I bought the SU650 version a while back and was just now getting ready to clone a system drive to it. If this version has the same potential errors, then it stays on the shelf. Any idea if this 650 version is a candidate for failure?

@calmist

They have lot of different configs under same name.

I suggest to never use such SSDs for system drives - get some Samsung SSD ala https://www.joybuy.com/600276700.html

480Gb with bad Maxio controller

Focusing on the off-cache behavior is a misleadingly bad test method with modern SSDs, especially on entry consumer lines as they will always be using high density flash with slow native write speeds. When your HDT results above show caches of 60-100GB, a user is unlikely to ever see that type of severely low transfer rate in normal usage patterns. These caches are likely dynamic and will perform worse as the drive fills, but nobody buying a 120GB SSD these days should expect miracles. Sadly, some manufacturers like Kingston cut worse flash density corners on 480GB vs 240GB drive models of a line such as the A400--which could be considered either unacceptable for the 480GB drive or an unexpected surprise for the 240GB model.

A far more useful testing method is to use AS-SSD or Anvil and perform the tests in 10-12GB data file sizes (versus their 1GB defaults), as it will push the caching enough to show its latency in heavy / mixed use. Anything showing an inflated latency score on AS-SSD writes in 10GB test mode will likely have noticeable lag or stuttering in "heavy" average user operations (this is badly controlled on Phison-based drives, and some Marvell-based drives like the Sandisk X600). The random write access time / latency score varies widely during those tests, and using the copy mode will also show major differences between drives (but does not measure latency, it just reflects transfer rates during heavy loads).

Anvil is useful for running in 1GB and 12GB test modes, and noting how much of a drop is seen at 12GB. Drives designed or optimized for sustained writes (eg: most enterprise drives) will often increase slightly at 12GB, whereas drives for consumer use with heavy cache reliance will see a significant drop. This is an effective reflection of the drive's mixed performance, whereas HDT full drive sustained write tests are artificially pessimistic and ignore how effective an SLC cache is in average use patterns.