Western Digital VelociRaptor - BeHardware
>> Hard Disks

Written by Marc Prieur

Published on July 24, 2008

URL: http://www.behardware.com/art/lire/727/


Page 1

Introduction



Since 2003, Western Digital is the only one to offer Serial ATA drives that function at a speed of 10.000 rpm. This rotation speed up until now was reserved for SCSI drives that could even go as high as 15.000 rpm, gives an additional boost to performances compared to classic models that function at 7200 rpm.

The history of the Raptor started in February 2003 with a 36 GB version while 7 months later its capacity was doubled. We then had to wait until December 2005 for the new generation with a capacity of 150 GB, NCQ and a doubled cache size of 16 MB. Western Digital is now back in the spotlight with the successor to its Raptors, the VelociRaptor.

What are the 10.000 rpm for?
Rotation speed has a double impact on the performances of modern hard drives. The first is in terms of speed. At equal data storage densities, the faster the platters turn, the higher the speed.

However, this is compensated for by the fact that it is difficult to use the same density on faster turning platters. Therefore, 7200 rpm drives make up for their slower speed with platters that generally have higher densities.

The second effect is in terms of latency, an essential characteristic that is added to pure access times. Actually, to access specific data on a platterís track, the reading head has to properly align itself (what manufacturers call access time) and also the platter has to turn in a way that the data is found under the reading head.

Depending on the situation, the reading head may have to wait a 1/4 or 1/2 turn or more in order to read data and this is the latency which is shortened depending on the rotation speed of a drive. In fact, at 7200 rpm, 8.3ms are needed for a complete rotation versus 6ms at 10.000 rpm.
Le VelociRaptor

While the speed always increased with each new Raptor generation (aided by increases in density), it wasnít the case for access time which remained more of less the same. So how can this value be decreased without using a rotation speed of 15,000 rpm which isnít necessarily compatible in terms of sound levels with what we expect from modern hard drives?


Western Digital in fact adopted a simple idea which is already used in professional drives, notably with Seagate models. To decrease the time needed to position reading heads, why not reduce the distance they have to move by decreasing the size of platters?

So here the drive is actually in 2"1/2 format but it comes attached to a 3"1/2 drive radiator, the WD IcePack. Note that it will not fit into a laptop as its height is 15mm versus 9mm for classic drives in this format. Moreover, this wasnít the only option available to Western Digital as Seagate kept a 3"1/2 format on its Cheetah professional drives despite the smaller platter size (the bigger casing notably enabling better sound isolation). Otherwise, Westernís choice means a non standard placement of connectors and for this reason it is denied access to certain NAS hotswap chassis, for example.


On the inside, we find two platters with a density of 150 GB or a surface density that is roughly 10% less than that found on recent 3"1/2 drives with 333 GB per platter. Compared to the previous Raptor generation, Western announces an access time reduced from 4.6 to 4.2ms and speed which goes from 88 to 120 MB/s.


Page 2
The drives, the test

The drives
For this test, we compared the VelociRaptor to a ď2005Ē Raptor, a 150 GB, as well as a ď2003Ē Raptor with 74 GB. We also included the latest current stars in this area, the Samsung SpinPoint F1 and Western Digital SE16 640 GB both of which are equipped with 320 GB platters. They are equivalent to 1 TB models but of course are less expensive.

- WD Raptor 74 GB : WD740GD-41FLC2 (31.08F31)
- WD Raptor 150 GB : WD1500ADFD-00NLR5 (21.07QR5)
- WD VelociRaptor : WD3000GLFS-01F8U0 (03.03V01)
- Samsung SpinPoint F1 640 GB : SAMSUNG HD642JJ (1AA01109)
- WD Caviar SE16 640 GB : WD6400AAKS-00A7B0 (01.03B01)


test protocol
Various measurements were carried out in tests. First of all, we were interested in a driveís ę synthetic Ľ performances: cache and sequential speeds and average access time. Next, were more practical tests, first of all involving an applicative performance index based on PC Mark Vantage and then a server load type simulation of files with IOMeter. This was followed by an evaluation of writing, reading, close (on the same partition) and far copying (on a partition which starts on 50% of the drive) with various groups of files.

