View Full Version : drive performance variables

07-29-2002, 12:09 PM
OK, here is the situation:

On an ATA/133 PCI card, what will give you best performance:

An ATA/100 120 gb 2mb cache 7,200 rpm drive, or

An ATA/133 120 gb 2mb cache 5,400 rpm drive, or

An ATA/100 120 gb 8mb cache 7,200 rpm drive (a la WD Special Edition model)

I am wondering how one particular feature of a drive compares to others? Do you get more bang for your buck with the rotational speed, the cache or the fact that the drive is ATA133 v. ATA100?



07-29-2002, 02:23 PM
ATA/133 drives. Probably 8MB cache. There is now a 16MB cache drive for portables (2.5" form factor ATA). I think Serial-ATA drives with 8MB cache are the future.

have you looked at the drive upgrade reports and seached over on www.xlr8yourmac.com (http://www.xlr8yourmac.com) ?

07-29-2002, 03:37 PM

?Now that's a question that could take a book to answer.

Rotational speed first:
?I would say there is no financial reason to purchase a 5400 drive today. The faster drives do most everything faster and costs are so close that I don't think you even consider the slower drives.

Cache size:
?Cache plays a huge role in small file size operations. In OSX the system is making constant reads and writes to the system disk and cache will play a bigger part in performance there than in a file system where you mainly store large files.

?I would say that for storage drive performance vs. dollars the IBM 120GXP is the current through-put leader even over the WD Spec Edition. There are some tests where the larger cache will blow away the smaller cache of the IBM but most of the testing shows comparable results.

?I an running Barracuda IVs for my main drives. I give up a little performance and gain peace of mind from the better press the Seagate's receive. I also have a couple IBM 60GXP from before the 120's were available.

ATA100 vs ATA133:
?I don't think there is a lick of difference between so called ATA100 and ATA133 drives. ATA133 is a bus specification and not an indicator of individual device performance. New ATA133 drives 'MAY' reach peak reads and writes (burst rates) that exceed ATA100 drive performance capabilities but you will be hard pressed to find that performance in real world use.

?Just like Gregory stated Seagate has announced a new serial ATA drive spec that should be showing up in the winter. Until then I don't think you determine your drive purchase by its ATA133 or ATA100 rating.


07-29-2002, 03:55 PM
Here is a comparison http://www.barefeats.com/hard24.html k

07-29-2002, 03:59 PM
I just purchased a pair of IBM 120GXP 60GB ($72 ea) and was really impressed with
how much better they do in a FW case than even the 60GXP was.

If I do get a IDE card, it would be ATA/133 for sure, probably Acard for $89 (non-RAID).

- Gregory

07-29-2002, 06:05 PM
Gregory said If I do get a IDE card, it would be ATA/133 for sure, probably Acard for $89 (non-RAID).

I wholeheartedly agree. I did not address the host card issue and my answer did not state that very good. There is no reason, at least cost and performance reason, that a buyer should even look at less than the highest specification on host cards.

Thanks Greg.


07-29-2002, 10:39 PM
As usual, a plethora of informative respones!

I kinda figured that there was no need to get 5400 rpm drive. I was curious if cache is all that it is cracked up to be. 8 mb is a lot of cache! I wasn't sure about the difference between 133 v. 100... I kinda figured it had to do with the max burst r/w rates.

So really, the key is to get the fastest rotational speed, then step up to 8 mb cache if it is offered and you can afford it.

I'd be running this all off an ATA133 raid card... I figured there wasn't much point in asking if it was an ATA66 card as presumably any 100 or 133 drive would "power down" to the ATA66 rate.

The general cost seems to be minor from 5400 to 7200 and ATA100 to ATA133... but the 8 mb cache WD adds about 50 bucks... or so it seems.

Thanks for the tips.

As a follow up... if I wanted to do an ATA mirror RAID with a 120 gb drive to a striped 60 gb RAID.... is that possible? I believe so... I think Rick told me it was a RAID 0 + 1 a couple of months ago. Do I need to get a Seagate 120 if I have two Seagate 60s on the striped RAID? Can I mirror a 133 drive to two ATA 100s?

I am trying to put all my MP3s on the 60 stripe for speed on my LAN, but I want to back them up to the 120, because I plan on selling all my CDs once they are on the drives. If I have catastrophic drive failure, I am up the proverbial processor pipeline without a... well, ya know! http://macgurus.com/infopop/emoticons/icon_smile.gif



07-30-2002, 12:19 AM
?So if you accomplish your mirror with one side being a let's say a 2 drive 75MB/sec RAID0 and the other side being a single 38MB/sec drive on another bus you will get all the drives performing at best 38MB/sec.

?The only way I see to get your RAID 0+1 at speed seems to be to run two separate Hardware RAID card and four drives. Then software can run the mirror.

?I have yet to find a way to 0+1 for OX except with that method. Instead I run as fast as I can for my scratch disk space, run standard singled up drives for my OS storage and backup automatically every night to another drive, or another computer with a sync application like SwitchBack.


07-30-2002, 05:48 AM
I see... so under current OS X conditions, a RAID 0 + 1 is not possible taking advantage of the full striped bandwith. Essentially, you might as well have two unRAIDed drives and backup one to other because you get the same amount of throughput.

So my set-up would be a RAID 0 with a back up to another drive. I put the 120 gb RAID (actually 110 or so formatted) on one bus of the ATA 133 card, and put the 120 drive on the other bus with my system drive (the original 40 gb) and just back up to the 120 each night. Eventually, there may be a software solution to doing the RAID 0 + 1 at speed (SoftRaid X perhaps?). Sorry for getting off into a RAID post in the IDE forum!

Thanks again!


07-30-2002, 10:54 AM

?Well, it really doesn't have anything to do with OSX. If you connect your High Speed RAID0 array to anything in a RAID1 mirror the mirror is going to operate at the speed of the slowest component. In the case of hanging a single drive on one side and your RAID on the other the single drive will determine the speed of your new RAID configuration. You just hung a big anchor on the striped array.

?The only way that I have figured out to build your internal 0+1 is with two matching striped hardware arrays. That way the speed is equal.

?The hope is that SoftRAID for X will have 0+1 capability. It may not since it didn't for OS9. There are downsides to even wanting a software 0+1. CPU overhead. And security. If I NEED data redundancy I NEED RAID5 or at least Hardware RAID 0+1. Dumping the parity building chores off on a controller would seem to be smart.

?Don't let RAID1 give you a false sense of security either. RAID1 solves ONE possible failure: A drive failure. That's just about the very least likely thing that can happen to your data. That's not to say they don't happen, entropy rules.

?There is no replacement for a good backup plan. The problem with RAID1 is that you contaminate your mirror with most any kind of failure where the failure starts at the computer. Heck, the simplest is the 'erase' command. Oooops. Wish I'd done a backup last night.


[This message has been edited by ricks (edited 30 July 2002).]