by Sergey Romanov
07/22/2004 | 10:41 AM
The life of a hardware tester is never dull. First, Maxtor pleased me with an unpredictable variety of the stuffing in its products; then some inexplicable surprises came from IBM. Samsung has remained an island of stability but it doesn’t remain one anymore! And really, the previous SpinPoint P40 series was unassuming and simple and yet very reliable. It was a kind of classics of hard disk drive making. And we might have expected the same from the new SpinPoint P80 generation, yes? None of that!
<%BANNER[article]%>The fierce competition makes the manufacturers go for various tricks to reduce the manufacturing cost of their products. There is a limit to simplifying any design, beyond which no hard disk drive would function properly, so they have to resort to the reduction of overhead costs. One of the most important operations performed over an assembled hard disk drive is factory formatting, the successful accomplishment of which determines the consumer qualities of this particular sample. For example, if one “head/surface” pair doesn’t comply with the specified characteristics, they just disable this pair to output a drive of a smaller capacity. If all the surfaces don’t fit into the norm, the device is defective, i.e. the manufacturer suffers a straight loss.
Of course, the manufacturing companies sought to avoid the situation when an already assembled device went to the trash bin; for example, they started pre-selecting heads with desired characteristics. Another approach, pioneered by Maxtor, employs formatting of disks for a smaller density. As you may remember, the last DiamondMax series used platters differing in their track as well as bit densities. It seems unlikely that different-capacity models of the one and the same series should be made of different parts, whereas formatting of platters for a unique density solves two problems at once: they can roll out junior models, much demanded by the market, and also utilize samples that turned to be non-operational at the maximum data density due to some reason.
Driven by this need to economize, Samsung has advanced further in this direction. This manufacturer developed a technology to format each head/surface pair for an individual capacity. And it seems like the bit density (sectors per track) and the track density (tracks per platter) may both vary on each surface of a drive! IBM employed a similar technology in its latest mobile HDDs, yet they only matched the number of sectors to each head and each density zone, but here we can even have different number of tracks! It doesn’t prevent the drive from functioning normally due to “per-track” rather than the “per-head” translation of the linear address (at sequential reading the heads move through the entire zone on one surface and then switch to another).
You may recall that this translation method was employed by the now-retired Fujitsu in its IDE drives, and it brought about the legends about their super-reliability. The trick was simple: if one of the heads of a Fujitsu drive failed, some data could still be restored in most cases, while “per-head translation” drives needed a complex operation of replacing the heads unit, without any guarantee as to the outcome.
The individual formatting technology was first used with SpinPoint V80 and P80 series, announced back on June 13, 2003. We will talk about the latter series, with 7200rpm spindle rotation speed, today.
The low-key case design has never changed for the last half a decade…
…only the electronic stuffing is different:
In the SpinPoint P80 series, Samsung for the first time realized such innovations as 48-bit LBA translation (BigLBA as Windows terms it) and faster versions of the interface (ATA/133 and SerialATA), thus creating some headroom for the future. Also new in Samsung’s drives, the allowable amount of onboard cache memory has increased from 2MB to now-standard 8MB, made possible by the multi-purpose 88i5522 controller chip from Marvell and its later 88i6522 modification that we know by the SpinPoint PL40 series. As you may recall, the PL40 was the leader among its peers in a majority of test exercises. That controller was also employed in V80/VL40 families, and its predecessor 88i5520 had worked in the SpinPoint P40 and V60, also very successful families.
The new generation of controllers features a higher read/write channel bandwidth (1Gb instead of 800Mb) and support of UltraDMA mode 6, so I haven’t found any qualitative changes. The specifications of the current and previous families are listed below:
Power consumption issues have grown very important nowadays, so I wanted to compare the spin-up current and the consumed power of the main rivals. Regrettably, Western Digital hasn’t published the characteristics of its new models, so I didn’t include them into the table.
