The Samsung 2 TeraByte HD204UI runs at 5400 RPM for quiet low power operation, a type that best fits a USB case or NAS. You would not use this type of disk for most servers or a desktop workstation because of the lower rotational speed and the poor write speed cased by the
Advanced format. I ran some benchmarks to see what the disk could do. 5400 RPM disks are not fast enough to fill a SATA II connection and definitely do not need SATA III benchmark. I ran a benchmark using a USB 2 enclosure plus a USB 3 benchmark.
- Advanced format
- Power requirements
- USB enclosure benchmark
The minimum write rate for this disk is randomly bad, a problem caused by the disk using the unadvanced
Advanced Format system where 512 byte sectors are stored in 4096 byte sectors. Normal disks write data in an ancient format where the data is cut up into 512 byte sectors. In the 1980s I was advocating 1024 byte sectors but the industry did nothing. In the 1990s they should have jumped up to 4096 bytes. Now the computer industry is making the change only because they are forced to change.
Instead of actually making the change real, they introduced a fake change of the type that produces more problems than it solves plus a lot of the problems are not obvious, which lets the manufacturers tell you lies when they sell the hardware.
When an Advanced Format disk wants to write a block of data, the disk has to read a 4096 byte sector, replace the 512 byte block, then write the 4096 byte sector back to the disk. The write times are pathetic.
The large RAM cache on the disk can reduce the impact of the read before write for sequential writes. The cache can combine with the read ahead feature of some disk controllers and have the next 4096 sector in the cache before it is needed. Lower cost disks have less cache and less intelligence in the controller.
The HD204UI has 32 MegaBytes of cache and should perform at a reasonable speed if that is all you use. My email exceeds 32 MB. The index files for my email exceed 32 MB. Just browsing my email will exceed the cache. When I am browsing email and the email software is updating the mailboxes with new mail, the data will continually flush out of the cache and force the time consuming read before write.
Command queuing can also reduce the impact of the read before write for sequential writes. In command queuing, a series of writes are sent in one request and the disk can schedule the commands for optimum performance. If there are several writes to one 4096 byte sector, the disk reads the sector once then applies all the changes then writes the sector once. This works for the Ext4 file system where data is written in 8192 byte blocks. A write of one 8192 byte block will produce a command to write 16 sectors of 512 bytes. The disk will read 2 or 3 sectors of 4096 bytes, apply the changes, then write 2 or 3 sectors of 4096 bytes.
Some operating systems can align disk partitions to the 4096 byte real sectors. The previous example of writing an 8 Kilobyte Ext4 block will always align to 2 sectors of 4096 bytes and never be scattered over 3 sectors. This alignment lets the disk controller skip the read before write.
Looking at the Samsung HD204UI write performance on a chart, I noticed the first writes were very slow then there was a buildup toward maximum speed then a period at nearly maximum. The pattern fits what you would expect if the first writes had to drag the 4096 byte sectors into the disk RAM cache. The last part of the chart shows what you expect to see when the writes are mostly to sectors already in the cache. The result is worse when you access more data than can fit in the cache.
Linux reports the decimal capacity is 2,000,398,934,016 bytes. Linux reports the disk as 1.8 TB total capacity and 1.7 TB available capacity with 93.3 GB already used by the Ext4 file system. When I switch to NTFS under Windows, Windows says the capacity is 2,000,396,288,000. NTFS uses 128,622,592 bytes (121 MB) and leaves 2,000,267,665,408 bytes (1.81 TB) free.
You might use the disk in a backup server because low cost space is more important that speed. You could use the disk for a print server because printers are far slower than the disk. The disk is fast enough to deliver music and video because of the fast read speed. The disk might not serve full high density video while also recording a couple of digital channels as full high density video. The reads will swamp the cache. The writes will have to read fresh data all the time. The disk would only work in this situation if you can align the file system blocks to the disk sectors.
The serial number starts with S2H7JD2B4 and may indicate a variation within the model. The disk firmware is 1AQ1 and it should not make any difference for this testing because the disk is well developed and stable. Some early Advanced format disks had horrible performance variations because of under developed firmware. I have not found the problem with the Samsung disks I have used.
Would I use this disk for a NAS (Network Attached Storage) device? No. I would select the 1.5 GB or 1 GB Samsung 7200 RPM disks to avoid the randomly low disk write speeds. Gigabit networks are more than twice as fast as USB 2 and faster than the minimum write speed revealed for the Samsung HD204UI.
Samsung say the disk makes noise at 2.5 Bel when idle and 2.8 Bel when active. What they do not mention is the annoying clicky sound some disks have when the heads change direction. This disk, most Samsung disks, and most Seagate disks do not click. Hitachi and Western Digital disks produce the most distracting and annoying clicks.
The disk uses 5 volts and 12 volts, placing the disk outside the range of USB 3 power. 2.5" disks can run on USB 3 power and could be a better choice for USB 3.
You need 2.0 Amps to spin up the disk. This could be significant power usage if you use the disk in a device that frequently drops to very low power mode. You need 6.3 Watts to read and write, a typical amount for 2 TB low power disks. You burn 5.1 Watts when the disk is idle and still spinning, a large amount compared to some of the competitors. You need 1 Watts when on standby power with the disk spun down.
While standby mode saves power, there is the power draw required for spinup and the limitation of a disk wearing out after 50000 spinups. If your disk spun down and up once per hour, your disk would last 5.7 years. If your disk spun down and up once per minute, your disk would last only 37 days. There is also the 13 second wait each time the disk spins up.
Western Digital and Seagate offer more expensive disks with a lower idle power. You might never save enough power to pay the extra cost. Your computer, or NAS, might burn far more electricity. You can save more power by upgrading your CPU or power supply to the latest power efficient models.
