Pros: Fast, quiet, low heat...and fast!
The MAW3073NP is the third generation of Fujitsu's Ultra320 (320MB/s) SCSI introduced in September 2005. It is the 10K RPM spindle-speed SCSI hard drive and currently available in three family members: MAW3073, MAW3147 and MAW3300. They are available in three types: NP (68-pin), NC (80-pin) and FC (fibre channel). I am reviewing the MAW3073NP with storage capacity of 73.5GB and uses 68-pin interface connector. It is the smallest storage-capacity hard drive in the MAW series. The storage capacity of the three members can be identified by their model number: MAW3073NP (73.5GB, 68-pin), MAW3147NC (147GB, 80-pin), etc.
The MAW series features fluid dynamic bearing (FDB) to ensure quiet operation and smooth response. Being the smallest member in the family, the MAW3073NP uses a single platter with 73.5GB storage density and two read/write heads.
Currently priced at around $180, the Fujitsu MAW3073NP, though seems expensive, is the most affordable performance SCSI hard drive from Fujitsu. Due to its ultra-fast 320MB/s data transfer rate and its fast 10,025 RPM spindle-speed active response, the Fujitsu MAW3073NP is best suited for a busy data storage partition in a multimedia workstation.
Installation and Hardware compatibility
Only high-end workstation motherboards are equipped with SCSI controller that readily supports SCSI hard drives without any added peripherals. Another draw back is that only high-end motherboards are equipped with a PCI-X slot to take advantage of the high bandwidth in the SCSI configuration; that is, these boards offer PCI-X slots with bus speed up to 64-bit/133MHz compared to the legacy (standard) 32-bit/33MHz PCI slots. My motherboard does not have built-in SCSI controller but supports PCI-X with bus speed up to 64-bit/133MHz, and therefore I can use high bandwidth SCSI controller to take advantage of the MAW3073NP's high performance.
Installation of the MAW3073NP is quite straightforward, similar to that of IDE. The only difference is that SCSI setup is not limited to a master/slave configuration imposed on IDE devices. In fact, this is the superior part of SCSI to IDE. Every SCSI controller supports drive configurations virtually in the same way. For instance, a single-channel SCSI controller can support up to 15 SCSI devices, with any ID number from 0 to 15, except ID 7. The controller reserves ID 7 for its own use.
I have used (and tested) Fujitsu MAW3073NP on two different SCSI controllers on two different motherboards. These are Adaptec 29160LP Ultra160 64-bit/66MHz and LSI Logic LSI22320 Ultra320 64-bit/133MHz. The Adaptec 29160LP is capable of transferring data up to 160MB/s, and the LSI22320 is capable of 320MB/s. These two controllers are backward compatible with regular PCI 32-bit/33MHz. Thus, it is readily possible to use the MAW3073NP with any of these two SCSI controllers on a standard ATX desktop motherboard. The only obvious disadvantage is that the setup will not benefit the fast data transfer rate compared to the 64-bit on 66MHz, 100MHz or 133MHz bus. More on this later.
My workstation consists of five SCSI and one SATA hard drives. My SATA and one SCSI are served for permanent storage while the others are used for system and program drives as well as active reading/writing partitions. I configure my MAW3073NP to be the last SCSI in the chain, since I often swap it with another 73.5GB SCSI drive (MAP3735NC). Thus, for ID configuration, I set ID 4 for the MAW3073NP. It can of course take any ID number, provided no two SCSI drives in the chain have the same ID number. I use the MAW solely as data storage. Therefore, setting the entire drive as a single partition is very convenient and highly practical.
Its terminal on the back provides a set of four SCSI ID combination from 0 to 15. In addition, the terminal provides other configurations, such as Write Protect, Remote Start, Remote LED, or Terminator power. The terminator power setting allows the hard drive at the end of the chain to terminate the signal. I use a manual terminator on the cable instead, and I only concern with setting the right unique ID number for each of my SCSI devices. This also eliminates the need to keep track of which hard drive acts as a terminator. There is also a terminal setting that allows the drive to be connected to an 8-bit SCSI bus (primitive SCSI bus). The MAW3073NP uses separate interface ports for data (68-pin) and 4-pin molex power (similar to IDE).