Against the common opinion, Samsung first used fluid dynamic bearings in the second version of the SpinPoint P40 series rather than in the described family. Such devices had the letter “A” in their model name like “SP40A2H” instead of “SP4002H”. Setting the SpinPoint P80 against its predecessor, we see the mechanical characteristics like seek time unchanged. The noise is much lower now, but the power consumption grew in all operational modes. The last fact is nearly fully explained by the use of those fluid bearings. Anyway, the SpinPoint P80 remained among the most power-economical HDDs, especially at seek operations. It only loses to the new product from Hitachi, and only at “doing nothing”. Besides that, the spin-up current consumption is reduced, and that’s good if you’ve got several hard disk drives fed by a not-very-powerful PSU.
The acoustic characteristics of the SpinPoint P60 are approaching the ideal; it is no inferior to any of its competitors in this aspect. Moreover, no one can challenge Samsung’s superiority in the “fast” seek mode, for which I put the values into the table above. The patented NoiseGuard technology does wonders, achieving an excellent seek time by practically noiseless positioner movements. I should confess that some users reported a noticeable vibration of SpinPoint P80 drives that would set the whole PC case abuzz, but this tremble is only felt in cheap system cases of thin tin, or if several same-type drives (7200rpm) are in use. Otherwise, this problem is not observed.
Now we can move on to performance characteristics. We see that the average access time of the Samsung drives is worse than that of Hitachi’s products (Hitachi inherited high-performance mechanics from IBM) and Seagate’s (this is only true about SATA models from Seagate, see this article for details). Yet, Samsung surpasses all its rivals in the track-to-track seek time, and this parameter has a greater impact on the performance than the average access time, since there’s a high locality of data on the disks of a home computer and “far seek” occurs less often than track-to-track movements of the heads.
The data density parameters are average with Samsung. Maxtor, on the one hand, went for an extreme, increasing the bit density, while Seagate, on the contrary, improves the track density. Samsung’s characteristics are more balanced, but we shouldn’t forget about the above-mentioned technology for choosing an individual density for each head/surface pair, making such comparisons incorrect. Our experiments suggest that the SpinPoint P80 series, like Maxtor’s DiamondMax Plus 8/9, can include models on 80GB as well as 60GB platters. Thus, the skipped-over SpinPoint P60 series is now “combined” with the P80. When the 80GB platter is redundant – for 60GB and 120GB models – they just use a lesser-density platter of 60GB capacity.
I’d like to offer you the zone maps of several samples of SpinPoint P80 family drives, given to me by Mikhail Mavritsin a.k.a. hwdiver. For better readability I represent them as graphs for each surface independently. Each of the tested drives had 24 density zones, but the zones themselves vary considerably.
This sample has a lesser track density on the first surface than on the second one, while the number of the tracks is identical. Thus, the capacity of one surface is a little below 40GB and of the other – a little more. The two give us the required 80GB in total. Comparing the total number of sectors on the zone map with the number reported by the HDD itself, we learn that about half a percent of capacity vanished somewhere. These sectors must have been left as a reserve for replacing defective sectors.
The different surfaces of this sample of the SP1203N have practically the same sector density, but the required capacity of 120GB had to be accumulated in some other way. The second surface has the same number of tracks as the previous tested sample; the first surface has fewer tracks (but the sector density is maximal); the third surface has more tracks.
We see some original combinations more in this biggest-capacity model. Two surfaces have an absolutely identical format, exactly coinciding with the format of the second surface of the previous sample. One more surface has a slightly lower sector density, and third has record-breaking sector and track densities!
Overall, you have seen three variants of the possible amount of tracks (80033, 83153, and 86448) in three samples of the SpinPoint P80, and numerous variants of the sector density, among which we can single out the most common one. Thus, Samsung does use adaptive formatting in the latest generation of its hard disk drives. With this technology, each density zone may vary in the number of sectors as well as tracks, maximizing the use of the platter surface without compromising data reliability. Some discernable regularity makes us think that Samsung uses different track densities (in tracks per inch – TPI) for each surface, but we haven’t yet received an official confirmation or denial of this hypothesis.
Lastly, I’d like to dwell on another HDD characteristic, which is directly linked to reliability but is often left in the background – ECC.
