Bad to secure-erase a Samsung SM951 AHCI M.2 PCI-E SSD?

[VirtualLarry]

I did so, a few days ago, because I wanted to re-format my desktop, possibly in prep for selling it off.

Anyways, while I was re-installing drivers after installing Win7 64-bit, during some of the driver .ZIP file extracts, it said my data rate was in KB/sec, not even MB/sec.

I know that on my Brix J1900 units, with Intel 320 300GB SSDs, I was getting like 10MB/sec for extract speeds.

I did a benchmark with CDM on the SM951 after installing things, and it mostly seemed OK, although the 4K random writes QD32 were a tad low compared to what I had benchmarked during the prior install of Win7 64-bit.

Then I remembered something that I had read about NOT doing a "Secure Erase" on the SM951, for whatever reason.

I did some web searching, and in one review that I read, it had a mention that one of the distributors of these OEM drives, RAMCity, made mention to NOT SE these drives, lest they become "bricked".

Now I'm wondering, if perhaps there is a firmware bug, regarding their SLC write cache, and the SE procedure. What if the SE code, resets the entire drive, and doesn't properly set up the SLC cache region of the drive again properly? At least, that's the stab-in-the-dark guess at what happened, why write speeds seem to be slightly down.

Hellhammer, can you comment on this issue? Is it in fact not recommended to SE a SM951?

 

[Elixer]

Hmm, secure erase is part of the specs, highly doubt they would miss that in testing, and if it was an issue, you would think the SM951 would just ignore the command.

 

[Hellhammer]

Chris Ramseyer from Tom's (he was at TweakTown back then) broke one by secure erasing it using Parted Magic. I never experienced such issue (secure erasing is a must for reviews), but I made sure to erase using an Ubuntu distro and hdparm commands instead of Parted Magic. Not sure if it really matters, but it's possible that Parted Magic runs the commands a little differently and causes the SM951 to break somehow.

 

[VirtualLarry]

I used Linux Mint 17.3 Cinnamon 64-bit LiveUSB, and used hdparm commands manually.

So I guess I should be good? Still a bit confused over what seems like slightly slower 4K QD32 random-write speeds. Maybe that's just... what it is. But it seemed slightly perceptibly slower too, in usage, after the secure-erase and re-install.

Edit: Thanks for the feedback!

 

[Hellhammer]

I used Linux Mint 17.3 Cinnamon 64-bit LiveUSB, and used hdparm commands manually.

So I guess I should be good? Still a bit confused over what seems like slightly slower 4K QD32 random-write speeds. Maybe that's just... what it is. But it seemed slightly perceptibly slower too, in usage, after the secure-erase and re-install.

Edit: Thanks for the feedback!

What sort of speeds are you getting? What capacity and OS is this? The SM951 doesn't have very high 4K32 performance, but then again that test is irrelevant in real-world usage.

 

[VirtualLarry]

OS is Win7 64-bit SP1 w/updates.

Platform is ASRock Z170 Pro4S ATX, with "Ultra M.2 32Gbit/sec". (PCI-E 3.0 x4)

SSD is SM951 128GB AHCI M.2 PCI-E 3.0 x4 model.

Before CDM:
Seq read 2000MB/sec, seq write 680MB/sec
4K QD32 read 240MB/sec, 4K QD32 write 230MB/sec

After CDM:
seq - same
4K QD32 read, same, 4K QD32 write, under 200MB/sec

 

[Hellhammer]

What if you run just the 4K QD32 test, does it make it faster? One possibility is that there are backgrounds tasks going on in the OS, which is typical for new installs. Might be worthwhile to give it a day or two and then benchmark to see if anything changes.