Solid-State Drives

In an ideal world, every PC would use a solid-state disk (SSD) to store data, rather than a hulking big mechanical hard disk. Using either DRAM or flash memory, SSDs are not only physically smaller than hard drives, but they're also silent, much more reliable (due to the lack of moving parts), more power-efficient, produce less waste heat and are faster too. However, until recently, SSDs have been so expensive that they've only been adopted by the defence and finance industries, where high performance and robustness are far more important than a low price.

However, as with everything in the industrial era, the economies of scale are bringing down the cost of SSDs to more consumer-friendly levels. This isn't to say that any of the SSDs in this Labs test are cheap, however - using a simplistic price per gigabyte index, they're still far more expensive than hard drives. However, SSDs offer a number of advantages over their mechanical competitors.

In conjunction with the tumbling cost of SSDs, there has also been an increase in the number and type of SSDs available; they vary depending on the type of memory they use, their form factor and interface. For example, some SSDs use volatile DRAM, which is generally faster than flash memory but can only retain data while it's being fed power. Flash memory, on the other hand, is non-volatile, which means that data is retained even when the device is switched off.

SSDs are manufactured in a wide range of different form factors, from 2.5in drives designed to replace laptop hard drives, to 3.5in drives for desktop PCs and servers. What also sets the SSDs in this article apart from the humble USB memory key are their interfaces. While most use common-as-muck S-ATA or EIDE interfaces, others employ the seriously heavyweight Fibre Channel interface, which is more commonly found in massive data centres than in desktop PCs.

Despite these differences in design, all of the SSDs in this Labs test have one thing in common - they want to replace the mechanical hard drive in your PC. One thing's for sure; if you want a silent PC, more robust storage for your laptop, or if you're just desperate to eek out more performance from your PC, then buying an SSD is the answer.

While an SSD (solid-state drive) works in a completely different way to that of an HDD (hard disk drive) it performs exactly the same function, which is to store data. For this reason, we began testing each SSD by measuring how fast data can be read from and written to the device. First of all, we used Simpli Software HD Tach 3 RW (www.simplisoftware.com), which runs a series of benchmarks to record sustained throughput when reading and writing, plus average seek times. These tests exposed one of the great advantages of SSDs. Hard disk drives perform worse when data is read/written to the inner zones of its platters compared with outer zones, since the outer sections (which have longer tracks) contain more sectors per track. However, the read/write performance of an SSD is identical, regardless of whether it's accessing data from its first or last gigabyte. These tests also revealed the Achilles heel of flash memory, which is that it's much slower at writing data than reading. This means that, while a flash memory-based SSD may be faster than a hard disk at read-intensive tasks, it will be much slower with write-intensive applications.


However, an SSD's performance in real-world applications is affected by other factors too, such as the memory buffer size, not to mention the algorithms used to calculate which data to store in the buffer and which to discard. For this reason, we also tested the SSDs in several real-world scenarios based on the kinds of tasks that you're likely to perform with your PC.


Our first real-world test is the Paint Shop Pro image editing test from our Media Benchmarks 2006. This test opens a large number of high-resolution photos, edits them and then saves the altered files back to the drive. This means that the read and write performance of an SSD can have a significant effect on the test. In addition, we also ran the multitasking test from our Media Benchmarks 2006. This runs the same Paint Shop Pro test, while also creating and writing a large RAR file archive in the background, which further strains the SSDs.


Our two other real-world tests time how long the test system takes to load games. The first test times how long it takes to load the Fort level of Far Cry, while the second test times how long it takes to load the Hunt for the Wounded Bear scenario from Silent Hunter 4 (reviewed on p92). At first glance, loading games seems to be the kind of task that would benefit from a drive with a fast sustained read throughput, since games tend to store much of their content in large archive files. However, in reality, most games grab smaller files from within these archives, which means that access time is just as important. This helps to explain why SSDs are so fast at loading games compared with hard disk drives, which can't start a read or write operation until the read/write head has physically moved to the correct part of the platter. In an SSD, the access and seek times are almost instantaneous, since all the memory controller has to do is address the memory.


In addition, we also ran a number of tests using Iometer, which is an enterprise-level, open-source storage device benchmark utility. Unlike HD Tach, Iometer can be scripted to run multiple scenarios, allowing us to simulate several different usage models. For starters, we ran a sequential 64KB block-size write and read test, followed by a random 64KB block-size read and write test. This provides a good indication of the maximum transfer rate and IO/sec (input and output operations per second) of each drive. We than ran a further four tests with a 4KB block size on the high-end SSDs. These smaller block-size tests offer a good approximation of how well each drive will perform when running a massively parallel database, which is the task for which high-end SSDs are optimised.


Our current storage device test rig consists of a 2.13GHz Core 2 Quad Q6400 CPU installed in an Asus P5B-E Plus motherboard, with 2GB of Kingston PC2-6400 RAM. All the tests were carried out using Windows XP SP2. This is typical of the kind of performance rig to which most of us aspire, as the motherboard is based on the excellent Intel P965 chipset with Intel's ICH8R Southbridge. For the Fibre Channel devices, we used a superb QLogic QLE2462 4x PCI-E HBA controller.


To make it easier to put the performance of SSDs into perspective, we also added the benchmark results for several high-performance hard disks to our graphs. These disks comprise the 2.5in Seagate Momentus 7200.1, which is a direct competitor to the 2.5in flash drives, and the 3.5in Seagate Cheetah 15K.4 and Western Digital Raptor X to provide a comparison with high-performance SSDs. We defragged the hard disks between each test, thereby allowing them to perform at their very best, but this wasn't necessary with the SSDs.


The overall Speed score is based on a weighted combination of all the different real-world and synthetic performance tests. Unlike hard disks, we've included a separate Features score for each SSD, as the different memory types, interfaces, form factors, warranties and error correction technologies are all important considerations. The Value score is then calculated by dividing the Speed and Features score and capacity by its price. The Overall score is the sum of these scores.


Custom PC would like to thank QLogic and Kingston for the loan of some of the test equipment used to conduct this Labs test.