Besides advanced data coding methods (for further error-free decoding), hardware and software ECC methods contribute much to the reliability of hard disk drives. Each data sector has traditionally been including a special ECC field, which helps to restore erroneously read data. The use of ECC often helps to read the data even if the read/write head couldn’t position precisely on the track and some of the information got lost. The employed Reed-Solomon code uses some redundant information to provide a higher data reliability than if they were just duplicated!
Regrettably, Samsung has now become as secretive about its products as the other hard disk drive makers, but a few years ago we could learn a few details about the structure of a data sector. Thus, the SpinPoint P20 had 480 bits of 10-way interleaved ECC and 7 bytes of CRC code (13% redundancy) per sector, which allowed restoring up to 10 three-byte errors or one 233-bit error “on the fly”, without compromising the speed. That’s hardware correction – the more sophisticated software processing of the same data allows restoring more, up to 60 bytes, i.e. each ECC bit allows restoring one erroneous bit of the original data at best. Besides the correction codes, each sector contained a 32-bit ESN field (Embedded Sector Number) for the correct work of the ECC generator, while CRC was added, most likely, for a faster identification of errors.
Each manufacturer takes a different approach to the ECC implementation. For example, IBM prefers distributing the code evenly along several sectors in order to have more chances to restore the data in case of a physical damage of a sector. This is the so-called interleave technique with respect to ECC. It is a pleasant surprise that Maxtor and Hitachi still give out information about the structure of the ECC block in their products. For the DiamondMax Plus 9 they specify 320 bits of non-interleaved ECC with cross-validation that allows correcting up to 15 errors, 10 bits each. The Deskstar 7K250 family has 4 interleaved ECC fields, 96 bits each, which allow correcting up to 5 errors to the total length of 153 bits.
With any implementation, the basic characteristic of ECC is the total number of wrong bits that the code can correct. In the characteristics table you saw above I post the specified data for “on-the-fly” correction, i.e. correction without compromising the read speed. As you see, although “thinner” than in the SpinPoint P40, the ECC correction in the P80 remains among the most powerful for today. Seagate (and truant Western Digital) prefer to keep silent about their technologies, but we can get an approximation of the data redundancy by correlating the specified recording density and the media data transfer rate with “end” data-transfer rates, which can be found in the same documentation or measured. Considering that Seagate nearly achieved the point of 60MB/s, like Maxtor but having a much smaller bit density, the amount of housekeeping data (including the ECC code) is very small in Seagate’s drives.
This testing session is wholly about the Samsung SpinPoint P80 series, i.e. I will be comparing models of different capacities, amount of cache memory, and interface (ATA/133 and SerialATA). Thus, we will see the impact of all such factors like the firmware version, cache buffer size and others on the performance of the drive.
We, at X-Bit Labs, have accumulated some statistical data and have files on the following models:
* Regrettably, we didn’t keep the record of the exact firmware version of that sample.
PATA-interfaced HDDs were attached to the Promise Ultra133 TX2 controller (BIOS 2.20.0.14, driver 2.0.0.29). SATA-interfaced drives were tested on the Promise SATA150 TX2 controller (BIOS 1.00.033, driver 1.0.0.27). We also include the results of one drive on the Promise Ultra100 TX2 controller, for illustrative purposes.
The testbed was configured as follows:
We used the following benchmarking software:
For WinBench tests we formatted the drives in FAT32 and NTFS as one partition with the default cluster size (FAT32 formatting was performed with Paragon Partition Manager). We ran the tests seven times each, chalking up the best result. The HDDs didn’t cool down between the tests. For the FC Test we partitioned the drive into two logical volumes, 32GB each. For IOMeter tests, we used Sequential Read, Sequential Write, Database, Workstation, File Server and Web Server patterns.
The File Server and Web Server patterns are identical to the ones StorageReview makes use of, and the Workstation pattern was created by us based on the access statistics for NTFS5 as given in the StorageReview methodology. This pattern differs from the server ones in a smaller load range and a higher percentage of write operations (a regular workstation must have much memory, a big portion of which is used by the OS for data caching).
