I’ve been using m.2 drives and even u.2 drives for a little while now. I love the performance boost, and size over spinning disks and even over most SATA SSDs. The question I am posting here is one I’ve been curious about myself. Gen3 NVMe drives claim 2-3x higher transfer rate (read and write) over SATA SSDs. For example, top end SATA drives claim 550 MB/s -ish speeds, whereas NVMe claim up to 3500MB/s transfers. This should be monumentally different when using them. Right? Gen4 has taken that blistering speed even farther by claiming as high as 7,000MB/s read and 5,000 MB/s write. This seems ungodly. I mean this should be time warp speed. Compared to the speeds when I was first building PCs of 66 MB/s (PATA speeds) this is certainly a huge leap in performance.
To add to the confusion, manufacturers have a plethora of acronyms that are used to describe features and types of components on the m.2 drives. These make a bit of difference as well. You have different controllers on the SSD that can affect the speed and then you have different types of memory chips. SLC (Single Level Cell), MLC (Multi Level Cell), TLC (Triple Level Cell), and now QLC (Quad Level Cell). If you want a good explanation of what each of these means, you can head over to here and read up. Essentially SLC is expensive and the highest performing and longest lasting. QLC is the cheapest, slowest, and lasts the least amount of time. Keep in mind however that for home use, even QLC is in most cases longer lasting than most people will need and is better performing than the fastest spindle disk.
So how do we cut through all the marketing fluff and figure out if there is actually a benefit? Specifically, in this article, to PCIe 3.0 vs Gen 4. On the face of it, seeing the above numbers you would think there would be a good performance increase. In my own testing – I have not seen it. If you are a fan of benchmark numbers and you like bragging rights, sure the PCIe 4.0 number put up higher numbers. Here is a case in point. Here is a benchmark for a Crucial P5 1TB PCIe 3.0 SSD.
Not bad, right? Here is the same test on the same system from a PCIe 4.0 WD Black SN850 512GB SSD.
While under what is the “up to” claims, still big numbers vs the previous. This is a bit of unhappiness in the claims vs what I actually got as well. The claims were up to 7,000 MB/s Read and 5,000 MB/s write and up to 1,000,000IOPS. I never hit those numbers not once. Crystal Disk was the most flattering numbers. When I tried a few other good benchmarks out there such as Anvil I got smaller numbers.
As shown here, the numbers received were 4675 MB/s Read and 3864 MB/s write. Max IOPs at 4k blocks with a queue depth of 16 was almost 550k IOPs. A great number to be sure, but not 1 million (Dr. Evil insert). This was the first slight disappointment. None of the other tests gave me numbers any better than Crystal Disk.
The next question was loading times. Loading times were extremely similar between a Gen 3.0 and a Gen 4.0 SSD. I was curious if other people saw the same thing and so did a bit of research. I found an article on PCGamesN by Jacob Ridley that had the same question and tested it. His article is here. TL:DR – there was essentially no difference and at least once, the Gen 3.0 drive actually performed faster.
At this point unless you are just buying faster because you’re (like me), someone that loves bench numbers, I don’t see any cost benefit of buying a PCIe 4.0 drive vs taking that same money and buying a larger PCIe 3.0 drive. Or buying a better type of memory chip on the SSD (like a Samsung Pro drive). If someone out there has experienced different – contact me!
Disclaimer: This may change in the future if games and other programs figure out how to take advantage of the higher bandwidth and lower latency that PCIe 4.0 DOES afford. Perhaps going the way of Intel and making personalized NVMe drivers for their drives vs the vanilla Microsoft stornvme.inf driver.