I have 2 gen4 SSDs in Raid0, which give me 14000MB/s in sequential large benchmarks (equivalent/faster than a single Gen5 SSD).
Expected results:
Copying large file on same Gen5 drive = 7GB/s
Actual results:
Copying large file on same Gen5 drive = 2GB/s
Using a 3rd party program to copy the large 10GB files, the transfer speed seems to be largely affected by the buffer size. By default, it's 256MB. Upping it to 512MB and above massively increases transfer speed.
It is clear that Windows has reached the limit of its file handling capability. I cannot find a way to increase the default windows buffer size for copying files, after exhausting my google-jitsu.
For years, storage benchmarks show no real-world improvement in loading times etc. For years, this phenomenon has been assumed to be due to a CPU bottleneck, which may be true to an extent.
I postulate that the issue may well be Windows - Windows simply cannot make use of modern SSDs to their maximum performance capability. If this is the case, then all our storage benchmark results and conclusions have been misguided for quite some time - i.e. no real difference between game load times vs SATA SSD / PCIe 3 / PCIe 4 / PCIe 5 / NVMe drives.
This is an interesting discussion in and of itself, but I would also appreciate any insights into finding out how to increase Windows buffer size for transferring files. I would like to do more tests...
Godawful video.
A few side by side pictures would have been a superior substitute. That's 10 minutes of my life I'll never get back.