by Sergey Romanov
07/29/2003 | 07:07 PM
Maxtor has not been long alone in its effort to bring a low-profile and high-speed (7200rpm) hard disk drive to the market. Surprisingly, it is followed by Samsung and not by Seagate, Hitachi or WD!
Customers are often wary about HDDs from Samsung, although there are few who do remember why :). Meanwhile Samsung engineers don’t waste their time, but keep on improving their creations. First of all, they eliminated the reliability issues we used to complain about, then did some thorough research and now Samsung HDDs are among the first to feature high capacity platters. Several latest generations of Samsung drives have been among the most noiseless and shockproof. The assortment is now extended with 7200rpm disks. The last thing to do is to optimize the firmware and we can see the results of software developers’ work: Samsung drives are getting ever closer to the leaders.
At a press conference held in April, Samsung representatives even expressed the intention of their company to become a leader of the ATA-HDD market by 2005! To uphold the claim, the company moved to 80GB platters and liquid bearings, implemented ATA/133 and SerialATA/150 interfaces in its new products, and once again extended the product line. Two more families, SpinPoint PL and VL were added to existing SpinPoint P (7200rpm) and SpinPoint V (5400rpm). The “L” in the name stands for “Low-profile”.
Just like Maxtor, Samsung started working on the junior members of the product line with a cool head, following the rule sounding like: “we’d better lose a little now, but rake in a lot later”. Why use a full-size case and an actuator developed for four heads in a product that’s in fact as big as a half of an ordinary HDD (junior models use only one side of a single platter)? Why save on development if you can save much more on manufacturing costs?
The new design came as a result of this approach:
As you see, the case has an original look; its height is smaller than that of the thin Maxtors: 17mm. By the way, it makes sense to compare the two immediate competitors.
According to the specs, Samsung PL40 looks just like Maxtor DiamondMax Plus 8, but better :). It doesn’t yield to the Maxtor drive in any characteristic, except somewhat lower maximum physical speed. It makes up for it with its higher durability, lower noise level and lower energy consumption.
The robustness and noiselessness of the new Samsung match the latest products from Seagate.
To form the basis for interesting comparison, we have picked up several HDD models with similar specifications from all manufacturers, except Western Digital. All recent WD drives which came through out test lab had very big capacity, which is mostly due to WD itself. The company has made a rule to debut with super-capacious novices, and to keep producing models of smaller storage capacities on older and smaller platters. A little while ago, we reviewed WD400JB, but it would be incorrect to include it into this review as it has 8MB of cache, while all other drives – 2MB. Besides, it costs more than Samsung PL40 and is based on a smaller platter.
So, here is the list of our today’s participants:
Numbers 1, 2 and 3 have full-capacity 80GB platters and one read/write head – they are going to show fair play. Number 4 is based on an out-dated 40GB platter, and number 5 has a 60GB platter “shortened” to 40GB. It’s going to have the best average access time and favor IOMeter tests, but this will make it more interesting to compare the new Samsung with it.
To make this article an easy reading, I refer to the drives by their family names. For example, Hitachi 180GXP rather than Hitachi Deskstar IC35L040AVV207-0.
Our testbed configuration looked as follows:
We used the following benchmarking software:
All drives that support “quiet/fast” operation modes were switched to the fast mode by means of Hitachi Feature Tool. For WinBench tests, the drives were formatted in FAT32 and NTFS as a single partition with a default cluster size. We used Paragon Partition Manager for FAT32 formatting. The benchmarks were run seven times each; the maximum result was counted in. The HDDs didn’t cool down between the tests. The tests in Intel IOMeter were run in SequentialRead, SequentialWrite, DataBase, WorkStation, FileServer and WebServer patterns. If you are looking for the detailed description of these patterns, please, see our previous articles.
HDTach, chastised but indispensable, opens up the show.
First of all, I would like to point out the level of CPU utilization. The UDMA/133 protocol perceptibly lowers the CPU utilization for Samsung PL40. We saw this with the Maxtor drive very rarely, I should say. However, the new model didn’t catch up with its predecessor, which is no surprise considering its higher linear speeds.