As usual, we start out by analyzing the performance in extreme cases of the load. First, sequential reading/writing:
Click to enlarge
From the rows and columns of numbers, I’d like to present a few characteristic cases in the diagrams:
Early versions of the SpinPoint P80 on the ATA/100 controller had the same speed of reading small data blocks as the SpinPoint40 delivered, but the result is somewhat better on the ATA/133 controller. Later versions improved the speed further – this must have been due to the modernization of the controller, accompanied with certain changes in the firmware. Anyway, starting from firmware version “-23”, the read speed of the PATA drives grew on small data blocks. For SerialATA HDDs this improvement came with firmware version “-25” and later.
The physical properties of the platters and heads (number of sectors per track) put the limit to the read speed on large data blocks. Here, we see a sharp separation between models on 60GB and 80GB platters. Compare the results of the SP0612N, for example, with the SP1604N TM-23, which developed the highest data-transfer speed due to the fact that all of its four surfaces had the maximum and practically identical bit density. We found a smaller-capacity platter in both samples of the SP0612N, while the 120GB model can have 60GB as well as 80GB platters. In today’s tests, the SP1213C and the SP1213N represent the “faster” platter, while the SP1203N – both variants. Note that Samsung always denoted the number of the operational surfaces with the last digit of the model number, but this time the model is the same, but the platters are different!
The firmware influences the write speed less, but the tendency “higher version means higher speed” persists. Any model of the SpinPoint P80 series surpasses its predecessor on small data blocks, probably due to the faster controller, but the maximum write speed depends heavily on the platter density. As you see, a drive of the same capacity can be writing at a maximum speed of 42MB/s or 56MB/s or something in-between like 53MB/s – that’s the matter of chance. Certain lucky specimens achieve a data-transfer rate of 58MB/s. The difference between 42 and 53MB/s is quite perceptible and annoying. Unfortunately, you cannot tell between Samsung HDDs on different platters by reading their markings, so you only have to rely on your lucky star when shopping.
Now, let’s examine the average access time at reading and writing, taking two extreme cases of the Database pattern (0% writes and 100% writes) under a load of 1 request.
Click to enlarge
The 120GB models are the best here, and they lined up in order of diminishing bit density. By increasing the bit density, it is possible to decrease the number of tracks, necessary to reach the desired storage capacity. The reduction of the number of tracks is not necessarily compensated by a higher track density, thus the zone in which the heads are moving becomes smaller, resulting in a better average access time. As you see, the theory conforms to what we see in practice, but there’s also another factor interfering. A closer examination of the results suggests that the newer samples of the Samsung SpinPoint P80 have a higher average access time, but it’s difficult to determine the reason for that. Overall, we can say that the average seek time hasn’t greatly changed since the SpinPoint P40.
Except for the “strayed-away” SP0612N and SP1213N with new firmware, all SpinPoint P80 drives have learned how to write random-address data faster. Contrary to random reading, small-capacity models win the test, and the SerialATA interface works somewhat worse then the classic ATA, while the bigger buffer doesn’t affect the average write speed in the slightest. The reason for that lies in Samsung’s being very discreet with deferred writing.
What is deferred writing? The hard disk drive reports a success of a write operation to the system before this operation is actually completed. So the system wastes no time waiting for the drive to find the necessary track and the necessary sector. In other words, the drive “deceives” the OS, because the data may not be actually written, while the system thinks it’s Ok. That’s why Windows 2000/XP has an option of disabling deferred write for the hard disk drives.
To minimize the possible data loss, Samsung uses the following technique (I suspect some other manufacturers do the same): a random write command is deferred for a time, not more than necessary to find the desired sector. The subsequent read and write requests wait for the operation to complete, save for the case of writing into a neighboring sector, which will also be deferred (added to the first one). In the world of central processors, this is called Write Combining. The natural drawback of this scheme is its inability to optimize the route of the heads, while its natural advantage is an acceptable effectiveness in real applications and a low risk of loss of the written data.
We find the lazy write efficiency coefficient by calculating the ratio of the average read access time to the average write access time:
A result of more than 1 is indicative of the availability of lazy writing; if the number is over 1.5, the efficiency of lazy writing is very high. The results of this test are usually directly dependant on the HDD’s firmware, but we can’t find any clear regularity in the present case.
