How much faster is SSD
Speed: Faster?
Theoretical speed calculation:
Let’s take an example with:
- A hypothetical 7200 rpm hard drive with 150 Mb/sec of write sequential bitrates and 15 ms of access time
- An old SSD with only 50 MB/s write throughput and 0.1 ms of access time
- A recent SSD with a throughput of 500 MB/s and an access time of 0.1 Ms also.
Let’s start with big files (about 150 MB). What we call big files (> 100 MB) are typically movies:
- The hard drive puts 15 ms (0.015 s) to access the data and 1 s to write it (because read 150 MB with a throughput of 150 MB/s takes a second ^ ^). For our action, it will have put 1.010 s.
- The old SSD puts 0.1 ms (0.0001 s) to access the data and 3 s to write it (because 150 MB/50/s = 3 s), a total time of 3.0001 s
- The recent SSD puts 0.1 ms to access the file and 0.3 s to write it (because 150 MB/500/s = 0.3 s), or a total of 0.3001 s
- In this example, the hard drive is faster than the old SSD. In our action, it will have had a real throughput of 150/1, 015s = 147.78 mb/s. The old SSD when at it will have had a real throughput of 150/3, 0001 s = 49.998 MB/s. While the recent SSD had a throughput of 150/0, 3001s = 499.83 mb/s.
Of course, the newer SSD is doing better than the other two, but if we only take care of the old SSD and the hard drive, the latter is more efficient, thanks to its higher sequential throughput.
But let’s now look at a 1 MB file (this order of magnitude, between 1 and 100 MB, is often used for music files):
- The hard drive always puts 15 ms to access the data and 6.66 MS to write it. Altogether it will have put 21.66 Ms.
- The old SSD puts 0.1 ms to access the data and 20 ms to write it, a total time of 20.1 ms
- The recent SSD puts 0.1 ms to access the file and 2 ms to write it, a total of 2.1 ms
- If throughput is calculated, this is 1 MB/21, 66ms = 46.16 mb/s for hard disk, against 1 MB/20, 1 = 49.75 MB/s and 1 MB/2.1 ms = 476.19 MB/s.
We see here that the recent SSD always exceeds the others from afar. The old SSD is passed in front of the hard drive, despite its 3 times lower sequential flow. What for? Thanks to its access time, much lower. Indeed in the case of the hard drive, in this example it will have taken more time to reach the data than to interact with it, thus diminishing its actual flow.
Good and now let’s try with small files, 10 KB for example:
- The hard drive always puts 15 ms to access the data and 0.066 MS to write it. Altogether it will have put 15.066 Ms.
- The old SSD puts 0.1 ms to access the data and 0.2 ms to write it, a total time of 0.3 ms
- The recent SSD puts 0.1 ms to access the file and 0.02 s to write it, a total of 0.12 ms
- On the debit side, the hard disk collapses because 10 KB in 15.066 MS is 664 KB/s, compared to 33 MB/s for the old SSD vs. 83.33 MB/s for the recent SSD.
Concretely, it is seen here that no matter (or almost) the reading rate, it is almost instantaneous given the size of the file. On the other hand, the access time takes a huge proportion compared to the almost instantaneous reading of the file. Thus, the hard drive will collapse to a few hundred kb/s while the SSD’s remain at several tens of MB/s.
Morality:
In the end, you’ll understand, the smaller the files, the more access time is paramount in performance. And most of the software, and especially operating systems, are made up of a myriad of small files. Thus, new or older SSDs are by far more efficient in these conditions, thanks to their very low access time. Because, as we have seen, a low access time allows the small files to be processed very quickly compared to a disk with a much larger latency, regardless of the sequential bitrates.
In the games, the finding is less glaring, because there are also many small files, but also large texture files, for which there is not a blatant discrepancy between hard disk and SSD.
Of course, the figures that we have released are not measures but theoretical figures derived from simplified calculations. But they give an idea of the overall behavior of a storage unit depending on the size of the files manipulated.
Morality of the whole story: for a system disk, sequential bitrates matter little in the end, especially look at the performance on small files, which depends a lot on the access time. In addition, the files, even large, fragmented into small files, behave like small files. So very high performance drops for the hard drive, while the SSD is doing much better. Hence the absolute need to defragment the hard drive, while a SSD does not necessarily need it.
So if you’re using a SSD as a system disk, say you bought it more for its 0.1 ms access time (see less times) than for its 500 MB/s of sequential bitrates, read or write.