|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The “More”Syndrome
PC sales are in a rut. But its not for want of power. Even as AMD and Intel ship faster and faster CPUs, and as graphics cards eat up more power than the AGP port can deliver, hard drive companies have continued to deliver bigger and bigger — and now megacapacity — hard disks.
Much can be done with one of these huge drives — but which is the right one for you? Recently drive companies started shipping hard drives with 60GB per platter (or roughly 45 Gbits/square inch, which compares to about 33-35 Gbits/sq inch a year ago). Thats all we needed, as we decided that now was the right time to check them out.
But being performance-oriented, though, we bypassed the 5400RPM set and waited for the 7200RPM drives to arrive. So we gathered up four of the biggest, and fastest, drives using these new 60GB platters, and put them through our exhaustive tests to help you figure out what to buy.
Theres more to the hard drive decision than raw performance, though. All hard drives make noise, although theyre certainly less noisy now. For example, all the drives we tested offered fluid-damped bearings, which reduce idle noise substantially from past drives.
Another factor is reliability. Surprisingly, some drive makers have recently begun reducing the warranty period for desktop ATA drives from three years to one — not exactly a ringing endorsement of their products. The good news is that both IBM and Western Digital maintain the three-year warranty period for drives with an 8MB cache.
This hands-on review starts with a look at the specifications for the four drives we tested, then details our exhaustive performance benchmarks. If you just want to know what to buy, head right to the Recommendations and Product Summaries section, where we wrap it all up. Otherwise, lets take a look at the contenders in this head-to-head review.
The Contenders
We took a look at four drives, from the four main US manufacturers. (Note that IBM will be transitioning its desktop storage business to Hitachi in 2003). All these drives offer 60GB or more per platter. Three of the drives ship with three platters, though Seagates Barracuda V ATA maxes out at two platters. Note that the focus here is on 7200RPM drives, so we didnt look at drives with even larger capacities, since those are all 5400RPM drives.
Lets take a look at how the manufacturers rate their drives.
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
All these drives now come in configurations with 8MB of cache, which offers some additional performance over the more common 2MB buffers. Only the Maxtor offers support for the ATA133 interface– most drive companies are waiting for Serial ATA to make the shift to higher speed interfaces.
Now that weve seen the manufacturers own specs, lets move on to our testing.
Testbed and Benchmark Setup
We tested on an Intel Pentium 4 system, running an 845PE chipset with 512MB of RAM. Heres the actual system configuration:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
As a baseline, we benchmarked our winner in last years big drive shootout, the Western Digital WD1000BB. This drive, which is still available, offers 100GB in three platters (33.3GB per platter), and a 2MB buffer.
The one potential limiting factor was the ATA/100 interface built into the ICH4 I/O controller hub on the Intel motherboard. Given that we were running single drives, not RAID configurations, this really wasnt really a big issue, since the maximum transfer speeds off the physical drive surface for sequential reads on outer tracks is well under 100MB/sec. However, the extra burst speed in ATA/133 drives to and from their 8MB drive cache can help performance during repeated random accesses when data is resident in the cache.
The hard drives were mounted in a removable drive cage with four screws, and installed into a Chieftec aluminum case. The hard drive cage was firmly clamped into the case.
We ran three sets of benchmarks, plus two environmental tests:
- Winbench 99 version 2.0 Disk Tests. We ran the two Winmark tests (Business Disk Winmark and High End Disk Winmark. We also ran the CPU utilization, access time and transfer rate inspection tests. The transfer rate was normalized to 8MB (8000KB), as suggested by the designers of the Winbench 99 disk test — that is, the transfer rates for the applications playback tests were set to 8000 kilobytes per second.
- Multimedia Content Creation Winstone 2003, to see how the drives performed under a set of real-world applications, executing scripts that represent real user workloads.
- Business Winstone 2002, to test performance with a standard suite of desktop business applications.
- Thermal test. We collected thermal data using a Fluke 189 digital multimeter and thermocouple assembly. The thermocouple was attached directly to the non-PCB side of the hard drive. The ambient air temperature was also recorded.
- Audible noise under max load was tested using a Tenma 72-860 sound level meter. The microphone was mounted on a tripod six inches above the surface of the drive. For this test, we removed the drive from the drive cage. The drive was somewhat isolated from the test system with a long IDE cable and power extension cables (only for the sound and thermal tests). Ambient noise without the hard drive present was also recorded to establish a baseline for comparison.
Winbench 99 2
.0 Business Disk Winmark Test Results”>
At first, we thought something was wrong with the Seagate drive. We double-checked DMA settings, and everything seemed correct, but the Barracuda V ATA got a Business Disk Winmark score almost half that of the nearest competitor. However, the Seagate drive had a few interesting surprises up its sleeve as well soon see. At the other end of the performance envelope, the DiamondMax Plus 9 outpaced the nearest competitor, the IBM Deskstar 180GXP. Note that even the slowest drive here still ekes out a small lead over last years fastest drive.
One thing to keep in mind, though: weve heard rumblings from certain drive manufacturer technical reps, who prefer not to be identified, that their competitors actually detect popular benchmarks and tune their onboard caches to the benchmarks access patterns. However, given that the Seagate drives rated disk-to-buffer transfer rate, as supplied by the company, is lower than the competitions. This result turns out to be less surprising after all,
WinBench 99 2.0 High End Disk Winmark Test Results
Again, the Barracuda V ATA trailed the pack, though the difference wasnt as substantial as it was with Business Disk Winmarks. Note that the IBM drive took the lead, with the Maxtor drive coming in second. Its also worth pointing out that the older Western Digital WD1000BB is the slowest by a significant margin.
The WinBench 99 high end disk tests can be broken down by application type. Lets look at a couple of them.
