[Eugene]

Hi all,

Testbed4's been a long time in coming- I know many readers have been eagerly awaiting to results to see how their favorite drive or prospective purchase fares under more contemporary access patterns and hardware. The good news is that formal measurements are complete, all that lays in front of me is the formal writeup (which, if you can remember from the TB3 article, can be -long-).

In the meantime, however, I'm pleased to present results from TB4 for those yearning for a sneak peek. The figures that follow are "official"- they'll eventually represent their unit in our revamped performance database.

First up, let's take a look at the cornerstone single-user measure, the SR Office DriveMark 2006:



A couple notes:

1) "no CQ" next to a drive means the bar represents the drive's performance (in I/Os per second) with NCQ disabled (in the case of the Barracuda 7200.8 and the MaXLine III) or TCQ disabled (in the case of the Raptor WD740GD).

2) The "D" and "S" next to Seagate's Cheetah drives represent the results obtained when the drive was set to Seagate's "Deskstop" and "Server" buffer segmentation strategies.

These results come from newer applications running on newer hardware. Some drives coped with the change better than others... there are losers, and there are winners. When we take the new results and constrast them with the SR Office DriveMark 2002 settings currently found in the performance database, the following changes occur.



These figures are the difference that the given drives exhibit when comparing their respective scores with the average of all these drives in the given test. In other words, relative to other drives, the Cheetah 15K.4 (in desktop mode) loses much of its lofty standard in this updated benchmark while the Caviar WD2500KS does much better with new hardware and new applications.

Enjoy. Comments and questions welcome... and more results to follow very soon!

[KingGremlin]

Hey Eugene,

Would you mind simply listing what applications comprise this suite?

It seems in general, the new test bed and suite brought the drives closer to each other, as the fastest drives generally did weaker while the slower drives did better relative to each other. That might be caused by the updated hardware more than the benchmarks, as the updated hardware with more RAM I would assume may contribute to minimizing the impact that hard drives have on overall performance thus slower drives don't hurt overall performance as much, while faster drives aren't utilized as much to display their superior abilities.

(Edit by Eugene- KG is actually referring to a graph that was briefly posted but that didn not represent the change in results for this test (it actually represented the change in the high-end test, results of which are going to be posted soon). The correct "change between 2002 and 2006" graph is now posted. Sorry about that, KG... I'm demonstrating right here why posts should not be edited after the fact- they create race conditions /edit)

Also, Seagate's SCSI division should be embarassed by those results. I understand that the drives are targetted for servers, but if you are going to have a "desktop performance" mode, it should not be equal to your 7200RPM ATA counterpart which isn't exactly a performance king by ATA standards itself. When compared to Fujitsu's and Maxtor's offerings, there are really no excuses for abysmal results like those.

Note #2, avoid TCQ like the plague if you own a Raptor and don't run a server. What a beating it takes when you enable it.

The very close proximity of Maxtor's 15k and 10k offernings is rather odd as well. Difficult to say whether the 15k II is a disappointment despite still coming in 2nd, or whether the 10k V is simply that good since it put a beating on second place Fujitsu.

[Eugene]

KG,

The test is a capture of the Veritest Busines Winstone 2004, which is a script that consists of launching and running MS Office XP, Winzip v8, NAV 2003, and a couple other productivity applications.

Regards,

Eugene

[udaman]

Say Eugene (oh, almost forgot, while '36k' is asleep... "you rock!", hehe, I beat him to the gratuitous pandering ),

I'm wondering how this giant jump for the Samsung SpinPoint jibes with your own 'hands on' feel for it? A really quiet drive, but before in prior testbed's seemingly an under performer which has been indicated in the review writeups? Hey, that Kennyshin must be some smart guy then (he has that drive IIRC).

I don't see my favorite 2.5 laptop drive on that list-4yrs is an awful long time for us to wait on the results before we upgraded our drives, lol; think of all the corporate buyers sweating bullets over this decision ;-) *hint* .

<insert> celebration, party icon moving gift</insert>

[Olaf van der Spek]

Eugene, on Jul 30 2005, 09:47 AM, said:

The test is a capture of the Veritest Busines Winstone 2004, which is a script that consists of launching and running MS Office XP, Winzip v8, NAV 2003, and a couple other productivity applications.

This is just the office benchmark. I'm not sure, but are there really (many) people that are 'bottlenecked' by the HDD when running those apps?

[Gilbo]

Congratulations on the success of the project Eugene. I'm very glad and excited that we're starting to see new results.

The Fujitsu SCSI's pull back to the top in the new suite. It's a remarkable tradition they've maintained here at SR. Also, the Maxline III puts in a rather gonadial performance...

[Gilbo]

Olaf van der Spek, on Jul 30 2005, 06:06 AM, said:

Eugene, on Jul 30 2005, 09:47 AM, said:

The test is a capture of the Veritest Busines Winstone 2004, which is a script that consists of launching and running MS Office XP, Winzip v8, NAV 2003, and a couple other productivity applications.

This is just the office benchmark. I'm not sure, but are there really (many) people that are 'bottlenecked' by the HDD when running those apps?

I think people have an interest in application launch speed. Accesses that represent those patterns will represent a significant proportion of this test, so I think many people will find it quite useful.

Personally, I've advocated and practiced pre-loading applications at startup as a way to greatly speed application 'launching,' but I think most people still load applications off the disk despite the numerous performance disadvantages associated with that.

[Olaf van der Spek]

Gilbo, on Jul 30 2005, 03:49 PM, said:

I think people have an interest in application launch speed.  Accesses that represent those patterns will represent a significant proportion of this test, so I think many people will find it quite useful.

Personally, I've advocated and practiced pre-loading applications at startup as a way to greatly speed application 'launching,' but I think most people still load applications off the disk despite the numerous performance disadvantages associated with that.

But what app (except Paint Shop Pro, PhotoShop and apps with nasty plugin architectures) take that long to load?
Especially the second time?

[fallbreak]

Wow...thanks for the numbers Eugene...

....guess, I will have a hard time next week, ignoring swear words about this test...Greetings from Minnesota ;-)

[Gilbo]

Instantaneous just feels good .

What would you recommend for a Office Test Suite? I think Eugene's choice of the Veritest Suite is excellent. Firstly, it is a standard performance metric. Secondly, it is actually representative of many common tasks.

As for its significance in the grand scheme of things, I don't think anyone on this board really believes that anything more than a modern 7200 RPM drive is going to make a difference for most people's experience, or makes recommendations based on such beliefs. On the other hand, SR provides a tremendous service to those who are curious about observing the quantitative performance of disk drives. People come to this site to satisfy their curiousity. The measurements matter to us, but I don't think anyone's trying to pretend that hard drive performance is going to change the way most of us work or play.