10000 RPM SATA vs 7200 RPM SATA vs 7200 RPM PATA

jrfuda

Active Member
Guys, I'm considering changin my automation server's system drive from a single IDE/PATA drive to two SATA drives in a RAID1 configuration - becuase I need the redundancy RAID1 gives becuase I don't like downtime, and I may get deployed for up to a year next summer (leaving the wife alone with the PC), so I really need the reliability.

Anyway, how much of a true perfomance gain is there in going from a 7200 rpm PATA drive to a 7200 RPM SATA drive, or to a 10000RPM SATA drive.

Can any of those drives even approach the specs for the interfaces (100 or 133MB/sec for PATA and 150MB/sec for SATA). In actual use, without resorting to one of the RAID striping methods, is there going to be any real performance increase in going from the lowest priced PATA drive to the highest price (per MB) SATA drive.

The drives are system drives, so size is not important.

1. I can get 40MB 7200RPM SATA drives with 2MB buffers for $40/each
2. I can get the same with 8MB buffers for $55/each
3. I can get 36.7GB 10000RPM SATA drives with 8MB buffers for $85/each refurbbed and $110/each new.

The warranty on option 2 and the non-refurbed option 3 are 3-5 years (I think 1 year for option 1).

Also, in a RAID1 setup, if one HDD goes kaput, will the array work off the 1 good drive, without intervention, until the bad one is replaced?
 
Hi John,
I am building a new desktop this weekend and am faced with same decisions.
I am looking at a WD Raptor 74 gig 10,000 rpm. www.outpost.com has a sale / $30 rebate on them, but are out of stock till end of Nov. (I called yesterday).

RAID 1 will save your butt if one drive croaks. You will continue to run until both drives fail. You can put a replacement drive in a rebuild the array and be back with redundant hardware. This has saved me a few times already on my HS server. I am undecided if I will do same for this new desktop, or just regularily Ghost the drive to a standard drive.

If you use a backup software, be careful because most don't like RAID. I found out the hard way. My experience was with Ghost.
 
If this is for an automation server, you may just want to get a regular drive and get larger ones and do raid 1 or equivalent.

The case I've seen raptor's being used is with high end gaming rigs primarily as they approach scsi performance without the price or requirement on additional cards, etc.

In this case two raptors in raid 0 are used for the gaming areas, and larger drives used for other scenarios.

I'd say save your money if it truly is an automation server.

For the price of two refurbed raptors, you can buy 2 250GB maxtor drives with 16mb cache and use them in a raid configuration.

I guess the question is: where and why do you see the need to extreme performance coming in? That might give further insight on where things might help.
 
Just ordered two 74gig SATA raptors for Raid1 for my desktop via NewEgg. Read a good review for it and said what the heck.
 
I use a Raptor..

It's a bit loud.. but a slight bit faster.
I will get a SATA2 drive next.
 
I've decided on two 7200 RPM SATA drives in RAID 1.

I can get 80GB models for a little over $50 each. I don't need that much capacity, but they appeared to be the best balance of price, performance, and quality. They're WD models with a 3 year warranty and come with the SATA cables.

I'm also going to up my system's RAM from 512MB to 1GB, as I discussed with several of y'all over in HS land.

I got my drives, RAM, and shipping for less than $200.

If I had gone the Raptor route, it would have been at least $140 more.

I'll let y'all know how the transitoin from the one PATA to two SATA RAID1 drives go... if I ever have the time to do it ;)
 
I just go my two 74 gig Raptors via NewEgg. Always happy with NewEgg prices and shipping.

I am installing of of the Raptors as my desktop boot drive. 40 gig forWindows, the rest for Program Files (helps to keep the boot drive defragged) and Data.

I plan on "regularily" ghosting the drive to a IDE large drive to minimize data loss in case of Raptor failure. The Raptor is suppose to have a longer MTBF than a standard drive.

So far, Windows is installing in record time. Whir from drive is not bad at all. Fans are much louder (gotta do something about that...)

When my new desktop is done, I will use the other Raptor in my Server rebuild.
 
I'm considering building a new top end gaming machine after the 1st of the year (tax reasons). I want to use 2 raptors in a raid 0. What backup software will handle a raid 0 backup but allow restoring to a none raid or non-restore access ala acronis ?
 
Raid for speed is a joke...not worth the time, expense or the decrease in MTBF.

