For our last benchmark, we have decided to use PCMark 8 Extended Storage Workload in order to determine steady state throughput of the SSD. This software is the longest in our battery of tests and takes just under 18 hours per SSD. As this is a specialized component of PCMark 8 Professional, its final result is void of any colorful graphs or charts typical of the normal online results and deciphering the resulting excel file into an easily understood result takes several more hours.
There are 18 phases of testing throughout the entire run, 8 runs of the Degradation Phase, 5 runs of the Steady State Phase and 5 runs of the Recovery Phase. In each phase, several performance tests are run of 10 different software programs; Adobe After Effects, Illustrator, InDesign, Photoshop Heavy and Photoshop Light, Microsoft Excel, PowerPoint and Word, as well as Battlefield 3 and World of Warcraft to cover the gaming element.
- PRECONDITIONING -The entire SSD is filled twice sequentially with random data of a 128KB file size. The second run accounts for overprovisioning that would have escaped the first;
- DEGRADATION PHASE – The SSD is hit with random writes of between 4KB and 1MB for 10 minutes and then a single pass performance test is done of each application. The cycle is repeated 8 times, and with each time, the duration of random writes increases by 5 minutes;
- STEADY STATE PHASE – The drive is hit with random writes of between 4KB and 1MB for 45 minutes before each application is put through a performance test. This process is repeated 5 times;
- RECOVERY PHASE – The SSD is allowed to idle for 5 minutes before and between performance tests of all applications. This is repeated 5 times which accounts for garbage collection; and
- CLEANUP – The entire SSD is written with zero data at a write size of 128KB
In reading the results, the Degrade and Steady State phases represent heavy workload testing while the recovery phase represents typical consumer light workload testing.
As you can see, performance is recorded in terms of Bandwidth and Latency. Bandwidth (or throughput) represents the total throughput the drive is able to sustain during the tests during each phase. Latency, at least for the purposes of PCMark 8, takes on a different outlook and for this, we will term it ‘Total Storage latency’. Typically, latency has been addressed as the time it takes for a command to be executed, or rather, the time from when the last command completed to the time that the next command started. This is shown below as ‘Average Latency’.
PCMark 8 provides a slightly different measurement, however, that we are terming as ‘Total Storage Latency’. This is represented as being the period from the time the last command was completed, until the time it took to complete the next task; the difference of course being that the execution of that task is included in ‘Total Storage Latency’. For both latency graphs, the same still exists where the lower the latency, the faster the responsiveness of the system will be. While both latency charts look very similar, the scale puts into perspective how just a few milliseconds can increase the length of time to complete multiple workloads.
For a more in-depth look into Latency, Bandwidth, and IOPS check out our primer article on them here.
AVERAGE BANDWIDTH (OR THROUGHPUT)
These results show the total average bandwidth across all tests in the 18 phases. In this graph the higher the result the better.
AVERAGE LATENCY (OR ACCESS TIME)
These results show the average access time during the workloads across all tests in the 18 phases. In this graph the lower the result the better.
TOTAL STORAGE LATENCY
These results show the total access time across all tests in the 18 phases. In this graph the lower the result the better.
PC Mark 8’s consistency test gives us some much better insight into the performance difference between the drives capacities and controllers. When it comes to heavy workload performance, the 1TB model blows the 120GB model out of the water. Average latency is almost three times lower and bandwidth is twice that of 120GB. And when it comes to light workloads, the performance gap becomes smaller, however, the 1TB capacity model still edges out all others.
What is interesting to note is the performance between the 1TB 850 Pro and the 1TB 850 EVO. They both features a 3-core MEX controller and 3D V-NAND, however, here we can start to compare the performance differences in 2-bit MLC vs 3-bit TLC. The inherent issue of increased latency and slower speeds is still present in the heavy workload results, however, we can see that when it comes to light workloads, the results are practically the same. Also, throwing in the Mushkin Reactor with 2D MLC, we can see that the 1TB EVO outperforms the value drive, which is still quite impressive considering the downfalls of TLC. The 3D manufacturing really does prove to make a difference.
For our power consumption testing, we have the drive connected to the system as a secondary drive. To record the wattage, we use an Amprobe AM-270 multimeter connected in line with the 5v power on our SATA power cable to the drive. The multimeter records the min/max amperage draw from the drive over our testing period.
We also record the drive’s sequential and random read and write power draw using Anvil Storage Utilities. We then take the values recorded and calculate the wattage of the drive. Some of the results may seem high compared to a standard notebook HDD because as these are peak values under load. When we see average power draw, SSDs are still more power efficient because they only hit max power for a short period of time.
With just one quick glance of the above graph we can see that the power consumption of the 1TB 850 EVO is much higher than that of the lower capacities. Startup hits 4.25W, at idle it consumes nearly twice the amount at 50mW. And for the rest of the results we can see it continues to consume more power as well.
REPORT ANALYSIS AND FINAL THOUGHTS
Throughout testing, the 1TB Samsung 850 EVO has proven to be quite the performer. Utilizing the same controller as the 850 Pro and 840 EVO, we were able to see great speeds in our synthetic tests. Reads and writes both pushing past 500MB/s. In our PCMark 8 testing, we were able to clearly see the performance difference between the two controllers in this product series. The 3-core MEX controller in the 1TB 850 EVO provides much lower latency over the MGX controller, however in our power testing, the 2-core MGX controller design definitely provides for greater power efficiency. Furthermore, in PCMark 8’s consistency test we were able to see just how much of a difference there is between TLC and MLC NAND. Overall, for light workloads there really isn’t much. It isn’t until the drive is put under heavy workloads that we can start to see a difference, but even then, Samsung’s 850 EVO still was able to outperform another value drive with planar MLC NAND.
3D V-NAND truly helps to alleviate a lot of the issues with TLC NAND flash. Samsung’s early move to both 2-bit and 3-bit 3D V-NAND should continue to give them a bit of an advantage down the road when more manufacturers start theirs, especially with pricing.
The increase from a 3-year warranty to a 5-year warranty is great and helps to improve this drive’s competitiveness in the market. Performance of the new Samsung 850 EVOs under light workloads is very impressive, however, the 1TB model seems to provide a clear advantage under heavy workloads. One thing we would have liked to see out of the new 850 EVO series would be some sort of extra power loss protection, as we have seen in other value oriented drives. Besides that, there isn’t much else. With the following Samsung has developed with their previous drives, we don’t see why this one won’t continue a similar trend in popularity.
If you are in the market for a new high capacity SSD, the 1TB Samsung 850 is a very competitive option to take a look at.