danann

I'd go for the larger SSD. Games are crazy huge nowadays. He can always add a hard drive later. There's not much noticeable difference in real world usage speed between NVMe and Sata SSDs - I think there's even a LTT video about it where people picked wrong the fastest on blind tests.

For comparison, here's (attached) sequential copy to a ST4000LM024 (2.5" 4TB SMR) drive. Slower because it's nearly full and 2.5" over USB, but surely consistent with not as crazy response. It's bad, but consistently at it. It's ok to be slow, I'm just worried my new drive is defective.

Aight. Now for some serious constructive answer, please? I have 5 Barracudas, including another SMR one, and none of them behave like this new one.

I purchased a Seagate ST4000DM004 (4TB SMR drive) for archival of media files and I'm wondering if it's defective or what. I knew it was SMR but it should be fine for the use case and there was a decent price difference. Plugged in, simple benchmark, run Seagate's tool tests, full format as a simple test - write speed (as indicated by Windows Task Manager) was about constant during the whole format, apart from the expected gradual slowing as it gets into the inner tracks). Checked SMART and everything looks fine. Let it idle for a few hours. Enabled Bitlocker and proceeded to copy some files into it. I'm copying about 2TB of some ~3000 files to it, so it's a very sequential workload, but I noticed the write speed is all over the place and response time is over 10s (hitting 20s sometimes). Like, what? See screenshot. Is this expected behavior because it's SMR or is it defective? Or maybe something else is wrong? I thought SMR drives were fine for sequential workloads. As a matter of fact, I have another Seagate SMR drive (an external 4TB one) and this doesn't happen with it. I'm keeping an eye on SMART and all still looks fine. Checked for firmware updates and there's none. Thanks!

Looks like cinebench was run in Silent mode (fan LED was green rather than white, and he only changed it back at the end during the photoshop test). Rather, he should just neuter Windows and disable the store, uninstall all pre-installed apps, and disable windows update before connecting to the internet, like people do after a windows 10 install to prevent that madness. A simple powershell script and he could have this done in a few seconds on every machine he needs to benchmark.