Stick with a single Raptor, or get a sata2 drive.
 
Skibum said:
Raid for speed is a joke...not worth the time, expense or the decrease in MTBF.
I will be assuming you have a disk system that allows for concurrent transfers. The MTBF comparison is made with respect to a single drive. I'm ignoring Raid-2 and Raid-4. Then:

Raid-0 does not normally increase speed at all. All it really does is increase capacity. It does indeed decrease MTBF.

Raid-1 allows twice the throughput on reads, but half the throughput on writes. You may see a marginal increase in total throughput, since reads outnumber writes. It dramatically increases MTBF, because of the full redundancy at the drive end.

Raid-3 allows for a significant increase in throughput for large contiguous files, like HD-Video. You can (and I have) achieve a total throughput equal to N-1 times the throughput of a single drive (where N is the total number of drives). The MTBF is increased where a degradation in performance can be tolerated, and decreased where a degradation cannot be tolerated. Raid-3 is only used in special circumstances, since most file systems do not adequately support large contiguous files.

Raid-5 allows for a significant increase in throughput for large contiguous transfers, such as occurs in large database systems. Again, you can achieve a total throughput equal to N-1 times the throughput of a single drive (where N is the total number of drives), but the file system rarely allows large enough transfers for that to happen. As in Raid-3, the MTBF is increased where a degradation in performance can be tolerated, and decreased where a degradation cannot be tolerated.

Note that, with Raid-3 and Raid-5, throughput can peak at N-1 times the throughput of a single drive (where N is the total number of drives). So a three-drive Raid-5 will only have a 2x peak. You really need a wide array to get a significant performance increase.

Note also that any redundant system increases MTBF, when a single failure can be tolerated. This is normally the case with Raid-5, and is not normally the case with Raid-3, where it's high throughput does not survive a single drive failure.
 
Raid-0 does not normally increase speed at all. All it really does is increase capacity. It does indeed decrease MTBF.

I thought it did provide a speed increase as, when implemented the way I've seen it, it uses two SATA pipes rather than one. Since it sees both drives as one unit it can get a faster throughput.

That is, if it properly uses both drives (meaning that the data is essentially evenly split so both pipes can be fully utilized at a time (simplified a bit however).

I thought there were only two reasons for RAID 0:
1. See multiple drives as a single storage device
2. Speed

Does it really not provide any speed increase?
 
I went back to check, and Mike is right. A Raid-0 CAN increase throughput if the drives are 'striped'.

There are two ways to implement Raid-0: concatenate the drives and striping the drives. When you concatenate the drives (the second drive starts where the first drive ends) a file access will normally be limited to a single drive, and you will see no speed improvement.

But when you stripe the drives, you distribute the data back and forth on both drives evenly. An example of a very fine-grained, two-drive stripe would be to put all even-numbered blocks on drive 0, and all odd-numbered blocks on drive 1. Any file access will be evenly distributed across both drives, increasing performance.

Of course, as Ski pointed out, you have doubled the chance of failure, and a drive failure will take out all of the data.
 
also... you dont nearly double your speed with striped drives. I think the improvement is between 30 and 40%

Forgetting numbers and theories. I've been there. I tried it. I've seen the speeds. I had a single drive failure. I lost all my data.

I now use a single raptor and am happy. Next will be sata 2.

Striping is best left to candy canes.
 
I just converted my DVR PC to a RAID 0 configuration with 2 250GB Seagate SATA drives. I was getting some audio choppiness when all 3 TV tuners were recording at the same time and Snapstream's support pages recommended the Raid 0 to fix this.

I hope it was worth the hassle (of figuring out how to do it and rebuilding the PC from scratch). I'll be testing it with BTV 4 tonight. My surveillance cameras also run on this PC.
 
I have used Raid 1 or Raid 5 HARDWARE controllers on all my pc's for years and any pc's I have sold or recommended to people.

There have been many times a drive has failed and all I had to do was put a new drive in and wait until it finished rebuilding.

Raid 1 mirroring is pretty cheap, the hardware controllers are pretty cheap and you only need a second drive. And some motherboards have the raid 1 built in.

To me and most of my clients the time it would cost me to rebuild a machine is just not worth it. Getting a Hardware raid controller is the best way to go.

Software Raid is a wast of money and time. It is slow and difficult to recover from failures.

Just my 2 cents.

StevenE
 
Back
Top