These groups were composed of:

- A collection of large files: 6 files (on average 2.2 GB) totaling 13.2 GB
- Medium sized: 7.96 GB of 10,480 files (each averaging 796 KB)
- Small sized: 2.86 GB of 68,184 files (each averaging 44 KB)

The source or target of reading or writing on the drive was a RAID of two Raptor 150 GB drives in order to not be limited by this parameter. This type of measurement is interesting because, while the sequential speed gives us an idea of the performance in copying large files, things can be different with smaller ones.

All measurements were made with acoustic management deactivated; however, later on we proceed with a few tests with its activation. The test machine was based on an X38 chipset mounted on an ASUSTeK P5E motherboard while Serial ATA ports were configured in the bios in AHCI (Advanced Host Controller Interface) so that NCQ could be used.

Of course, in addition to these performance measurements, we took temperature readings after 2 hours of intensive use as well as measuring power consumption. Finally, the sound levels of each hard drive were evaluated based on their dBA values. We also provide you with a recording of each model in various states of activity.


Page 3
Synthetic tests

Cache speed

We start with cache speed measured with h2bench and its ę Core Test Ľ. Contrary to the Raptor, the VelociRaptor adheres to the SATA 3 Gbit/s standard which enables it to attain speeds in cache access worthy of this name. It was the first area in which we saw a significant improvement.
Access time

Also measured with h2bench was the VelociRaptorís access time which was also roughly 1ms lower compared to previous Raptors. While the activation of acoustic management did not have an impact on the Raptor 74 GB and it only slightly affected the 150 GB, here we immediately noticed the difference (later we will see if noise levels are significantly reduced). Western and Samsungís 7200 rpm drives are logically behind with an advantage for the first without AAM.
Sequential speed
Finally, h2bench enabled us to measure reading and writing speeds in sequential access.


With a peak at 119 MB/s, the VelociRaptor is the current fastest SATA drive. However, the gap between it and 7200 rpm models is small at the beginning of the track because they benefit from the high density of their platters. Moreover, 7200 rpm models are largely ahead of the other Raptors with the Samsung being a bit faster than the Western. In writing we noted a slight decrease in speeds.


This graph shows the precise speed depending on how far along the drive is in the operation. Note the excellent behavior of the VelociRaptor which enables it to widen the gap even more at the end compared to the Samsung and Western 7200 rpm drives.

We observe the same phenomenon in writing where otherwise the Samsung drive shows rather erratic behavior at the beginning of the track.


Page 4
PC Mark Vantage

PC Mark Vantage
We now move on to less synthetic tests, starting with an index of hard drive performances in PC Mark Vantage. FutureMark reproduces a series of recorded reading/writing operations on the drive in diverse tasks such as the startup of Vista, loading of applications (Word, Photoshop, IE, Outlook), manipulation of multimedia files (photo, videos, and music), a game (loading in Alan Wake) and scanning the drive via Windows Defender.


While we didnít note a significant evolution between the Raptor 74 and 150 GB, performances of the VelociRaptor explode. This isnít necessarily that impressive given the performances of Western and Samsung 640 GB drives which are comparable to those of the Raptors. Otherwise, between the Western and Samsung the former is better with or without acoustic management.


Page 5
File copying

File copying
This brings us to copying files. We measure reading and writing speeds, as well as the close copying (on the same partition) and far copying (on a partition which starts on 50% of the drive) of our various groups of files.

These groups of files are composed of the following:

- Large sized files: 6 files (on average 2.2 GB) totaling 13.2 GB
- Medium sized: 7.96 GB of 10,480 files (each averaging 796 KB)
- Small: 2.86 GB of 68,184 files (each averaging 44 KB)

The source or target in reading or writing on the drive is a RAID of two Raptor 150 GB drives in order to not be limited by this parameter. AAM does not have too much of an impact but you can consult its performances here.



