WD6400AAKS faster than Raptor?

[crspyjohn]

WD6400AAKS faster than Raptor?

http://forums.slickdeals.net/showpos...5&postcount=71

It looks right but is it true?

[LagunaX]

This I would like to see proven with real numbers and not theoretical.

Also seek times by his numbers are still slower at 6.6 vs. 4.6.

[G80]

What happens when you partition a Raptor then?. It will probably blow his numbers away.

[One_Hertz]

I don't think its linear... I know it doesn't work in practice. I am using 10% of my raptors in the first slice and getting 5.6ms. According to his calcs I should be getting 3.4ms.

[tiro_uspsss]

i highly doubt a 7.2k rpm hdd will *ever* have the access time of a 10k rpm hdd - & thats will all this tricky partitioning wizardry out of the pic too... my 2 cents

[Big Myke]

I would like to "shortstroke" my 4 raptors in raid. How does one correctly shortstroke? I've tried creating partitions through windows but that does not reduce my access times. can someone help me and others who are also wondering.

[EvilBellyLint]

np doubting Thomas's

I'll post some benchies in a few days or so when I get this rig built.


EBL

[EvilBellyLint]

I don't think its linear... I know it doesn't work in practice. I am using 10% of my raptors in the first slice and getting 5.6ms. According to his calcs I should be getting 3.4ms.

Of course you would not be getting 3.4ms. the Raptor's Rotational Latency
alone is 3ms, and there is a limit to how low you can shave Seek Times.
Drive heads are mechanical devices, and as such have a limit to how fast
they can physically move.

I state as much in the above quoted thread.

EBL

[EvilBellyLint]

This I would like to see proven with real numbers and not theoretical.

Also seek times by his numbers are still slower at 6.6 vs. 4.6.

Look again, closely this time.


EBL

[RAW-Raptor22]

Nobody is faster than Raptor!

[LagunaX]

Look again, closely this time.


EBL

2ms seek time
Ok, I'll wait 4 your benchies

[One_Hertz]

Of course you would not be getting 3.4ms. the Raptor's Rotational Latency
alone is 3ms, and there is a limit to how low you can shave Seek Times.
Drive heads are mechanical devices, and as such have a limit to how fast
they can physically move.

I state as much in the above quoted thread.

EBL

So the raptors shave off less than half their seek time by using fastest 10% and you think your wd6400 will shave off 3/4s of your seek time by using fastest ~25%?

Optimistic just a wee bit, no?

I predict ~9ms at best

[EvilBellyLint]

So the raptors shave off less than half their seek time by using fastest 10% and you think your wd6400 will shave off 3/4s of your seek time by using fastest ~25%?

Optimistic just a wee bit, no?

I predict ~9ms at best

What will matter most is I/O per sec. I'll try to get a hold of
IOMeter or similar to test... along with HDTack/Tune just to
compare sustained throughput.

Of course I could always be optimistic as you say, but limited
initial data strongly suggests this drive will outperform the Raptor.
I'm not trying to push these new drives on anyone...lol just trying
to offer some solid alternatives to outdated technology.


peace

EBL

Edited to Add: even if your prediction of 9ms is correct,
there are factors like caching algorithms, data throughput,
etc... that determine "real world" performance, so we really
need to do some actual application testing if we want the
true full picture. Benchmarks are fun and all, but they don't
really tell the complete story.

[One_Hertz]

What will matter most is I/O per sec. I'll try to get a hold of
IOMeter or similar to test... along with HDTack/Tune just to
compare sustained throughput.

Of course I could always be optimistic as you say, but limited
initial data strongly suggests this drive will outperform the Raptor.
I'm not trying to push these new drives on anyone...lol just trying
to offer some solid alternatives to outdated technology.


peace

EBL

Edited to Add: even if your prediction of 9ms is correct,
there are factors like caching algorithms, data throughput,
etc... that determine "real world" performance, so we really
need to do some actual application testing if we want the
true full picture. Benchmarks are fun and all, but they don't
really tell the complete story.

Yep IOPS will be the best measure. We will see, too difficult to make any real calls right now... I don't believe it will beat the raptor, but that is just a guess...

[Serra]

There are actually a number of issues in the theory posted. Some for, some against. Most prominently:

1. The WD drive uses multiple platters. It is not all a straight seek, there is also an electrical shift from one platter to another, which will actually boost speeds (searching near the edge of a second platter versus searching towards the center of the first).

2. You cannot possibly assume that you can multiply the average seek time by .234 to obtain an adjusted seek time... it just doesn't work that way. The average full-stroke seek time of a 7200rpm drive is ~21ms - at closest approximation, that's what you should have looked at (versus ~10ms for the Raptor).

Now with this all said, I do believe it's likely possible to beat a raptor in IO performance thanks to the significantly faster sequential read speeds offered by current 7200rpm drives... but the fact is that the a good portion of the raptor must necessarily be faster than much of the 7200rpm drive.. now, exactly where - if ever - a crossover occurs, I cannot say for sure. I do know however that as a general rule I try not to fill a partition to the point where I would have to find out (excepting mass storage drives that I don't care about speed with).

