@Fernando
I observe when using Samsung 950 PRO as system drive, Windows registers 'unexpected shutdown' time stamped events and Samsung 950 PRO registers 'unsafe shutdowns' non time stamped events.
So, I agree it is difficult to correlate these events.
May be the source of the problem is when I am asking to shutdown the PC system via the Windows panel control and if for any reason the shutdown process is an excessive time long (may be due to any software conflicting buggy applications) an automatic Windows 'time-out' process occurs somewhere and this 'time-out' is logged as an 'unexpected shutdown' event by Windows.
May be in that case Windows stops abruptly SSD 950 PRO which registers an 'unsafe shutdowns' event.
For me this cascade of all these events is possible but it is strange that when using the i750 SSD system drive no such 'events' are observed.
May be 'abrupt stop' is less 'clean' for 950 PRO than for i750. (it depends of driver/controller/firmware/bios).

@Fernando,
Many thanks for your data analysis.
I agree that a problem is with Samsung NVMe BIOS module or may be Samsung NVMe driver itself.

For i750 does an 'Intel NVMe BIOS' exist ?

You are lucky to have only "83" 'unsafe shutdowns', it is 4 times less than on my 950 PRO device, but i see you have 8 times less than me on 'Power On Hours'.
So, it is sure that a proportianility rule does exist with 'Power On Hours' and 'unsafe shutdowns' and that problem does exist from the beginning of the product.

For me it is a 'misconception' of Samsung pieces of firmware/Bios NVMe, not the 950 PRO by himself, and we discover this "side effect" after some hours of usage and also we can compare , in the time, with other SSD NVMe manufacturer's solutions.

On my side, i don't understand the huge value of 'unsafe shutdowns' because I have observed only 1 or 2 'freezings', no more, from the time I have plugged in this device in a system.
'freezing' is not the unique explanation of the 'unsafe shutdowns'.

@Fernando,
I agree with you Samsung NVMe driver v1.4.7.17 is buggy.
I have uninstalled it and installed the native Standard NVMe driver.
Additional a comment:
Reliability (native driver) vs Speed (Samsung driver).
Here are screenshoots to compare ANVIL benchmarks results:


Here are screenshoots to compare (SIVx64 tool) SMART infos with 'native NVMe driver' vs 'Samsung NVMe v1.4.7.17':


After installed MS native Windows 10 native NVMe drivers (v10586 or v14366) on all my PCs equipped with a Samsung 950 PRO I can observe NONE '"Windows was not properly shutdown" reliability error reported !

So, for me I have a solution to this issue with NVMe 950 PRO SSD device I had observed from several months ago on all my PCs equipped with: don't use the Samsung NVMe driver v1.4.7.17 !!

Hey 100PIER, I was wondering. Was there not a Samsung NVMe driver which was one version down from the 1.4.7.17. I think it was 1.4.7.16. Wondering if that one is still around and if it too is buggy or maybe with the .16 does not exhibit the unsafe shutdowns.

I can add to confirm that since removing the 1.4.7.17 I have not seen an increase in unsafe shutdowns.

Samsung 950 Pro
Driver: Samsung v1.4.7.17
Power On Hours: 2839
Unsafe Shutdowns: 486

I am seeing this also.

Hi!

I have the same problem, Gigabyte GA-Z97X-Gaming 5. Both Windows 7 and Windows 10 (same driver), after reboot the computer freezes after the boot menu. If I turn off the computer wait 5 secs, turn on, the boot is successful.
The unsafe shutdowns property is incrementing.

The problem is not present with the default Windows 10 NVMe drivers - although the AS SSD Benchmark reports ~50 MB/s writing speed with it, so I keep the Samsung drivers.

Thanks Fernando. I have installed the 1.4.7.16 and I will run it for a while to see if the unsafe count increases under this driver.

FYI, since having removed the 1.4.7.17 driver and using the standard one I have had no increase; still sitting at 12 since July, pic included

http://ssd.samsungsemi.com/ecomobile/ssd...s_Driver_21.zip

v.2.1

http://www.samsung.com/semiconductor/min...uide_Rev2_1.pdf


Unsafe shutdown bug fix
on Windows 8.1 & 10

@imsims
Yes, i know v2.1 will finally close this NVMe driver issue I observed several months ago when posting the start post.
I had strong doubt about the Drivers v1.4.7.16, v1.4.7.17 and v2.0.0.1607
I wait for the 24h download limit exceed autorisation delay to get it and do tests.
I ignore if a possibility does exist for the "0D" smart parameter value to be zeroed...

here is