Power requirements and efficiency
Storage device manufacturers by law must
provide power consumption specifications with their storage device products. Quite often these specifications are quite vague, and rarely, if ever, publish
the power efficiency of their storage devices with regard to how much work a storage device can do for a given amount of energy consumed. In this article we
will disclose with unprecedented precision, the energy efficiency of some popular storage devices.
Myce has now secured a piece of ‘state of the art’ test equipment, which takes a sample every four micro-seconds, that I will be using to measure the power consumption of consumer grade SSDs and HDDs. I’m so very proud to be able to announce that Myce.wiki, in partnership with Quarch Technology, now aims to bring our readers the most comprehensive, and accurate, power consumption tests ever carried out on consumer grade storage devices, to be found anywhere on the Internet.
Myce’s Power Testing will be carried out using Quarch Technology products. More specifically we are privileged that Quarch has allowed us to use their latest Programmable Power Module (‘PPM’) and we would also like to take this opportunity to give a huge ‘thank you’ to Quarch for providing this equipment. The PPM is specifically designed for testing low power sleep states on modern SSDs and as such has a remarkably accurate low level current measurement, down to 100mA (micro amps, or millionths of an amp). Please click here for details.
Quarch Technology is a world leader in the supply of testing solutions for the data storage industry and if you would like any further information please visit their website by clicking here.
Let’s take a closer look at the Quarch PPM box in a bit more detail.
Quarch Technology PPM
The Quarch Technology PPM is able to provide two power supply rails to the target SSD. A 12V (volt) rail is required for PCIe based SSDs, and also for SATA HDDs, SATA HDDs also require the 5V rail to function. All the power requirements of a SATA SSD are handled by the 5V rail. The Quarch Technology PPM can switch between 5V and 3.3V on the secondary power output channel as required. So for SATA based SSDs it is set to 5V, and for PCIe based SSDs, it is set to 3.3V.
On the right of the Quarch PPM, you can see the socket where the main power injection lead connects.
On the rear of the box (not shown) you will find a USB 2 socket, a power socket (to supply power to the unit) and a Torridon connection interface, for connecting to external equipment.
Although the Quarch Technology PPM can be used on a single PC, which can act both as host and measurement system, I will be using two PCs to run the tests. One PC will handle the measurements, and the second PC will act both as a host for the target SSD, and will also be used to load the target SSD with data. This will allow me to do some pretty fancy power consumption tests.
I will first show the type of workload being used to load the SSD during the power consumption test. I will then present the power consumption graph, and power consumption statistics of the SSD.
I will display the results in the form of bar graphs, at the end of each test carried out in this article, so one can compare the results obtained on all the SSDs featured in this article.
I will use the following IOMeter test patterns to load the SSD or HDD.
- 4K random read and write at a queue depth of 1 (to emulate a lightweight consumer workload).
- 4K random read and write at a queue depth of 4 (to emulate a medium workload).
- 4K random read and write at a queue depth of 32 (to emulate a heavy workload).
- 512K sequential read (to emulate reading a sequential file from the storage device).
- 512K sequential write (to emulate writing a sequential file to the storage device).
Power requirements for a lightweight consumer workload – 4K random read and write QD1
A typical lightweight consumer workload will generally be at very low queue depths. Typically at a queue depth of one or less. I’m testing random data at a block size of 4 Kilobytes, as this block size of small random files is generally accepted as the most frequently occurring in the consumer environment.
I will show the chart generated by the Quarch PPM for the drive that I have tested. I will then show the results in the form of bar graphs, so one can easily compare with other recently tested SSDs.
There will actually be two bar graphs for each test. The first graph will show the average power consumption during the test run. The second graph, which is much more important, will indicate the power efficiency of the storage device, showing how much work the storage device can do for each Watt of energy it consumes.
4K Random Read – queue depth 1
Crucial BX200 960GB – 4K random read QD1
It looks like the BX200 requires more power than the drive that it replaces.
The read results when it comes IOPS per Watt are acceptable, especially for a TLC SSD.
4K Random Write – queue depth 1
Crucial BX200 960GB – 4K random write QD1
The same thing happens when the drive is writing data, the BX200 is more power hungry.
Well, the performance of the Crucial BX200 is not anywhere near to what I would have liked to have seen from the BX200 SSD.
Power requirements for a medium weight consumer workload – 4K random read and write QD4
A typical medium weight consumer workload will generally be at a queue depth of four or lower. This workload would typically involve some multitasking, with perhaps two or three applications running, and processing data simultaneously. I’m testing random data at a block size of 4 Kilobytes, as this block size of small random files is generally accepted as the most frequently occurring in the consumer environment.