Now, let’s learn the physical characteristics of the drives.
The average rotational latency time is about 4.17ms for all participating drives. By subtracting this value from the measured time we get the average seek time.
The Seagate drive turns to be an outsider in terms of average access time, while Samsung PL40 did quite well and proved up to its specs. It is just a little slower than the predecessor. In fact, that is exactly how it should be: a low-profile drive must have a less powerful actuator, and the fashion for noiseless work brings about its own restrictions, too.
The next diagram shows the speeds of the drives.
As for burst read speed, Samsung PL40 is no record-breaker: working on UDMA/133, it is a tiny bit ahead of the UDMA/100 IBM/Hitachi disk. Anyway, the result is the same as that of the competitors, and better than by the predecessor. The new Samsung loses in average read speed to only the Maxtor drive. As for write speeds, we already know we shouldn’t trust HDTach here.
Now, let’s have a look at the internal platter organization of the drives. The diagram below shows their speeds converted into millions of bytes per second, as manufacturers specify them.
So, Samsung PL40 is a little slower on the edge of the platter than Seagate 7200.7, but a little faster in the two other categories. Its doubled data density resulted in a 25% linear speed growth. Still, the analogous platter from Maxtor is notably faster. However, high linear speed doesn’t necessarily mean higher performance at work. Moreover, it might have been achieved by packing less servo information per track, which usually leads to a reduced reliability and life term of the drive.
Summing up, I would say that Samsung PL40 won no nomination, but has never been the last. It shows stable and good results, which is very nice. Let’s see what it will show next.
The good old WinBench 99 is a hale man and knows a lot about “household” performance of hard disk drives.
Linear read speed graphs are given below (click to view):
Let’s start analyzing the WinMark tests for FAT32 file system:
IBM/Hitachi drives have been the favorites in WinBench for a few years already and this time is no exception. Strangely enough, Seagate drive won the second place. This is even stranger, as the sample we tested some time ago boasted no better performance than the old SpinPoint P40 that shows the worst result here. Seagate managed to pull up its drive to the leading group by simply changing its firmware version from 3.00 to 3.06. That’s really mysterious, but this review is not about Seagate :).
Samsung PL40 does quite good, outrunning in High-End WinMark its direct competitor – Maxtor DiamondMax Plus 8. To better understand why it happened, let’s consider the results for separate applications.
So, Samsung PL40 is far ahead of Maxtor DM+8 in SoundForge only. This test is sensitive to lazy write implementation. We thank Samsung for not following the distasteful practice of deliberately worsening the firmware features, so that they could distinguish between the product lines. We saw this in products from Maxtor, Seagate and WD. In all remaining tests SpinPoint PL40 goes neck and neck with the Maxtor, and they both do better than the overall leader (IBM/Hitachi) in Visual C++, Premiere, Photoshop and FrontPage.
But why do we explore performance in professional applications using FAT32? Let’s go over to NTFS.
The new Samsung is on equal terms with the IBM/Hitachi in NTFS. How did it manage that?
The IBM/Hitachi won in AVS/Express only (traditionally, by the way), but the Samsung paid back in Premiere. The Maxtor drive turned very slow in SoundForge once again, while Seagate HDD lost to Samsung all the sub-tests, except AVS/Express.
Here we are! Hard disk drives from Samsung used to be weak in WinBench, but now there is no trace of this weakness left. We have only the best or near the best results, and only the AVS/Express sub-test proved to be a hard nut!
Time to have some Intel IOMeter here. This powerful tool can give out indispensable info about peculiarities of a hard disk drive.
First of all, let’s check the drives’ ability to read and write data blocks of various sizes.
Maxtor is beyond any competition in sequential read, with the Samsung closely following. That’s quite good, as it left the IBM/Hitachi and Seagate as well as the previous Samsung model, in its wake.
The Maxtor drive was quite confident at writing, too, but allowed Seagate to defeat it starting from 8KB blocks on. Once again, a specimen from the first shipments of Barracuda 7200.7 was slower than Samsung PL40, which is ahead of both: IBM/Hitachi and its own ancestor.