Since a SCSI controller has its own BIOS and detects SCSI devices in the chain during bootup, there is no need for software driver or installation. Under Windows XP Pro SP-2, the device is recognized; it can be formatted and added into the drive configuration readily. The MAW3073NP (with its 68-pin) is not hot swappable. If I need to swap a hard drive, I will have to take the MAW3073NP off line, shut down the system, disconnect the MAW3073NP and reconnect its replacement. Then boot the system back up. The process is quite similar to adding a new hard drive in IDE or SATA configuration.
I use the MAW3073NP as a storage partition that holds original video clips and photo images. Many of the data often get moved out of the drive when the project is completed, either onto a DVD disc or to a permanent data drive. Therefore, its performance reported here is mostly due to data storage performance. I did, however, install Windows XP to test the MAW3073NP's responsive as a system drive (at least 20GB of it) before using it solely as a data storage device. Because SCSI hard drive is quite expensive, I felt that using a 73GB for a system drive is a waste of disk, even though the MAW3073NP is quite responsive and highly capable as a system device.
In my workstation, I have mainly Fujitsu SCSI drives that include MAS3367NP, MAP3367NP/NC and MAP3735NC. All of which are extremely high performance SCSI devices. Therefore, it is impossible to get a feel of how the MAW3073NP performs in comparison to the rest by just the feel of it, from a human-experience standpoint. So I set up a benchmark to test the MAW3073NP in the perfect-world environment and obtain results to compare with its predecessors. Below are the two benchmark results conducted under two different SCSI controllers and motherboards running under Windows XP Pro SP-2.
HD Tach (long zone analysis)
using LSI22320 PCI-X 64-bit/133MHz SCSI controller on Tyan Thunder S2885ANRF motherboard
Burst Rate: 236.7MB/s (higher is better)
- Inner Zone: 92 MB/s
- Outer Zone: 50 MB/s
Ave Read: 74.6 MB/s
CPU Time: 1 %
Access Time: 7.5 ms (lower is better)
A burst rate of 236.7MB/s with an average read of 75MB/s is phenomenal. According to HD Tach's sequential read graph, the MAW3073NP put out an impressive constant slope graph from 0GB at 92MB/s to the middle zone, then slowly dips to 50MB/s towards the outer zone. The graph from 0GB to 73GB is pretty smooth indicating its constant read speed and is very consistent with its access speed.
HD Tach (long zone analysis)
using Adaptec 29160LP PCI-X 64-bit/66MHz SCSI controller on Tyan S2466MPX motherboard
Burst Rate: 133.3MB/s (higher is better)
- Inner Zone: 90 MB/s
- Outer Zone: 52 MB/s
Ave Read: 73.6 MB/s
CPU Time: 3 %
Access Time: 7.7 ms (lower is better)
Apparently, its limited burst rate of 133.3MB/s is evident due to a bandwidth cut-off by the Ultra160 controller. Based on my repeated tests, the transfer rate cannot reach maximum value of 160MB/s (even though the hard drive is fully capable); it has to stay well below that theoretical 160MB/s. And the reason is clear that when matched with Ultra320 controller, its impressive transfer rate has gone up to over 200MB/s as reported above with the LSI Logic LSI22320 Ultra320 SCSI controller.
With the exception of burst rate, HD Tach benchmark results from Ultra160 and Ultra320 SCSI controllers are roughly identical. Therefore, the MAW3073NP can be used on an Ultra160 SCSI controller without sacrificing much bandwidth.
SiSoft Sandra2005 gave a nearly identical drive index (DI) from both controllers. The Adaptec 29160LP on Tyan S2466MPX gave a drive index of 74MB/s. The LSI Logic LSI22320 gave 77MB/s, though the LSI 22320 gave a faster 3ms access time compared to Adaptec 29160LP at 5ms. Both controllers gave an impressively high sequential read and write with an average of 86.8MB/s. This makes the MAW3073NP perfect for a sustain, continuous reading and writing application.