The realization of lazy writing with the above-mentioned limitations doesn’t allow the Samsungs to reach a coefficient of 2 or thereabouts as we saw with some Western Digital models, but in real life thanks to Write Combining this lazy write still can save much time, by the OS’s measures.
The Database pattern helps us check how well a hard disk drive handles a stream of write and read operations:
Click to enlarge
The diagram contains the best and worse results under the linear load; all the other models drew graphs that have a similar shape and are very densely located.
That’s Samsung’s way of dealing with writes in combination with reads. The firmware has no problems with writes, increasing the performance proportionally to the writes percentage. I selected all samples with “23” in the firmware version here; as you see, the same number doesn’t promise the same behavior. I don’t know exactly what this number means, but I have some thoughts about the value of the letters. The first letter changed with the modification of the controller chip, and the second is clearly linked with the HDD capacity:
So, Samsung only distinguishes between major differences by the firmware version (the first letter), and even that is not always true. The second letter denotes the HDD capacity, and the digits stand for some mysterious parameter, which affects nothing.
It’s hard to come up with a method for comparing the overall performance of the drives, so I took it easy and averaged the results for each load. Here’s what I got:
We see that the difference in performance within the series can be as big as 13%, and that’s quite much. What’s funny, you cannot tell beforehand what drive you get – faster or slower than the average level!
Click to enlarge
First, let’s simulate a workstation. This pattern features short request queues and a high percentage of writes.
All in all, all the drives handle this task confidently enough: the difference in performance of different samples of the SpinPoint P80 series, without the latest specimens of the 60GB and 120GB models, is only 3%. As usual, a big cache buffer plays no positive role for IOMeter. Fortunately, it doesn’t play a negative role here, either.
What’s curious, a drive on the UltraATA/100 controller took the first place, outperforming itself on the UltraATA/133 controller. It means that Samsung has no special optimizations for the ATA/133 interface and you can use the new drives with ATA/100 mainboards without worrying about losing any performance.
The writes percentage is lower and the load became higher, but there’re no changes in this server pattern – the gaps between the participants of the test have even got smaller.
There are no write operations here and the previous SpinPoint generation suddenly gains the upper hand. The new revision of the SP0612N and the SP1213N cheered up and took positions in the middle of the diagram, while the ATA/100 interface lost its ground somewhat. Still, there’s no great difference in the results – Samsung’s HDDs have traditionally been strong in IOMeter.
On the contrary, Samsung was never brilliant in WinBench. Is it different with the new series?
I’ll start the today’s session in WinBench in a slightly uncommon way. Each drive of the SpinPoint P80 series is a unique product – the factory formatting assigned an individual zone map to each of them – so I offer you typical read graphs with my comments.
SP0612N
You can clearly see two surfaces in this graph, each of which has its own zone configuration. The average speed conforms to the overall data density, 60GB per platter.
The first sample of the SP1203N model produced the following graph:
SP1203N
The read speed is just slightly higher than that of the SP0612N and I’m about sure that this sample uses four surfaces, rather than three. Since the sector density has grown, it took fewer tracks to amass the desired storage capacity – you can see this by comparing the read speeds at the beginning and end for this and the previous models.
SP1213C
This sample obviously uses higher-capacity platters – its read speed is about 10MB/s higher than of the previous sample and we can discern three platters of a varying density.
SP1604N
Lastly, there’s a sample, whose all surfaces have a similar zone distribution. Without the characteristic “jaggedness” of the graph, it would be an ordinary last-gen hard disk on 80GB platters.
Note that the graphs serve only illustrative purposes and you shouldn’t associate the SP1203N model, for example, with reduced density. Most drives really have 80GB platters, but sometimes you can come across a slower sample. In 160GBmodels, however, you are sure to meet no low-density platters at all.
To end this section of the review, I offer you a table with estimated platter capacities for each of our samples of the SpinPoint P80:
Thus, only three out of thirteen drives have small platters, and both SP0612N samples are among them.
Now we can plunge into an analysis of Winmarks.
