[Eugene]

In the same vein as our preview of the Office DriveMark results ( http://forums.storag...showtopic=20450 ), here's a peek at the results from the SR High-End DriveMark 2006:



Again, a quick legend:

"no CQ" next to a drive means that NCQ was disabled on the MaXLine III and Barracuda 7200.8, and that TCQ was disabled on the Raptor WD740GD. The "default" entries for these three ATA drives all include their queuing in the results.

The "S" and "D" next to Seagate's SCSI drives indicate that that drive was set to "Server" (factory default) or "Desktop" page-mode strategies via their SeaTools Enterprise utility.

Here's a look at the relative changes:



The High-End 2006 DriveMark is a capture of Veritests Content Creation 2004 WinMark, which consists of execution of key tasks from the following applications:

Adobe® Photoshop® 7.0.1
Adobe® Premiere® 6.50
Macromedia® Director MX 9.0
Macromedia® Dreamweaver MX 6.1
Microsoft® Windows Media™ Encoder 9 Version 9.00.00.2980
NewTek's LightWave® 3D 7.5b
Steinberg™ WaveLab™ 4.0f

A look at the queue depths generated by these applications:




for an average of 1.24.

Enjoy!

[continuum]

Wow. The tight cluster of 10K and 15K at the top makes me wonder what's really going on, particularly given the abysmal performance of some units such as the 10K.7's...

[hmmm]

Goodness Deskstars are just getting destroyed by TB4.

[udaman]

Which brings up the question again. I thought the Spinpoint and Barracuda 7.2k.8 versions and the like where supposedly at the bottom of the heap, unimpressive in past SR's tests; and that surely someone with as much experience as Eugene, should have been able to ascertain the 'seat-of-the-pants' empirical feel of these supposedly underachievers. While like lemmings, or should I say the 'emperor has no clothes' convinced themselves of the superiority of Maxline or Deskstar models. If Eugene has no idea what the fastest drive is like compared to the slowest groupings, surely no one else can tell the difference. It's all imaginary!

Which means I can safely buy the slowest apparent laptop drive within a given rpm rating, and expect that it's really the best performer???

It would seem also then, that differences of more or less than as much as 50%+ would be unnoticeable for the vast majority, and we should buy for warranty, heat, power consumption, reliability, and price before considering performance...within a given rpm range?

[bobsykes]

I would agree with the second half of your conclusion. For SATA I use the 7200.8, 5-yr warranty, cool, quiet, ripping sustained transfer for video editing; and, the Fujitsu MAU for speed, reliability, and performance with queued access intense apps like ProTools, and it's $850 vs. $1,200 for it's competitors. To me this site is great for matching drrive performance to your requirements, but your stated criteria are critical too.

rb

It would seem also then, that differences of more or less than as much as 50%+ would be unnoticeable for the vast majority, and we should buy for warranty, heat, power consumption, reliability, and price before considering performance...within a given rpm range?

[/quote]

[Olaf van der Spek]

udaman, on Aug 4 2005, 02:33 AM, said:

Which brings up the question again. I thought the Spinpoint and Barracuda 7.2k.8 versions and the like where supposedly at the bottom of the heap, unimpressive in past SR's tests; and that surely someone with as much experience as Eugene, should have been able to ascertain the 'seat-of-the-pants' empirical feel of these supposedly underachievers. While like lemmings, or should I say the 'emperor has no clothes' convinced themselves of the superiority of Maxline or Deskstar models. If Eugene has no idea what the fastest drive is like compared to the slowest groupings, surely no one else can tell the difference. It's all imaginary!

Which of the apps (in this benchmark) do you (or the majority of users) run that's (significantly) IO bound for you?
Now that most systems come with a decent amount of RAM (although 256 mb is still not uncommon) disk performance doesn't matter as much as it used to.

This post has been edited by Olaf van der Spek: 04 August 2005 - 08:00 AM

[douglite]

Wow, 7K400 takes another beating Are we going to hear from T7K250 or 7K500 in the near future?

Also, the 7200 ATA drives are pretty close (if we exclude the abyssmal showing of 7K400 as an outlier), with a spread of about 10% between SP P80 and WD2500KS.

I am the keeper of the [H]ard|Forum [H]ard|Drive Buying Guide (I am known as DougLite there as well), and these results are shaking things up for me. WD2500KS wasn't very compelling on TB3, but is best in class on TB4. Hitachi has gone from penthouse to outhouse, and it looks like WD has moved into the penthouse suite.

Seagate looks a little better this time around than in Office, Maxtor is still sitting pretty, and MAU is still untouchable. Samsung is at least keeping up with the pack, admittedly still at the back, although the P120 drives show promise with their very high transfer rates. WD had better have a refresh on WD740GD in the works, or it is going to become an extinct dinosaur (pun intended) with the next generation of 7200 drives. Raptor's lead over 2500KS is a scant 7.5%, certainly not worth the premium of a 40% higher price and 70% less capacity.

The short queue depths explain the little-to-nothing impact of CQ, at least in my mind.

[MartinP]

I presume that the new benchmarks still enforce the same order of submission of I/Os to the drives, even when CQ is enabled?

cheers, Martin

[Gilbo]

MartinP, notice the differences between the same drives with CQ enabled or disabled.

[Eugene]

MartinP, on Aug 6 2005, 09:03 AM, said:

I presume that the new benchmarks still enforce the same order of submission of I/Os to the drives, even when CQ is enabled?

cheers, Martin

That's correct, request order and interarrival times are properly preserved.