The next pattern is both the longest and most informative, simulating work with an SQL database. The data block size is 8KB; the proportion of write/read operations as well as disk subsystem workload varies.
So, the results of this test can tell how well the drive manages random data block access.
The IBM/Hitachi HDDs have always been fast at seeking. Secondly, it is the only drive of the five that doesn’t use the entire platter area, so it has an advantage in average access time from the very beginning. If not for this, it would perform as fast as Samsung PL40. As for Samsung PL40 itself, it reads as fast as Maxtor DM+8, and writes even faster, having won the second prize here!
The ratio of average read access time to average write access time indicates the ability of the drive to accumulate and arrange write requests. In other words, it characterizes quantitatively the efficiency of lazy write algorithms.
And Samsung PL40 proves that it has the best lazy write algorithms of all participating devices. The IBM/Hitachi HDD follows right behind. The other Samsung comes the third, while Maxtor proved twice worse than the leader. As you remember, Maxtor hard disk drive was behind Samsung in SoundForge, too. As for Seagate, the situation is rather strange, really. The earlier sample with 3.00 firmware had 1.20 efficiency coefficient, better than the HDD with 3.06 firmware. However, the 3.00 HDD did considerably worse in SoundForge!
Now, let’s check the drives’ firmware and learn more about their “personality”.
When the workload is random, we have two leaders: IBM/Hitachi 180GXP and Samsung PL40. Samsung P40 is not bad either; the Seagate solution was perfectly indifferent to the type of workload, while Maxtor goes a bit down when write requests appear.
The next test will help to evaluate the “scalability” of HDDs firmware according to the workload. Random read comes first.
Here, average access time influences the result most. The IBM/Hitachi is the best, and Samsung SpinPoint PL40 is the third. Unlike the competitors, Samsung’s firmware gives up when the workload increases.
The next diagram shows averaged data for all read/write ratios, except the two terminal cases.
When read and write requests are mixed together, the IBM/Hitachi drive loses its leadership under linear workload. It loses to Samsung PL40 that even defeated its predecessor with a lower access time. I’d better not mention the outsiders: Maxtor and Seagate.
In case of 100% write requests the graphs take quite another look. The drives that feature lazy write functions manage data under low and medium workloads quite well, but starting from 32 outgoing requests it is the operation system that sorts the requests out and determines the overall result.
The shape of the Seagate graph tells that the drive does have lazy write function, but its efficiency is too low. I will try to explore this phenomenon and report to you in one of our future reviews.
Now that we have learned the main characteristics of the drives, it’s time to take a look at patterns that imitate real-life HDD jobs.
Professional use comes first. Imagine a workstation with an NTFS5-formatted hard disk where several simultaneously running applications create intensive traffic with a high share of write operations.
If the IBM/Hitachi drive didn’t have an advantage in the way of the shortened platter, both Samsung drives would be in the lead – they showed practically identical results.
Let’s calculate HDD performance ratings for this pattern with the following formula:
WorkstationRating = IOps(1) + IOps(2) + IOps(4) + IOps(8)/1.5 + IOps(16)/2
Maxtor and Seagate fell far behind under this type of workload. It would be interesting to see how efficiently they use the CPU resources. IOMeter calculates a special coefficient called CPU Effectiveness. The rating is calculated according to a formula with the same weights.
Once again, the IBM/Hitachi product is the first, the Samsung HDD – the second.
Another application field for a hard disk drive is a fileserver.
Maxtor and Seagate drives are both equally slow. The IBM/Hitachi solution is a confident winner, while Samsung PL40 is a little behind its predecessor.
The rating for the server use is calculated with a formula using different weight coefficients:
FileServerRating = IOps(1) + IOps(2) + IOps(4)/1.5 + IOps(8)/2 + IOps(16)/3
The IBM/Hitachi is inimitable as a server HDD – maximum performance plus minimum CPU utiliation. The results of the other drives do not show anything remarkable.