Samsung say the disk starts up in 13 seconds and this is similar to most of the disks with moderate start up power requirements. They claim an average seek time of 8.9 milliseconds and an average latency of 5.52 ms. The benchmarks show an average access time of 15.5 ms, a little bit slower than what you get by adding the latency to the seek time. Samsung claim a maximum media transfer rate of 250 MegaBytes per second and that did not occur in the benchmarks. The claimed 300 MB/s Interface Transfer Rate is the rate from the cache through the SATA II cable and is usually a little bit less because of the SATA overheads. 295 MB/S is closer to the practical limit for everything other than very long sequential reads.
I used a standard benchmark conducted on a raw disk with no file system to slow down access. The computer hardware is more than twice as fast as SATA II. The results are as follows.
|Measurement||SATA II||USB 3||USB 2|
|Average access time||15.5 milliseconds||15.8 milliseconds||16.3 milliseconds|
|Maximum read rate||149.8 MB/s||132.0 MB/s||33.5 MB/s|
|Minimum read rate||69.3 MB/s||65.9 MB/s||27.4 MB/s|
|Average read rate||117.7 MB/s||111.4 MB/s||31.9 MB/s|
|Maximum write rate||99.7 MB/s||95.4 MB/s||33.8 MB/s|
|Minimum write rate||26.3 MB/s||35.3 MB/s||26.1 MB/s|
|Average write rate||53.7 MB/s||89.4 MB/s||32.7 MB/s|
You can see a huge speed jump from USB 2 to USB 3. USB 3 is the minimum standard for external disks located near a computer and replaces eSATA because the cable can be longer. USB 3 also replaces eSATA for some disks, but not this HD204UI disk, because there is enough power supplied through the USB 3 cable to run some disks without a separate power supply. The smaller disks, 2.5" and 1.8", are designed for USB 3.
There is very little difference between USB 3 and SATA II. The minimum write rate is an anomaly caused by the disk using the unadvanced
Advanced Format system where 512 byte sectors are stored in 4096 byte sectors. When the disk wants to write a block of data, the disk has to read a 4096 byte sector, replace the 512 byte block, then write the 4096 byte sector back to the disk. This makes no difference on USB 2 because the disk is still as fast as USB 2. For USB 3 and SATA II, the write times are pathetic.
The large RAM cache on the disk can reduce the impact of the read before write for sequential writes. Command queuing can also reduce the impact. In command queuing, a series of writes are sent in one request and the disk can schedule the commands for optimum performance. If there are several writes to on 4096 byte sector, the disk reads the sector once then applies all the changes then writes the sector once. The cache can combine with the read ahead feature of some disk controllers and have the next 4096 sector in the cache before it is needed. Lower cost disks have less cache and less intelligence in the controller.
Looking at the Samsung HD204UI writes on a chart, I noticed the first writes where very slow then there was a build up toward maximum speed then a period at nearly maximum. The pattern fits what you would expect if the first writes had to drag the 4096 byte sectors into the disk RAM cache. The last part of the chart shows what you expect to see when the writes are mostly to sectors already in the cache.
As an experiment to test benchmarks, I ran the USB 3 test again, twice, in a similar setup. The access time is the same because the hardware is the same. The read times are similar because the hardware is the same. The write times are all over the place because the hardware changes. This is the effect of the Advanced Format where the hardware does not match what the operating system expects. The minimum and maximum write speeds depend entirely on how lucky you are with the fake sector size matching the underlying real sector size.
|Measurement||USB 3||USB 3||USB 3|
|Average access time||15.8 milliseconds||15.8 milliseconds||15.8 milliseconds|
|Maximum read rate||132.0 MB/s||149.1 MB/s||148.1 MB/s|
|Minimum read rate||65.9 MB/s||69.4 MB/s||69.3 MB/s|
|Average read rate||111.4 MB/s||111.8 MB/s||111.4 MB/s|
|Maximum write rate||95.4 MB/s||119.8 MB/s||122.2 MB/s|
|Minimum write rate||35.3 MB/s||16.3 MB/s||108.1 MB/s|
|Average write rate||89.4 MB/s||32.3 MB/s||114.9 MB/s|
If you want consistent write performance, forget the advanced format used for 2 TB disks. Buy a 1.5 TB or 1 TB disk and make sure it is not Advanced Format. Some brands are so dishonest that there is no warning in the product description about the fake format.
The disk temperature is sitting at a comfortable 22°C or 72°F after a long hard benchmark. This disk is really low power compared to other brands of similar capacity and speed disks. The low power translates to a disk running 10°C cooler than a full power 7200 RPM disk.
USB enclosure benchmark
This is the same standard benchmark conducted on a raw disk with no file system to slow down access. File opens can be slow in some operating systems and are often worse across USB 2. There is no file open during this test and no distortion caused by the operating system. The operating system happens to be the latest Linux. Windows 7 and XP will give the same result for USB 2. Vista is not predictable and I would not use Vista for any benchmarks. Apple Mac OSX is FreeBSD with a pretty face. I know from talking with FreeBSD and NetBSD developers ten years ago that their operating systems are not going to slow down something that is already as slow as USB 2. Compared to SATA II, USB 3 shows a slight drop in speed because of the extra communication overheads in USB.
The computer hardware is many times faster than USB 2. This test rarely uses any part of the hardware, other than the USB wiring, for more than ten percent of the time. The software recording the resource usage uses more resources than USB and the combined result is still under ten percent.
The USB enclosure is the Astone ISO gear 481U3 external enclosure for 3.5" disk with a USB 3.0 connection.
I am impressed by the low noise. The low power, based on rated specifications, is the best at this capacity and price. I am happy to use this disk for backups via long USB 2 cables running into the next room and through the network. I would not use this disk or any other Advanced Format disk where I need good write speed.