Click to enlarge
Back from the example of the WD Caviar SE, I know that 8MB-buffered models have a hefty advantage at office applications. The big buffer may account for a 25% performance gain – a nice bonus! The SerialATA drives perform like the PATA ones here, while the ATA/100 interface is slightly better than ATA/133. This test is rather indifferent to the linear speed, so the previous generation of drives have just a little worse results.
Advanced Visualization Studio, as far as I could see in previous test sessions, is highly sensitive to the deferred write algorithms. There’s no simple explanation of the results: the latest firmware for the SATA models did well in FAT32, but lost to the previous version (24-th) in NTFS. The previous generation did unbelievably well in NTFS, occupying the middle of the diagram, but suffered a defeat in FAT32 where the cluster is bigger size and linear speed matters much!
FrontPage is susceptible to Windows’ and to the controller driver’s caching, so the speeds exceed the theoretical peak of the interface bandwidth. There’s no sense in trying to explain this mess. Moreover, the difference between the modals is small enough.
Microstation produces a sharper picture. The ATA model with a big buffer is in the lead in FAT32, having a 25% higher performance than the model with an ordinary buffer. This gap diminishes in NTFS, and the SATA models come ahead. The SpinPoint P40 again gives out a sparkling performance, while the firmware version 25 for the SATA models is better than the previous one.
You may remember that Photoshop favors linear speed and is indifferent to the amount of the HDD’s cache memory. Here, the bigger buffer provides a 15% performance gain both in FAT32 and NTFS – that’s just fantastic! If we put aside this phenomenon, the position of a drive in the table is determined by its linear speed.
It’s also predictable in Premier: there’s a strong dependence of the results on the HDD’s linear speed and capacity, although the two SpinPoint P80 of 60GB capacity suddenly woke up in NTFS. There’s again parity between PATA and SATA interfaces.
SoundForge has always been sensitive to deferred write algorithms as well as to the linear speed. This subtest became a triumph of the SerialATA interface at large, and of the firmware version 25 in particular – 15% advantage over the ATA counterpart, which itself has a wider gap to its analog with the standard buffer, in FAT32! Otherwise, it’s all expectable and determinable by the linear speed.
The last subtest brings no surprises: ATA and SerialATA are equals, a larger buffer results in a 25% performance gain, linear speed is important.
We witness a similar situation in this overall performance rating – I think there’s no need to comment on it. Instead, I want to say that WinBench, in spite of its respectable age, hasn’t lost its qualification yet and remains a good judge of the performance of modern HDDs – you just need to examine the subtests independently. Of course, the amounts of data WinBench uses are small by today’s standards, but it doesn’t diminish its value as of a firmware test. That’s why we continue using WinBench in our reviews.
The WinBench results suggest a certain progress in firmware-writing: higher versions make the SATA models perform faster, other factors being equal. For the classic PATA drive, the latest firmware sometimes worsens the performance (we’ll talk about that a bit later). HDDs with smaller platters perform noticeably worse than their “full-size” counterparts, and the big buffer gives a performance boost, although not that strong as I saw with products from the competitor companies.
I can also come up with a supposition about the excellent results of the 8MB-buffered models at reading. I think an improved caching scheme makes the trick. My experiments suggest that 2MB models can support up to 7 parallel read streams, while 8MB ones – up to 25, and the size of each cache segment reaches to 936KB! Such advanced cache segmentation allows the HDD to effectively perform anticipated reading: irrespective of the sequence of the requests from the OS, reading of one file (a sequence of sectors) is not interrupted with reading of another. Moreover, the dependence of the disk performance on the file system’s fragmentation level is diminished.
As usual, the tests of real operations with files of varying sizes will help us make our final opinion.
Click to enlarge
Click to enlarge
Step 1: we create (i.e. write) files on the disk.
The SpinPoint P80 handles the problem of uninterruptible data writing just excellently. The bigger-buffer models leave their value counterparts behind on small- and average-size files. The newer firmware versions take effect with these drives, making them work faster with small files. For the 2MB-buffered models, the firmware version doesn’t matter much. There’s a stable advantage over the previous generation, which amounts to 90% at best.