I will show the charts generated by the Quarch PPM, for the drive that I have tested. I will then show the results in the form of bar graphs, so one can easily compare with other recently tested SSDs.
4K Random Read – queue depth 4
Crucial BX200 960GB – 4K random read QD4
Here again we can see that the BX200 consumes more power that the BX100.
Not much to say, as the graph says it all. The BX200 is still behind the BX100.
4K Random Write – queue depth 4
Crucial BX200 960GB – 4K random write QD4
Continuing the write tests and again the Crucial BX200 shows that it needs more power compared to the BX100.
Again the Crucial BX200 is behind the drive that it replaces, It requires more power and delivers less IOPS.
Power requirements for a heavyweight consumer workload – 4K random read and write QD32
Whilst this workload is unlikely arise for the casual consumer PC user, it could well appear in a semi-professional consumer environment, such as in a graphics workstation. This workload would usually involve heavy multitasking, and having several processes running concurrently that require constant access to small files located on the storage device for input or output.
I’m testing random data at a block size of 4 Kilobytes, as this block size of small random files is generally accepted as the most frequently occurring in the consumer environment.
I will show the chart generated by the Quarch PPM, for the drive that I have tested. I will then show the results in the form of bar graphs, so one can easily compare with other recently tested SSDs.
4K Random Read – queue depth 32
Crucial BX200 960GB – 4K random read QD32
In this test we can clearly see a big difference in the power requirements for the two tested drives. As we can see the BX200 wants 1038mW more.
No surprises here, again the new TLC based SSD is behind the older MLC SSD.
4K Random Write – queue depth 32
Crucial BX200 960GB – 4K random write QD32
And again the Crucial BX200 needs 1155 mW more than the drive that it replaces.
Here we can see that the difference between the two drives is huge, and in favour of the BX100 SSD.
Power requirements of a storage device when reading and writing sequential data
Not all of a consumer workload is based around the reading and writing of small random files. Many files are sequential in nature, and can vary in size from a few Kilobytes to several Gigabytes, so your storage device will spend a lot of time reading and writing sequential data.
I’m testing sequential data at a block size of 512 Kilobytes.
512KB Sequential read
Crucial BX200 960GB – Sequential read
No surprises here, the newer drive continues to require more power to complete the same task.
Again the Crucial BX200 SSD is behind the drive that it replaces.
512KB Sequential write
Crucial BX200 960GB – Sequential write
This is the first time that the Crucial BX200 is able to stay ahead of the BX100, a nice result.
When it comes to IOPS per Watt, the Crucial BX200 is faster than the drive that it replaces.
Power requirements of storage devices when they are idle and doing no work at all
The practical reality relating to power consumption is that it can be quite erratic and sometimes unpredictable. Some of us will invest in the most powerful PC we can afford, only to find that the PC can spend quite a lot of time running and doing absolutely nothing. Storage devices are no different.
Often we can be sitting idly pondering what to do next, or perhaps browsing the Internet. When we arrive at a page that interests us, we will read it, and that can take a fair amount of time to complete. During this period the storage device will most likely be idle, but still consuming energy.
In this test, I’m measuring how much energy the storage device consumes when doing no work at all.
Crucial BX200 960GB – Drive idle
As we can see the BX200 SSD is a little more power hungry even when it’s idle.
I will now show one more test, and this should be regarded as for information purposes only.
Power requirement trace of an SSD booting Windows 8.1, in real time.
This test is for interest only, and shows the power requirements of the review SSD booting Windows 8.1 to the desktop.
Crucial BX200 960GB – Real time trace of the drive booting Windows 8.1 to the desktop.
As we can see the crucial BX200 SSD requires more power, to boot into windows.
We can clearly see that the Crucial BX200 SSD requires more power than the drive that it replaces, and it also falls behind on the IOPS per Watt graph. I admit that I would have liked to have seen a smaller gap between the two drives, but unfortunately it didn’t happen. We also have to remember that the BX200 is the first drive that Crucial has with their new TLC NAND, and TLC NAND also requires more accurate error correction so this has an effect on the overall power consumption.
This doesn’t change the fact that the drive isn’t as power efficient as the BX100, although let’s hope that there is room to tweak the firmware of the BX200 to improve some aspects of the drive.
Now let’s head to the next page where we will look at how the Crucial BX200 SSD performs using Anvil’s Storage utilities….