Continuous activity test
We recently began testing SSDs to find out how well they perform when they are in continuous use for 15 minutes. We did this with the 250 GB and 256 GB models of the 840 and 840 Pro.
For an SSD to be able to write, the data cells first have to be emptied. This can only happen per block, and a block consists of 128 pages and is therefore 512 kB.
SSDs therefore have to employ some smart tricks. When a number of pages of data have to be removed, the rest of the data first has to be copied to another block, after which the entire block can be emptied. In practice that means that SSD controllers save up writes as much as possible. When the SSD is idle, the controller activates garbage collection, which removes the data from the chips and combines the left-over data as much as possible in complete blocks. This ensures that as many blocks as possible are made available for write actions.
However, if the SSD is continuously in use, the garbage collector doesn't get the chance to do this. Eventually this will cause all blocks to be occupied, which will force the SSD to employ garbage collection between commands. This has a negative effect on performance.
To illustrate this impact on performance, we performed a 15-minute test on several SSDs with the IOMeter benchmark. The settings were as follows: 32 simultaneous instructions (queue-depth 32), with 50% reading and 50% writing. The instructions consists of 25% 4 kB, 25% 64 kB, 25% 128 kB and 25% 1 MB. IOmeter uses a 160 GB test file that takes up about 60 percent of the SSD's capacity. (160 GB for 256 GB SSDs, 320 GB for 512 GB SSDs).
The Crucial M500 480GB SSD keeps it up for about 11 minutes, after which a slowdown occurs to about a third of the normal performance. We have not yet tested many 512GB SSDs in this manner.