In the playback of a Microsoft C++ 5.0 compilation, the Barracuda actually places ahead of the Maxtor drive. Compilers are constantly reading and writing from the disk in small chunks, so perhaps the firmware in the Maxtor isnt optimized for this type of disk access.
This chart shows the results of a playback of a FrontPage 98 authoring session. The Seagate wins handily. Again, the pattern is one of relatively small reads and writes, alternating. Interestingly, this is the only subset test in which the older WD1000BB drive scored well.
In this case, Premiere reads large chunks of data into memory, then renders out to disk in steady streams. The Maxtor drive seems to like this type of disk access pattern better than the other drives.
WinBench Transfer Rate Test
Results”>
The Winbench transfer test simply reads large blocks of data from the drive. As we can see, the maximum transfer rate, which occurs at the outer edges of the drive, reveals the Seagate drive performing poorly. This was suggested by our breakdown of the high end disk Winbench scores. Now lets look at the transfer rate from beginning to end.
Seagate Barracuda V
IBM Deskstar 180GXP
Western Digital WD2000JB
Maxtor DiamondMax Plus 9
Western Digital WD1000BB
We can now see that the Barracuda follows the same general pattern, but the entire curve is shifted downward from the other drives. Last years champ looks positively anemic compared to this years crop. Even the slowest of this years drives outpaces the 33.3GB per platter, 100GB Western Digital drive.
Winbench 99 2
.0 Access Time and CPU Utilization Test Results”>
While the average access times, as reported by Winbench 99, are pretty close, take a look at the CPU utilization scores. Suddenly the Seagate drive starts to look interesting — maybe it doesnt move large blocks of data as fast as the competition, but it uses much less of the CPU in the process.
Business Winstone 2002 and Multimedia CC Winstone 2003 Test Results
In our real world applications tests, things suddenly become less clear. The Barracuda drive still lags behind, but the difference in Business Winstone is about 10%, while the difference in CC Winstone is negligible. Again, the older WD1000JB staggers in behind the newer generation.
Environmental Tests
All of the disks ran relatively warm, but not nearly as hot as some of their predecessors. Its clear that fluid-damped bearing technology has progressed from the early days when some drives with fluid bearings would actually run quite warm. Still, if you were running a RAID array, youd want to have adequate cooling.
Interestingly, the IBM and Western Digital drives were the quietest when running at full bore — roughly 5dB above ambient. The Barracuda drive was surprisingly noisy, given the idle noise spec, but its still not as loud as some older drives. All these drives are quieter than earlier generations.
Analysis
These drives offer huge capacities. Even the lower capacity Seagate drive offers up 120GB — a capacity that would have seemed unbelievable only three years ago. However, big drives have become less of a luxury and more of a necessity in a world thats increasingly digital media centric. More and more users are storing digital photographs, ripping their CD collections, and editing the family videos downloaded from their mini-DV camcorders. Even games routinely eat up 1.5 gigabytes or more — Unreal Tournament 2003 consumes 3 gigabytes of disk space.
Media we use will eat up even more space in the future. Consider the new crop of 5 megapixel digital cameras. When you set the typical 5mp camera to “fine” compression, a 2560×1920 pixel image occupies in excess of five megabytes — and thats a single photograph. Imagine how much space all those vacation shots will take.
Its also clear that these newer drives outpace the performance of last years best drives. It seems obvious that higher platter densities will result in better performance — after all, more data is passing underneath the drive read heads each rotation of the disk — but its good to see that verified in the benchmarks.
Different drive companies clearly optimize their firmware and hardware for different access patterns. The Maxtor DiamondMax 9 seems to excel at transferring large blocks of data, suggesting that it might be a great drive for digital video editing or capture. Alternatively, the Seagate Barracuda ATA V might do well in a database environment, where small chunks of data are being read from and written to the disk. The Deskstar 180GXP and Western Digital WD1200JB seem to offer a balance between the two extremes.
Whats of more concern to the average buyer of desktop drives is the warranty issue. All of these drives have 8MB of cache, but both Maxtor and Seagate only offer one-year warranties. IBM and Western Digital also offer meager one year warranties, but on their drives with 2MB of cache. The drives with 8MB buffers are considered to be a premium product, and so ship with three year warranties. Given that IBM had some reliability issues with their Deskstar 75GXP line, its something of a relief that IBM will honor a three year warranty on the high end of the line.
All these drives are also offered in models with smaller capacities. Seagate only offers products with one or two platters, while the other companies offer drives with one, two or three platters. Note that not all of these drives are necessarily of the same type or have similar features. For example, IBM only offers the 8MB buffer on drives with two or three platters — the single platter version only has a 2MB buffer and only one years warranty. Western Digital has 8MB buffers from 40GB on up — but not all are the same generation. For example, the 120GB drive has three 40GB platters, while the 180GB and 200GB drives have 60GB or more per platter. The higher areal density can actually improve performance. The bottom line: check the actual spec of the drive you buy before plunking down your hard-earned cash.
The Seagate Barracuda ATA V is something of a paradox. Its slow on the standard benchmarks, but seems to excel at applications that read and write small chunks of data. Thats a key consideration for some application types. The Seagate drive also has a good idle sound spec, and the Seashield is a handy feature if youre upgrading the drive yourself, as it covers up all the sensitive electronics.
The Maxtor DiamondMax Plus 9 seems to be optimized for large data transfers. This would make it a very good drive for digital video capture, audio work or any application that reads and writes large streams of data. This would also make it a good drive for gamers, who are often waiting for large levels to load.
Recommendations and Product Summaries
Based on our observations, heres our bottom-line recommendations for various usages:
-
Streaming Media Content Developers, Digital Videographers and Streaming Media
The Maxtor DiamondMax Plus 9, also know
