Intel 600p1tb extremely slow writes after firmware update PSF121C

[JStid]

After updating to the latest firmware my write speeds are that of a normal HD.

[idata]

Hello everybody,

We have a final statement about this situation, please read it carefully as it explains the current situation:As referenced in the Intel® Firmware Update Tool Release Notes: https://downloadmirror.intel.com/26491/eng/Intel_SSD_Firmware_Update_Tool_2_2_3_Release_Notes_019.pd..., we made a change in the 121C FW to correct behavior of the Force Unit Access feature. This will impact write commands utilizing the FUA feature. When measuring user experience and the performance impact of this change, the results are dependent on the benchmark being used and how write commands are implemented. The AS SSD tool has a significant dependency on FUA and performance results are substantially impacted; however, other tools like CrystalDiskMark, IOMeter and PCMark show little to no impact. We feel user experience should not be judged based upon a single tool, but should be evaluated and validated with several tools. We do not rely on AS SSD to judge the performance of our Intel SSDs.   The change was made to reduce susceptibility to data loss in unsafe power off conditions. It has been noted that turning off Windows Buffer Flushing brings the write performance back up within the AS SSD benchmark. This is due to the fact that FUA is disabled in this case and should be done with caution.Any other questions you may have, please let us know.Regards,Nestor C

[GNo]

With all due respect, but this is total nonsense. It's not about AS SSD Benchmark, it's about the day-to-day use implications of the changes you made. It is well-known that the default Windows NVMe driver uses FUA, so - together with the fact that you chose not to provide a custom driver for these drives - in everyday use cases your changes have impeded performance tremendously. This is unacceptable, unless you either provide us with a way to get back to the previous situation (i.e. a firmware downgrade). Personally I don't give a rat's ass about the unlikely occurrence of power loss. It'll probably happen once in a lifetime at most, and the few kilobytes of data I might lose at that time can be missed. If that obliterates the performance for which I bought this drive, it's definitely not worth it.

So, basically you're again focusing on scoffing a benchmark tool, while completely ignoring the implications on everyday use.

"the performance difference was due to the Microsoft NVMe driver creating FUA (Force Unit Access) I/O write commands. These FUA commands bypass the DRAM cache on the SSD and directly write to the flash, increasing the response time and also lowering bandwidth. For the same access traces, this situation does not happen with the Microsoft AHCI driver." (Anandtech)

[idata]

Hello all,

It seems that other vendors are ignoring the FUA bit as a fix, while we decided to no longer ignore it. You can check the following link as a reference:

http://forum.notebookreview.com/threads/potential-bug-with-as-ssd-benchmark-and-nvme-drives.803017/ Potential bug with AS SSD Benchmark and NVMe drives? | NotebookReview

Regards,

Nestor C

[GNo]

Vendors ignore the FUA bit as a fix, because they have their priorities straight: for end users, the (extremely low) chance of data corruption due to power loss does not, in any way, justify a permanent performance drop, especially not AFTER people bought a product for its previous performance results.

There is, however, an interesting statement in that topic: "In real life, the FUA command is rarely if ever used, so the driver is of little benefit." (a PC World editor in http://forum.notebookreview.com/threads/potential-bug-with-as-ssd-benchmark-and-nvme-drives.803017/# post-10494122 Potential bug with AS SSD Benchmark and NVMe drives? | NotebookReview)

This is not what I've read so far. Perhaps Windows has updated their NVMe driver since, but I've read in multiple articles and forum threads that its NVMe driver does indeed depend on FUA in everyday use. I would very much appreciate it if someone could find me a source stating that the Windows 10 NVMe driver doesn't use FUA by default anymore, or a way to check whether FUA is currently used. That would mean that everyday performance is not affected. The quote of the Lenovo engineer in that same post states the same, and a lot more specific than I've ever read it before, especially in this very topic.

It sucks that I never did any testing with CrystalDiskMark before firmware 121c, would've loved to have had the opportunity to test the drive in my own system before/after.

[SDong6]

You tell us："It seems that other vendors are ignoring the FUA bit as a fix", it is only based one of the editors at PC World, can he represent other vendors? Lenovo engineer say "The problem is that the AS SSD Benchmark is issuing commands with the caching set for Write Through instead of Write Back mode", so I think other vendors may provider a special driver that can translate these special cache related commands to normal sequense, that can't labeled as "ignore" but like "another smart method". And can you tell me why DC3700 has a speical driver ?