Whether itís in reading or writing, the manipulation of large files produces results similar to synthetic performances measured with h2bench. Three drives surpass 100 MB/s: of course the VelociRaptor but also the Samsung and Western 640 GB drives. When we reduced the size of files, the Samsung has the lead over the Western model. Otherwise, while the VelociRaptor is ahead in reading, it gives up first place in writing.


In the copying of files on the same platter, the VelociRaptor takes the lead over the Samsung 640 GB in all situations; however, the latter manages to stay close behind in the close copying of small and medium sized files. Itís only in the manipulation of large files that the Western 640 GB is equivalent and even ahead of the Samsung model.

Note that while the Raptor 150 GB didnít represent a significant evolution in terms of performances during the manipulation of small and medium sized files, Western remedied the situation with the VelociRaptor.


Page 6
IOMeter

IOMeter
IOMeter is used to simulate the load in a multi-user environment by using a server type file load comprised of 80% reading and 20% writing all in a 100% random manner on the drive. In this type of situation, NCQ can be particularly useful because of multiple concurrent commands. In this test, we measured performances expressed in inputs /outputs per second (IO/s) with 1, 2, 4, 8, 16, 32, 64 and 128 simultaneous commands. Of course, with a single command NCQ has no effect.


The more we increase the number of simultaneous accesses, the more the VelociRaptor widens the gap between it and its predecessors. The Samsung and Western 640 GB drives start off equal but better management of NCQ for Western enables it to later take the advantage.


With acoustic management, performances collapse! This is particularly true for the VelociRaptor which remains behind the Raptor 150 GB for up to 8 simultaneous accesses, although it subsequently takes the lead. However, note that these results should be considered in parallel with the acoustic impact of this function which is not equivalent from one drive to another (as we will see later on).


Page 7
Consumption, temperature

Power consumption
Here we have the power consumption for each hard drive using a clip-on ammeter on the 5V and 12V lines of each model.


The reduction in the size of platters as well as that of reading heads enables strongly reducing the VelociRaptorís power consumption which is the lowest by far.
Temperature
Next, we measured the hard drive temperature after two hours of intensive use with IOMeter. These results are with the HDD out of its casing, without ventilation, and each slightly raised. Temperature is taken using an infrared thermometer in two areas; in middle on top of the drive and on the hottest part on the left side (the IcePack temperature for the Velociraptor). Finally, we also give you the temperature as reported by the driveís internal probe (SMART).


Here again we notice a significant improvement of the VelociRaptor over the Raptor. And we also noted that the probe integrated to hard drives isnít too reliable. Upon startup of the machine, while the room was at 24įC, the Samsung proudly showed 19įC (and to our knowledge it is not equipped with air conditioning). We noted a similar problem on the 1 TB model as well as the Caviar SE16 640 GB showing 27įC just after startup or a slight overestimation of its temperature.


In order to better illustrate the difference in temperature between the Raptor 150 GB and VelociRaptor 300 GB, we used infrared thermal imaging in a second test and still after two hours of intensive use.


The difference in temperature is immediately noticeable. Note that part of the VelociRaptorís surface emits less heat and therefore distorts infrared readings. The green part should actually be orange like the surface covered by the sticker.


Page 8
Noise levels

Noise levels
For this measurement, we place each hard drive in a Textorm jelly case in order to avoid any vibration from the desk with a sound probe 12 cm above. Of course, the PC we use in this test is fanless so that we can isolate the hard driveís sound level:


Up until this point, the VelociRaptor had been faultless; however, here the result is catastrophic. In fact, sound levels in rotation are so high that we get the impression we were 5 years in the past when hard drives used ball bearings systems instead of rarer hydrodynamic bearings.

