AIDA64 Random Access Time
Version and / or Patch Used: 1.60
Developer Homepage: http://www.aida64.com
Product Homepage: http://www.aida64.com
AIDA64 offers several different benchmarks for testing and optimizing your system or network. The Random Access test is one of very few if not only that will measure hard drives random access times in hundredths of milliseconds as oppose to tens of milliseconds.
Drives with only one or two tests displayed in the write test mean that they have failed the test and their Maximum and possibly their Average Scores were very high after the cache fills. This usually happens only with controllers manufactured by Jmicron.
As we mentioned in the overview article, Samsung moved to 128Gb die, at least that's what we think since no one is confirming all of the details in the move to 19nm. The latency is higher on the EVO than it was on the 840. That's a good indication of 128Gb die and the same thing we're seeing on the IMFT 20nm flash.
Samsung had a write latency issue on the 830 and other 840 products in the 128GB capacity sizes. The 840 EVO 250GB has just two packages and we believe each package has two die inside. The interleaving should be the same as the base 840 drive with 128GB of 2Xnm flash. We're seeing about the same write latency when writing across the span of the drive.
This is the plot with RAPID Mode. Again, you can see the different stages of caching. When we write to the entire span of the drive, the cache doesn't have a chance to flush the data to TLC, because there isn't an idle state between writes. If we were able to write for 30 seconds, then idle for 30 seconds and then continue on like a normal consumer application, the results would look different.
Samsung doubled the amount of low power DRAM in the 840 EVO over the 840. That DRAM caches the page table data. That's why the write latency is actually lower than the 840 at this stage of the test. The 840 write latency numbers published in the chart are from a special test that was performed after the drive sat idle for a period of time. When you let the 840 idle, it finished the garbage collection and the latency returns to what we would call a normal state.