[EvilBellyLint]

I think you won't have much problem with over-filling a 100-150GB
partition on the WD6400AAKS if you only place tose applications
that must reside on the OS drive there, and isolate all others on
another drive.


EBL

[NapalmV5]

lol the only way to take full advantage of lower ms and increased MB/s is not to use the castrated storage

create partition @ 10% 20% whatever% and thats it..

[Kondik]

Generally 7.2 k RPM Hdd - 3.5" plates
Raptors 10 k RPM - 2.5" plates - seek will be allways lower
because it spins faster and the head needs to do less movement to "find data"

Generalized a lot
Never HHD's can overburst rate raptors maybe overspeed it
But Raptors are Raptors , and Raptors in Raid 0 are destroyers (:

Btw i hate partitions, just get a customizable defragmenter and order him to put data on the "fastest line" that's all

[tiro_uspsss]

Generally 7.2 k RPM Hdd - 3.5" plates
Raptors 10 k RPM - 2.5" plates - seek will be allways lower
because it spins faster and the head needs to do less movement to "find data"

i doubt raptors have 2.5" plates...

[EvilBellyLint]

i doubt raptors have 2.5" plates...

correct. They do not.


EBL

[Serra]

I broke out the MS paint to make a graphical representation of what you should expect.

The raptor (red) will start off with the best access times, and those will degrade as you reach the end point of the disc.

The 7200rpm drive (be it WD or, better, Samsung F1) will start with worse access times and although they will degrade as well, they will degrade at a lesser rate than those of the raptor because the portion of physical disc they travel over is smaller.

What EBL seems to be stating is that the 7200rpm drive will be always faster, and that is not true. The questions that will have to be answered are (referencing the attachment):
1. Does the raptor ever fall below the access time of the 7200 rpm drive (I would say probably yes)
2. How large is "X"?
3. Where is the endpoint of "Y"?
4. How large is "Z"?

For example, if the crossover (assuming there is one) occurs at 149GB, then you're most likely better off with the raptor if you're looking for access times. The crossover could alternatively be around 1GB though, and that would make the raptor silly. However, given that the raptors performance does have to degrade notably before even the outermost track of a 7200rpm drive is its equal, I would suggest that "Y" is most likely to be a value of such size that if you're only using it for your OS and a few apps, they would probably all fit within the zone and a raptor would be best for access times.

None of this is to speak of access patterns, of course... I mean, if you're only using the outermost few GB for random access and the rest is just non-speed-critical reading, then it'll be another vote for the raptor. Conversely, if the sequential read speeds offered by todays 7200rpm drives is attractive for your usage habits, then you should certainly get that drive.

[MartinuZ]

Wouldn't it be possible to achieve way higher transfer speeds and access times if you combined the data density of a 1TB drive, 4 platters (is 5 possible?), only use the outer 20%, and have it spin at 10k rpm?

200GB, 115+MB/s (from 0% to 100% of the drive!) 5ms random access times? :>

[tiro_uspsss]

Wouldn't it be possible to achieve way higher transfer speeds and access times if you combined the data density of a 1TB drive, 4 platters (is 5 possible?), only use the outer 20%, and have it spin at 10k rpm?

200GB, 115+MB/s (from 0% to 100% of the drive!) 5ms random access times? :>

of course! I've always wondered why Seagate or other doesnt take just *one* dense-as-Paris platter (i think they have 320GB single platters now) & forget 10K rpm, go all the way - go 15K rpm - I mean Seagate has plenty of experience spinning HDDs that fast (SCSI range)...

single super dense platter
SATAII interface
32+ MB cache
15K rpm

= killer drive

[One_Hertz]

of course! I've always wondered why Seagate or other doesnt take just *one* dense-as-Paris platter (i think they have 320GB single platters now) & forget 10K rpm, go all the way - go 15K rpm - I mean Seagate has plenty of experience spinning HDDs that fast (SCSI range)...

single super dense platter
SATAII interface
32+ MB cache
15K rpm

= killer drive

= killer production costs and the price would have to be near what SSD are right now and it wont be able to beat SSDs regardless what they do to it.

[tiro_uspsss]

= killer production costs and the price would have to be near what SSD are right now and it wont be able to beat SSDs regardless what they do to it.

i will agree with the fact that it wouldnt beat a SSD - but production cost? puh-leez, they already make the platters left, right'n'centre (Seagates 7200.11 line-up), 32MB cache the already do in all higher end HDDs (again 7200.11), & the spindle motor - they already have them from their SCSI line... sure it'll be an 'enthusiast' part & it'll kick the Raptor thus it'll be more expensive than Raptor - but still cheaper if compared to GB/$ re: SSD - & while this dream HDD of ours cant match a SSD in ms, it'll beat or come oh so close to MB/s (go look @ MB/s that seagate is advertising for its new 15K.6 SCSI HDDs - 160MB/s )... the Raptor150 ATM is getting flogged of @ newegg for ~USD$170. Lets say the dream drive of urs has double the capacity (single ~300GB platter) & aforementioned specs.. what would u pay out of curiousity? me? mmm... USD$250-300...