To have a better idea of the level of noise produced by these hard drives, here are some recordings made with a Roland portable Edirol R-09 recorder placed 12cm above the drive. We recorded 15 seconds in rotation, then 15 seconds in access with acoustic management (except for the Seagate) and finally 15 seconds without acoustic management.

- Western Caviar SE16 640 Go
- Samsung SpinPoint F1 640 Go
- Western Raptor 74 Go
- Western Raptor 150 Go
- Western VelociRaptor

You may have noticed that in addition to being quite noisy, the sound emitted by the VelociRaptor is sharp making it even more unpleasant.



Whatís the cause of this unfortunate characteristic which according to various users seems to be rather inconsistent from one drive to another? Does it come from a drive which isnít as well insulated or is the WD IcePack a type of soundboard? To get to the heart of the matter, we equipped ourselves with a Torx T8 screwdriver and separated the VelociRaptor from its IcePack. In the dismantling we noticed that the IcePack isnít flat and has two contact points intended to transfer the heat of the driveís PCB.


The result was unequivocal. Sound levels dropped and the drive became more silent than the Raptor 150 GB (the recording is available here), whether in access or not. On the other hand and as you can see on the following thermal imaging, the drive heats up significantly more surpassing a surface temperature of 50įC. However, the most worrisome is found on the back of the drive where the IcePack is the most effective; the hottest point was measured at 64.2įC after two hours of intensive use in the open (24įC room temperature).



To confirm our hypothesis regarding increased noise levels, we conducted a simple experiment by adding some material between the VelociRaptor and IcePack. If this was done just at the pads, the result was the same. However, if we put material under the entire PCB, which means we have more air between the IcePack and drive, the noise level was almost identical to that recorded with the drive alone! So the IcePack is indeed a soundboard and amplifies the noise when the drive is in rotation. We can only hope Western will modify its design so that the entire IcePack is in contact with the drive and not just part of the PCB!


Page 9
Conclusion

Conclusion
With the VelociRaptor, Western Digital incontestably offers the best current SATA hard drive. The gain compared to the previous generation is significant. This is notably thanks to a reduction in size of platters which enables decreasing access time and increased surface density that allows much higher speeds. Of course, Western makes us pay a lot for its exclusivity on the 10.000 rpm SATA segment and you will have to plan on spending roughly 250Ä for a VelociRaptor 300 GB, or the price for example of two Samsung 1 TB SpinPoint F1s!


However, there is one hitch concerning noise levels. Indeed while the IcePack offers very good results in terms of cooling the drive, on the other hand, it also serves as a veritable soundboard rendering the drive unpleasant in simple rotation. This is true even if the defect varies in degree from one drive to another. Separating the VelociRaptor from the IcePack resolves the problem but we then have a drive that heats up a little too much for our taste. This may also compromise the product life as well as the fact that dismounting the IcePack also cancels the manufacturerís warranty!

Seagate Cheetah

With this in mind, we may ask if Western Digital didnít make a poor choice by making a real 2"1/2 drive equipped with the IcePack. In the past, manufacturers such as Seagate used 2"1/2 platters all the while keeping the 3"1/2 casing on high end models. Perhaps, this solution might have been more efficient in terms of noise levels as well as providing proper cooling. Or more simply, according to our tests it would have sufficed that Western slightly modify the design of its IcePack so that it is entirely in contact with the PCB and not just at two points. In the meantime, we noticed that they have now announced a version without the IcePack and therefore without the noise and warranty problems. However, you yourself will have to ensure proper cooling of the drive.

At a time when SSD type memory solutions (a subject we will come back to this summer) are starting to become more affordable and are actually completely silent, this isnít the best timing for Westernís mistake. For this reason, it is difficult for us to recommend the VelociRaptor with IcePack, unless if you donít care about the noise level. As for the two best 640 GB drives, the Western Caviar SE16 and Samsung SpinPoint F1, itís hard to choose a clear winner. The Western is faster in applications and as a file server while the Samsung is better in file management and is a bit more discreet.


Copyright © 1997-2014 BeHardware. All rights reserved.