The above figures from either controller (and its corresponding motherboard) gave impression that the MAW3073NP is greatly improved. For instance, its predecessor (the MAP3735NC) gave an HD Tach score of 216.5MB/s burst rate and an average read of 58.6MB/s with a 73MB/s inner zone and 40MB/s outer zone. Of course, the MAW3073NP is the third generation of Ultra320 and the MAP3735NC is the first. Therefore, it has improved greatly in terms of read/write and access time. The second generation of the Fujitsu's Ultra320 is the MAT series. But since I have never tested it, I cannot really make any comments.
Fujitsu SCSI hard drives have improved greatly in all aspects -- not just read/write and access time. Having used SCSI drives from different manufacturers allowed me an opportunity to make comparison. And, as a result, Fujitsu is my first choice for SCSI (second is Maxtor, and third Seagate).
During startup, when the controller "spins" the device, the MAW3073NP gives a soft low pitch sound almost unnoticeable. It can be heard only if I pay attention to it. By comparison, its noise is much lower than my IDE Western Digital and old Maxtor hard drives that give an out burst of high pitch sound. During accessing, reading and writing, the MAW3073NP is almost silent that often blends in with other noise in the system. Its operation noise seems to be more quiet than the MAP series.
In terms of heat, the MAW3073NP is also well improved. Its heat dissipation is quite low; it is only a tiny bit warm to the touch. My 250GB SATA drive seems a bit warmer than the MAW3073NP; and the 250GB SATA is a data drive that spins only occasionally, and both drives sit next to each other with a 92mm fan blowing onto them.
The difference between IDE (or SATA) and SCSI drives is shear speed. In an IDE-device system, it seems to me that there is always a lag in response between the time an application is launched and ready for use. In a SCSI system, it seems that everything responses instantly; it's because the system is able to access and receive data at a greater speed.
The MAW3073NP seems (and is) fast in everything, even disk defragmenting. I used to wait about an hour or so to defragment my 94GB partition in a 250GB SATA drive (and only about 40 GB of the 94GB is being used). And we already know that once disk defragmentation is started, and depending on which partition or disk is being defragmented, we virtually cannot do anything else, unless we have multiple disks and separate partitions. If you have only one drive, you are stuck until the system is done with its disk defragmentation. This is one good reason to regularly perform disk defragmentation and not wait for it to accumulate that will take a much longer time to defragment the disk. For SCSI hard drives, it seems that disk defragmentation time is reduced to a fraction of the time it normally takes to do on my SATA partition with similar amount of fragmented files. For me, time is very valuable, and SCSI saves me a lot of time.
I have been using SCSI devices for some time now. And I have found it to be very reliable; I have never lost any data. Of course, any technology is bound to failure, but with Fujitsu SCSI drives, I have yet to experience bad sectors that lead to a loss of my valuable data. However, there was one occasion where I encountered a RAID 0 problem with two 36GB MAP3367NP hard drives configured as striped. Since I often monitor the status of my hard drives, I quickly correct the problem before it escalates. Furthermore, since my RAID 0 setup does not store permanent data, the problem is not even a serious one. In fact, the problem is not really a technical one that results from a bad hard drive. It seems that my two drives do not have matched storage capacity (one with 33.91GB and the other 34.22GB usable space). And sometimes the left over disk space seems to cause the drive to want to drop itself from dynamic setting. After I delete the volume and recreate a new one, the two drives work as good as new again.
I believe the problem with this RAID 0 was perhaps because I imported it into a different Windows XP installation during my various testings. Importing the dynamic setting back and forth between various operating system (without performing a clean soft RAID 0 setup and I don't want to do that because data will be lost) may have disrupted the dynamic coding in both drives to respond to a RAID 0 setting improperly. Now that I permanently use the RAID 0 on my workstation system (where it belongs), everything is normal.
