Real-World Software Performance Testing
We didn’t have time to run both BX200 drives with Momentum Cache in the PCMark 8 Advanced Storage Tests and notebook battery life test.
The BX200s had a hard time making it though the heavy portion of the test (they hung around the 50 MB/s mark). Momentum Cache helped, but it isn't able to spin a different performance story under taxing workloads. Really, the drive wasn't designed for heavy workloads.
Unfortunately, it doesn't fare any better under moderate use either. Though Momentum Cache augments performance here as well, it only allows the 960GB BX200 to compete with OCZ's 480GB Trion 100. And it's not like the Trion 100 is a measuring stick for great SSDs. The Samsung 850 EVO holds that title in the low-cost space, and it's nearly three times faster in part of the test.
Total Access Time
The access time tests really tell the story. Even with Momentum Cache enabled, the BX200 lags behind the other drives. Without the DRAM cache feature, the latency is painful.
This set of charts is something we've worked on over the last year, but haven't displayed yet. They show the latency buckets in an 80% read mixed workload with 4KB data at a queue depth of one. We want to see more red (0 to 50us) and black (50 to 100uS) area. Blue (100 to 200uS) and beyond is bad.
The chart illustrates why Samsung's 850 EVO is such a good product and why the 480GB BX200 is its polar opposite.
Notebook Battery Life
We measure SSD power consumption differently than other sites. Instead of looking at four-corner power modes, we use BAPCo MobileMark 2012.5 with real-world software and a notebook. Outside of when it's actually being utilized, most of an SSD's power consumption is related to garbage collection and other background flash management activities.
Controllers take a couple of approaches to these tasks. Some start the clean-up process right after new data goes to the flash. Others wait until a defined amount of cleaning is needed, then wipe the cells at once. This consumes more power than heavy reads and nearly as much power as a 4KB random write test. Because of when it all happens, though, most reviews miss it. By recording across several hours using powerful software, we're able to capture a drive's behavior more accurately.
Unfortunately, we had very little time between receiving our BX200 sample and Crucial's embargo, so we could only test one capacity point. We chose the 480GB model, since it's compared to other 512GB-class SSDs. It fell towards the bottom of the drives tested today, though the result is still good on our Lenovo notebook. Just be aware that Adata's SP550 with the same controller and SK Hynix TLCmemory performed better.
For those who have been out of the loop or haven't bough their first SSD yet just avoid SSDs that use TLC unless it is 3D (stacked) and just pay a little more for MLC.
It's not a duplicate. The latency distribution is just the same for both. I'm working on fine tuning those charts a bit.
Also, you will notice that the bars on the side are different sizes even though it's still 0%. Data has fallen into those buckets but not more than 1%.