Four-Corner Performance Testing
We have all three Adata SP550 drives in today's chart. We're comparing them to a large collection of TLC- and low-cost MLC-based models.
Even the newest TLC-based solid-state drives deliver exceptional sequential read performance. What we don't know yet is how well LDPC will fix their errors long-term. The very small lithography processes are susceptible to bit-flip, voltage shift and noisy neighbor errors. As the flash wears, all of those issues should be corrected by low-density parity-check code at the expense of increased latency. In time, the drives will slow down. However, that should be years down the road.
Since we use light conditioning before running our test suite, many of the SSDs that rely on emulated SLC fall short in our write tests. The SLC buffer is too small for accurate performance measurements since they only facilitate a few moments (often less than two seconds) of accelerated write time.
Higher-capacity SSDs benefit from larger SLC buffers, so it's a little easier to benchmark 512GB+ drives. It doesn't come as a surprise that the 120GB SP550 appears at the bottom of our list. Really, our focus is on the 240GB SP550, since the chart highlights other 256GB-class products.
The four SM2256-controlled drives are all lumped together toward the bottom of the random read IOPS chart. Longsys' Foresee S500 240GB joins the Adata SP550s with the same controller. It appears that Toshiba's A19 TLC is a bit faster than SK Hynix's when reading random data. Thus far, the Toshiba flash has proved superior in almost every test.
At low queue depths, we clearly see two distinct groups in the random write tests with 4KB data. All three of the drives in the top tier cost about $25 more than Adata's SP550.
It's time to start worrying why the 480GB SP550 hasn't walked away from the lower-capacity models in these tests. The big SP550 has twice as much emulated SLC as most of the other drives, but isn't using it to outperform Samsung's 850 EVO.