Angelbird SSD wrk 256GB Solid State Drive Review

Test Procedure

Test system:

• Supermicro C7Z87-OCE
• Intel Xeon E3-1230Lv3
• Corsair Vengeance 8GB 1866MHz
• Corsair H100i
• BeQuiet Dark Power Pro 850W
• Kingston HyperX 240GB SSD

We would like to thank Be Quiet, Supermicro, Intel, Corsair, Kingston and Lian Li for supplying us with our test system components.

Many different software applications are also used to gain the broadest spectrum of results, which allows for the fairest testing possible.

Software used:

• Anvil’s Storage Utilities V1
• AS SSD
• ATTO
• CrystalDiskMark Tech Preview

In a bid to make our testing as thorough and as accurate as possible, we are always looking for ways in which to tweak and tune our testing methodology and to give a clearer picture at how a drive performs throughout its life, I have formulated a new way of testing that gives eight times the amount of data to analyse within the review. Whilst this means that testing each drive will take considerably longer to test – nearly 36 hours in fact, the picture that we are able to build up of the overall performance of each drive is far clearer, providing a more accurate analysis at how a drive performs under different conditions.

The basis around the new method of testing is to do with how the performance of the drive improves or degrades as the amount of free space is reduced. To simulate this I will first of all test the drive when it is perfectly clean and fresh out of the box. Following on from this I will test the drive at key points as the volume is filled up with data, namely 25%, 50% and 75% full. As the type of data that is on the drive is directly relational to how the drive performs, I will be using a sample set of data that includes a mixture of small files such as text documents through medium-sized files such as photos and large video files. Intel’s NASPT suite that I use for NAS testing generates a 15GB block of data and as this contains a wide variety of file types and sizes, it is perfect for this application. To fill up the drive I will copy this block of data in to the root folder and copy it to fill the drive as required.

On top of this I will also take wear on the NAND in to consideration and this process obviously takes a long time to perform – so long under real world conditions that it is simply not practical. In order to accelerate this conditioning process I will use the SSD endurance test that lies within Anvils Storage Utilities and read / write 5TB of data to the drive through a process of filling the drive with thousands of small files of various sizes and compressibility (with randomised duration between each of the writes), then reading each of them back with a randomised time between each file. After this has completed, the data is then erased and the process starts again. This process of conditioning the drive can take anywhere upwards of 12-15 hours on a typical 256GB SATA III SSD, however the performance and capacity of the drive will influence the time it takes to condition the drive.

Following the conditioning process, the benchmarking process as described above is repeated again with the volume filled with sample data to each percentage between test runs. Furthermore between each benchmark test, the drive will be left alone to allow the TRIM command to take place if the drive feels it is necessary. From previous experience, TRIM can heavily impact the drives performance as it takes place, so allowing a period of time between tests will eliminate this factor and any subsequent false performance figures.

On each page that follows with the benchmark results, I have inserted the screenshots from the benchmark results when the drive is 75% full. In the drive comparison charts, the performance at 0% fill is used as these are the performance figures that are recorded from drives in previous SSD reviews.

Sadly, accidents can happen and it looks like our test bench got a little knock on the transport to me. This has the little side effect that I couldn’t run some of the usual tests. I will however make it a priority to return to those tests as soon as the test bench is back to fully working state. There might also be slight variations to the old results Chris made, as I might run the test slightly different. I’ve tried to adjust it as close to the previous as possible.