Review of First 10K IDE Drive

Sivar writes "StorageReview has a review of the first 10,000 RPM IDE hard drive. Despite the speed that other technologies are improving, this is the first rotational speed increase in almost six years for standard IDE drives." The review is pretty thorough, but also warns to keep in mind that the reviewed unit is only beta hardware.

Comments

Finally...

[Anonymous Coward]

Nice to know they are finally starting to speed up the slowest part of the computer again.

It would be nice

[T5]

if the manufacturers of these 10K SATA drives would offer two different sets of firmware - one optimized for locality access for desktops and another for the more scatter/gather usage patterns seen on servers. How WD et.al. will position this drive for production remains to be seen.

Does that really help?

[Anonymous Coward]

But hasn't there been several articles around that show hard drive RPM to be a minimal factor in the performance of HDDs?

5400 -> 7200 wasn't that advantagous, but will 7200 -> 10000 be that much better?

Don't we get better performance improvements from tweaks to the file system and how it writes and spaces out its blocks and cylinders?? Or are we at those limits already?

An important paragraph...

[La Temperanza]

When Western Digital raised the bar nearly 1.5 years ago, we repeatedly pointed out that the Special Edition (JB series) Caviar was what readers really wanted when they speculated over 10,000 RPM ATA drives.

Equipped with an 8-megabyte buffer and accompanying firmware aggressively tuned for single-user scenarios, the WD1000JB easily matched and even exceeded the performance that the best 10k RPM SCSI drives of the era delivered when it came to desktop performance.

While SCSI drives feature superior mechanics, their server orientation forces them to trade away firmware optimized for highly-localized patterns in favor of strategies that maximize returns in random access scenarios. In the Raptor, WD faces much of the same quandary.

Things To Keep In Mind

[MBCook]

There are a few things to keep in mind about these numbers. Most of them are mentioned in the article, but they're scattered around. Just think about these things:

• Seperate Card - Remember that the SATA controller is on a seperate card, it's not integrated into the chipset. So these number could (and probably will) change for the better when we see SATA built into the southbridge later this year (was it Grandale from Intel that will do this? I'm too lazy to look it up).
• Drive Size - The drive in the review is up to 1/6th the size of some of the other drives in the review. So if you're comparing this drive you have to remember that it would perform better if it was a 160 gig drive and didn't have to work all over it's platter.
• SATA - All the other drives in this review are either ATA or SCSI. So as SATA goes, this drive might be king of the hill by far.

Those said, I have a few other things I'd like to say. First of all, it's nice to see that the drive is quiet. Even many 5400 and 7200 RPM drives are quite loud today. It's nice to know that going to 10k isn't going to turn my PC into a jet engine. Also, they mention that the reason that we haven't seen 10k IDE drives before was that servers didn't want them since they couldn't be hotswapped like SCSI. SATA supports hotswap in theory, but can you hotswap today? I don't think Windows lets you, IIRC (or if it does the system is a bit unstable afterwards). Does Linux let you hotswap SATA drives? If all the drives are one one controller (say RAID 5, or something else redundant) and you swap a drive, does the OS even know it happened? I don't have any expirence with hotswapping hard drives.

Re:Stand back and watch for now..

[Sivar]

My main concern is its actual capabilities when being used to store/delete/etc large, numerous files and how long until the hard drive finally crashes and dies out.

The Raptor has a 5-yr warranty (5 times as long as most desktop hard drives) and is targetted for the server market. Unless WD seriously screwed up, I am willing to be that it is about as reliable as other enterprise 10K drives (all of which are SCSI)--that is to say, incredibly reliable.

A 10k IDE drive is bound to have a ton of hard drive space

Actually, the faster the platter spins, the lower density each platter must be in order for the heads to keep up. For example, the Western Digital Raptor is a 36GB drive with a single 36GB platter (that's 18GB/side). This is the same size of platter as on the largest of 10KRPM SCSI drives.
To contrast, the largest platter size on a 7200RPM drive is 80GB/platter (or 40GB/side), and Weste3rn Digital is about to release a 250GB drive which will have three 83.3GB platters.

Higher platter density improves speed as well, but generally speaking (VERY generally speaking), increasing rotational speed improves drive performance more than having a somewhat higher density platter. Those of course varies based on what you are doing with the drive, whether it involves lots of random accesses (mail/webserver) or lots of linear accesses (video editing) or something in between.

In general, expect higher RPM drives to trail behind lower RPM drives in platter density, and therefore in maximum available disk space.

Re:I know it's a joke, but

[Sivar]

Have anybody ever actually thought about it? For the amount of extra money to blow, why not spend more for memory and have EVERYTHING run from there? what, 4G is not enough for your desktop system? x86 only addresses that much right now, ya know...

Compare the price of 4GB of RAM with a 10GB hard drive. Also note that all memory used for a RAMdisk (as disk which will vanish once power is turned off) will be unavailable to applications.

Notice that computers run on multiple tiers of increasingly large and decreasingly expensive storage. This has been found to have the best performance/cost ratio. First we have registers, then L1 cache (except for Pentium IV's), then L2 cache, then on some systems L3 cache, THEN RAM, then the hard drive.
RAM is simply not cost effective for mass storage, and the performance benefits of using a RAMdrive really aren't very noticeable for many tasks. They help immensely for extremely random I/O, like running a mailserver, but Office and Diablo2 aren't going to run so much faster that it justifies the huge jump in cost and huge increase in risk (RAM drive dying when power goes out).
Besides, if we used a slow hard drive to load 4GB of data into RAM, can you imagine how long booting the system would take?

That said, there are companies offering battery-backed RAMdrives which fit in a PCI slot, and there are those (Armadillo comes to mind) which offers huge, fast FLASH-RAM drives in both IDE and SCSI flavors, but they are very expensive. There's more to making one than simply collecting a bunch of DIMMS together, ya know. :)