Kusky :
I'm currently looking at SSDs to put in my laptop, and I came across the thinner type SSD that are called m.2 I think. I've learnt that they're way faster than their SATA3 counterparts.
They are not "way faster". They're slightly faster - you probably won't be able to tell the difference between the two in real-life use outside of a few esoteric use cases.
The problem is the benchmarks are all done in MB/s. Humans don't think in MB/s. They think in terms of how long they have to wait for the computer to complete an operation. That's wait time or sec/MB - the inverse of MB/s. When you measure theses SSD speeds on sec/MB, the huge MB/s increases end up making a very small difference. Imagine you need to copy 1 GB of sequential data.
125 MB/s HDD = 8 sec
250 MB/s SATA 2 SSD = 4 sec
500 MB/s SATA 3 SSD = 2 sec
1 GB/s PCIe SSD = 1 sec
2 GB/s NVMe SSD = 0.5 sec
Notice how every time MB/s doubles, the amount of wait time reduction is only half the previous step? So the huge 1000 MB/s speed increases we're at now translate into only tiny fractions of a second reduction in wait time. The biggest wait time reduction was from moving from a HDD to a SATA SSD. The increased "huge" MB/s speed increases of M.2 SSDs translate into only a tiny reduction in wait time. If you call the speed increase from a HDD to a 3 GB/s latest-gen NVMe SSD 100%, then:
250 MB/s SATA 2 SSD = 4 sec wait time reduction = 52%
500 MB/s SATA 3 SSD = 6 sec wait time reduction = 78%
1 GB/s PCIe SSD = 7 sec wait time reduction = 91%
2 GB/s NVMe SSD = 7.5 sec wait time reduction = 98%
3 GB/s NVMe SSD = 7.67 sec wait time reduction = 100%
So basically, the SATA 3 SSD gives you 80% the wait time reduction of the fastest M.2 SSDs. You're probably not going to notice the 20% difference between it and the M.2 drive, but the M.2 drive is probably going to cost you about 50%-100% more.
Also, because MB/s is the inverse of what's important, the
smallest MB/s figure actually ends up being most important. Say you wanted to copy 1 GB of sequential data and 200 MB of 4k data. Say you're comparing the 970 EVO NVMe SSD with 3 GB/s sequential read speeds and 33 Mb/s 4k read speeds, vs the SATA 860 Pro with 550 MB/s sequential read speeds and 50 MB/s 4k read speeds. Which will be faster?
If you think in MB/s, you see 3 GB/s vs 550 MB/s - the 970 EVO is nearly 6x faster. 50 MB/s vs 33 MB/s - the 860 Pro is only 1.5x faster. And you're copying 5x as much sequential data (where the 970 EVO is 6x faster) as you are 4k data (where the 860 Pro is only 1.5x faster). This should be an easy win for the 970 EVO right? Let's crunch the numbers.
970 EVO = (1 GB / 3 GB/s) + (200 MB / 33 MB/s) = 0.33 sec + 6 sec = 6.33 sec
860 Pro = (1 GB / 550 MB/s) + (200 MB / 50 MB/s) = 1.81 sec + 4 sec = 5.81 sec
Surprise! The 860 Pro ends up faster because it's the
smallest MB/s benchmark which makes the biggest difference in terms of wait time. Notice how both drives spend far more time on the 4k read than they do the sequential read. Basically, unless you're constantly copying large files between NVMe SSDs, or constantly reading large sequential files (e.g. real-time video editing), you should ignore the sequential speed benchmarks. Compare drives based on their
slowest benchmarks - the 4k speeds. That's the one which makes the biggest difference in real-world use. Because of its faster 4k speeds, the 860 Pro will be faster in most real-world tasks despite it being a SATA drive.
It's that pesky inverse. It means smaller is bigger, and bigger is smaller.