As for the MAW3073NP working as a single drive/partition, it has been phenomenal. Backed by a five-year warranty, the Fujitsu MAW3073NP provides a dependable and confident hard drive.
Different SCSI Controllers, Different Results
It is worth to mention that different SCSI controller yields slightly different results. My test results show that Adaptec 29160LP controller gave a higher buffered reading speed than the LSI Logic LSI22320. For instance, Adaptec 29160LP (using PCI-X 64bit/66MHz) gave a buffered read of 123MB/s with access time of 6ms for a MAP3367NC drive. When I used this same drive on my LSI Logic LSI22320 (operating at 64bit/66MHz) controller on the same motherboard (Tyan S2466MPX), I obtained a buffered read of 54MB/s with an access time of 4ms.
I have never had an opportunity to sample an Adaptec Ultra320 64-bit/133MHz PCI-X card. I suspect that its result would be somewhat better than what I obtained using LSI22320 Ultra320 PCI-X 64-bit/133MHz card. However, since LSI22320 gave a faster access time, it seems to provide a nice compensated benefit.
However, under a PCI 32-bit/33MHz, the result decreases noticeably. For example, HD Tach gave the following results for MAW3073NP when used on Adaptec 29160N (32-bit version) on Tyan Tiger S2875ANRF motherboard with PCI 32-bit/33MHz.
Burst rate: 93.4 MB/s
Ave Read: 64.4 MB/s
Access Time: 8.2 ms
Clearly, the MAW3073NP performs at a noticeably reduced speed under the standard PCI 32-bit/33MHz bus, but nevertheless still much better than IDE or SATA drives in my possession. This clearly indicates that even if you cannot afford a high-end motherboard and Ultra320 64-bit/133MHz SCSI controller, it is still possible to benefit the MAW3073NP's performance over the standard IDE device. But it doesn't make sense to spend $180 for a drive and not take full advantage of its great potential.
With a burst rate of 236.7MB/s (using Ultra320 SCSI 64-bit/133MHz controller), the MAW3073NP is quite a formidable SCSI hard drive. Its low noise and low heat add great features to the unit, especially for a SCSI drive, and definitely due to its fluid dynamic bearing technology. Based on its benchmark score, it rightly deserves to be the successor of its first generation, the MAP series.
As an active data partition, its 73GB storage space is quite practical for my applications. If used as a system drive, it might be over-killed for its unnecessary large space. For a system drive, a 15K rotational spindle speed SCSI drive introduced at the same time as the MAW series, the MAX series would make an excellent choice (even available in 36GB). I am currently using the first generation of 15K RPM SCSI drive (MAS3367NP) for my system partition and impressed with its performance. I can speculate one point that for gamers who demand swift responses in a system, there is nothing better than to have SCSI drives for the job.
Specifications: Fujitsu MAW3073NP
Interface: SCSI Ultra320 68-pin
Spindle Speed (RPM): 10,025 RPM
Ave Seek Time: 4.5ms
Ave Write Time: 5.0ms
Ave Latency Time: 2.99ms
Form Factor: 3.5-inch
Life Expectancy (MTBF): 1,200,000 hours
A note on model number identification. Here is how Fujitsu's MAW3073NP model number is deciphered.
AW = 10K RPM
3 = form factor (3.5-inch)
073 = storage capacity, 73.5GB (note, Fujitsu keeps using a four-digit format, and its fifth digit--5--has been dropped)
NP = Low voltage differential 16-bit SCSI Ultra320 68-pin connector.
AP = 10K RPM
3 = 3.5-inch form factor
375 = storage space, 73.5GB
NC = Low voltage differential 16-bit SCSI Ultra320 80-pin
MAS3367NP, MAU3036NP, MAX3036NP
AS, AU, AX = 15K RPM
367 = 36.7GB
036 = 36.7GB
NP = 68-pin
AU = 15K RPM
300 = 300GB
NC = 80-pin
You get the idea...