Extensive benchmarking done on WD Black SN750 - Help me make sense of this?

derpkotacasper · June 22, 2021

When I initially purchased my 500 GB WD Black SN750, I relegated it to the chipset M.2 slot on my board - it's a pcie 3.0 drive, so I put it in the pcie3 slot, leaving the cpu slot on my ASUS ROG Strix B550-F Gaming WiFi vacant for a future upgrade. Today, I came to the conclusion that I will likely be a ways out from an upgrade to my M.2 and I decided to swap it over and take advantage of the CPU lanes. However, being me, I decided to benchmark first and see exactly how much I was getting in gains. I saw losses in Disk Sequential 64.0 Write, Latency in the 95th Percentile and Maximum Latency, gains in everywhere else testing through winsat (Windows System Assessment Tool).

For each benchmark, I ran winsat ten times and averaged out each number. (Winsat is run through running command prompt as an administrator and entering "winsat -disk drive C:")

Build is an overclocked Ryzen 3 3100, 16 GB DDR4 2666 and ASUS ROG Strix B550-F Gaming. Rest of the build feels irrelevant, but is nonetheless in my signature. 86 of 465 GB usable storage was in use for the testing both in my baseline and after the move testing.

My questions: Why the increase in latency? In some individual tests, I found extremely large increases in latency. Why did I lose average Disk Sequential 64.0 Write speed? How do I identify which revision (controller, specific dram and nand flash) my SSD has? I'll enclose a picture of the drive.

Benchmarking:

Baseline in chipset m.2 lane

Disk Random 16.0 Read 1266.999
Disk Sequential 64.0 Read 2706.985
Disk Sequential 64.0 Write 2173.649
Average Read Time with Sequential Writes .0623
Latency: 95th Percentile .2152
Latency: Maximum 3.7993
Average Read Time with Random Writes .0658

CPU m.2 lane

Disk Random 16.0 Read 1312.743
Disk Sequential 64.0 Read 3088.771
Disk Sequential 64.0 Write 2025.292
Average Read Time with Sequential Writes .0564
Latency: 95th Percentile .2199
Latency: Maximum 4.3029
Average Read Time with Random Writes .0618

Increase in MB/s 45.744
Increase in MB/s 381.786
Decrease in MB/s -148.357
Read Time with Sequential Write improvement in ms -0.0059
Latency increase 0.0047
Latency increase 0.5036
Read time with Sequential Write improvement in ms -0.0040

Edit: I can also provide the data from each of the individual ten tests run for both of my averages on request. It's just a lot to type if it's not pertinent.

jaslion · June 22, 2021

Is windows on this drive? If so benchmarking it is kinda invalid as well windows will windows. For example my sx8200 (the original release one) varies MASSIVELY if I bench it now or in half an hour. Not due to temps but just because windows introduces too many variables.

If this is a windowsless drive but has programs those are also a good variable.

Could also literally be that from moving it down it has more airflow and runs cooler. Part from 64 read all these numbers are margin of error.