As usual, reading of files is less dependent on the cache buffer size than on the data density. Among curious things here, note the low read speed of the SATA models with very small files.
The speed of copying large files (digital video, for example) is evidently limited by the physical speed of the drives: the models with a reduced recording density are lagging far behind their bigger-platter mates. The 8MB-buffered models occupied the top of the diagram: they are faster on all file types, whatever their interface. They achieve a maximum speed of 25MB/s, and all the models reach as high a speed as 10MB/s and more at copying numerous small files.
Now that we have increased the distance between the departure and destination points (we’re now copying from one logical volume to another), the speeds diminished, although slightly so. The drives mostly kept their positions, too. The PATA and SATA models with an 8MB buffer are in the lead: all the models of the SpinPoint P80 family outperform the previous generation, and we see no differences between the firmware versions.
Here I’ll try to summarize my experience with the SpinPoint P80 hard disk drive series from Samsung.
First of all, I would like to draw your attention once again to the possibility of purchasing a “pig in a poke”. According to Mikhail Mavritsin, who provided me with detailed information about the zone distribution of the reviewed drives, up to 10% of Samsung’s P80/V80 families come out with a reduced data density. Our tests prove that the biggest share of those 10 percent falls on the junior 60GB models and, to some extent, on 120GB models. Theoretically, 80GB models can also use low-density platters, but we haven’t yet encountered them. You are guaranteed to get the maximum performance by purchasing 160GB models – to reach this capacity, maximum-density platters are needed. Logically thinking, I can assume that more expensive models (SP1213N and SP1213C) also use platters of the maximum density. At least, all 8MB-buffered samples delivered the maximum performance in our tests.
Overall, the speed of the SpinPoint P80 is at a very high level. There’s been progress made compared to the previous series: the 8MB-buffered models with the latest firmware versions have nearly doubled their speed! In fact, the larger data buffer of the SpinPoint P80 brings about some gains, but not in the tests where I had expected to see them. For example, Samsung left all its competitors behind in the copy speed long ago, and it’s probably very hard to improve anything in this respect. Anyway, the 8MB-buffered models are overall about 10% faster than their 2MB counterparts. Moreover, the new firmware versions increase the performance of the top-end PATA and SATA models in the first place.
Considering the implementation of the SATA interface through a Marvell converter, I didn’t expect the SerialATA drives to be a revelation. None happened. The SP1213C and SP1614C were faster than the SP1213N and 1614N in some tests, but slower in others, and overall they were equal. In fact, this difference can be written off to the discrepancies between the two Promise controllers that I attached the HDDs to.
I couldn’t help testing the new Samsung drive on an ATA/100 controller. You have already seen that there’s no difference with ATA/133 in tests. In fact, the Maxtor D740X, the first drive to support the faster version of the protocol, remains the only device to be perceptibly faster with ATA/133.
I can’t say anything objective about the noise characteristics of the SpinPoint P80, since we haven’t got the equipment and methods ready yet, so I can only give you some subjective impressions. The fluid dynamic bearings considerably reduced the idle noise and there’s no that characteristic noise of the spindle rotating at 120rps at all. Some users complained about a strong vibration of the SpinPoint P80, but they were doing meekly enough in our testbed. As for the actuator’s buzz, Samsung silenced it long ago.
The thermal characteristics fully comply with the specifications: the SpinPoint P80 is usually cooler than analogous models from the competitors (with the same number of platters and fluid dynamic bearings), but ventilation is required for any modern drive, especially those consisting of more than one platter.
So, can I call the new HDD series from Samsung a victory? I think, yes. You could see from our previous reports about 80, 120 and 160GB drives from different manufacturers that the SpinPoint P80 contends for the leadership in performance and has no obviously weak spots, while the SpinPoint V80 isn’t at all inferior to models with a higher spindle rotation speed. The programmers have been working on the firmware, improving the performance as well as reliability (lower risk of losing data through deferred write). Overall, they rolled out a very impressive product that leaves no doubt about Samsung’s serious plans of becoming a leader in the market of hard disk drives.