The last pattern imitates the workload of a typical web-server (when no write operations are involved).
This time the picture is different: the HDDs from Maxtor and Seagate have cheered up and outpaced Samsung PL40 under high workload.
The rating in this case is calculated with one more formula:
WebServerRating = IOps(1) + IOps(2) + IOps(4) + IOps(8)/2 + IOps(16)/3
The new Seagate is slightly worse in this test, while Maxtor DM+8 did otherwise and took the third place behind the IBM/Hitachi and Samsung P40.
On the other hand, the drives from Samsung were the best at effectively using CPU resources.
So, IOMeter patterns testify a progress of Samsung in HDD making. Only the failure in the WebServer pattern doesn’t allow calling Samsung PL40 definitely better than Maxtor DM+8. However, the new Samsung never was the last, and even won the WorkStation pattern.
The last test checks the drive’s file-copy speed and is the closest to reality.
When we developed this test set, we were aiming at two goals: to make the copy speed measurements automatic and to measure operations during the copy process separately (such as reading and writing files). We managed to achieve these two goals in the following way. First the test creates files on the HDD according to a certain pattern set by a special script, which provides the number of files to be created and their sizes. Then the saved files are read and copied within one partition and from one partition to another within a tested HDD. Of course, the time required to complete this entire process is measured. The results are given in the tables below:
To make the review more compact, I show only one graph if the results in NTFS and FAT32 are identical.
File creation (write) comes first.
A surprise from the start: Samsung is ahead (though sometimes slightly) of the traditional leader of this type of competition – Maxtor HDDs! Its results are the best among all participating drives. Note also that Samsung PL40 is up to 40% faster than its ancestor, P40. Recalling that the physical speed growth is only 26%, we have clear evidence that software developers did a great job on the firmware. It proves true not only in WinBench, but also in real-life tasks. Bravo!
One can do no more than one can (although we remember IBM/Hitachi drives do it). The Samsung drive lost to the one from Maxtor that has higher “raw” speed. Seagate solution had a good turn, too. Its results are close to those of Samsung PL40. The IBM/Hitachi follows close behind, while Samsung P40 is a quarter behind, as it should be.
Now, let’s go to copying …
I say it’s a shock! The new Samsung is faster by a half than the copy ex-champion, Maxtor! As for the remaining drives, the IBM/Hitachi did quite well, outperforming Maxtor on small and middle-sized files.
The Samsung is leading at copying both small and large files between two partitions. The previous model from Samsung was quick at copying, too, but the new one is better than everything I have ever seen.
Yeah, this part of the competition was the most spectacular. I guess there is no need to comment: a glance at the diagram is enough to tell the winner.
Thanks to its liquid bearings and house technologies of noise reduction, SpinPoint PL40 was the most noiseless, much better than an analogous drive from Maxtor. The spindle squeaked no louder than in Barracuda 7200.7, while positioning sound of the Samsung drive was much less perceptible. You can only hear it when the case is open. The buzz of the Seagate drive has more high-frequency components and is more irritating. As for shock-tolerance, SpinPoint PL40 can match the higher-priced Barracuda 7200.7, but beats it in performance!
The ATA/133 interface doesn’t bring any advantages to Samsung SpinPoint PL40, although it does load the CPU less than ATA/100. On both interfaces, the drive shows good results, and has no performance slumps we saw in the first ATA/133 drives from Maxtor. Another nice fact is that Samsung has grown past infantile sicknesses like fear of NTFS file system we saw in SpinPoint V60 and made a very fast hard disk drive. The new model is much faster than Maxtor DiamondMax Plus 8 and is on equal terms with the IBM/Hitachi. Although Samsung didn’t win all the nominations, it always stepped on the podium, and often outran its more renowned competitors.
Our today’s review produced a rather surprising outcome: the budget model from Samsung is an absolute champion among all (once again – ALL) hard disk drives we have ever tested in file-copy speed test. Even Western Digital drives of the JB series with a larger cache-buffer (which positively tells on copying speed) are left behind. And we have not yet tested the full-size SpinPoint P80, which should be even faster…