger the file, the more numbers it has to crunch (It has to go through every bit of it). Regardless of file-size, your CPU should be the bottleneck. Check Task Manager's Performance tab.

I'm checking it now while unRAR'ing a 762mb file and it spiked to 42% once, the rest of the time it hovered between 20% to 35%. So what is the real bottleneck?

Now I'm not saying I could be wrong and that RAID does spank the llama's ass, so I'm curious to hear what other people are using for their stripe size. I had set mine to 16kb after reading another thread where someone said it was the optimal size. But then I read in storagereview's forums that the optimal stripe size depends heavily on the specific hardware being used and what the usage pattern is. Maybe I had the wrong stripe size?

 

[XBoxLPU]

I am really enjoying my SCSI setup. Seek times are below 5ms, the I/O is high and my system is not dragged down in intensive situations. Yes a SCSI setup will set you back around $300 ( Atlas Maxor IV and LSI Controller ) but I do think the extra cash will give you more of a wow factor than going from ide to ide raid 0.