Cannot recognize Optane 900p M.2 on X399 Taichi

Has anyone gotten the Optane 900p to work in non-RAID mode on the X399 Taichi motherboard?  I cannot get the bios to recognize it at all and wondering if anyone was able to do so.  I am using the M.2 connector.

jamesjhhan wrote: Has anyone gotten the Optane 900p to work in non-RAID mode on the X399 Taichi motherboard?  I cannot get the bios to recognize it at all and wondering if anyone was able to do so.  I am using the M.2 connector.

What UEFI/BIOS version are you using?

An Intel 900P should work with any PC, it is a general use SSD. Only the M.2 Optane SSDs used to accelerate a SATA HDD or SSD require specific support from the UEFI/BIOS to work correctly.

Do you have the SATA power cable connected to the U.2 to M.2 adapter cable used by your 900P?

If you boot the PC into Windows, and then start Disk Management, you should be prompted to initialize and format the 900P. If you were to install Windows 10 RS2 (required by X399) it should be recognized fine even if not initialized and formatted.

jamesjhhan wrote: Sorry, I thought I had posted my follow up response.  I did end up getting it to work.  Because its an M.2 it's not like an nvme and needs a separate power source.  I didn't think to connect the SATA power cable.     One thing that is a little odd.. while it is fast, the SSD speeds I am getting is nothing close to what others are getting on other Intel platforms.  On AS SSD benchmarks and Crystal Diskmark, I get a score thats barely a bit above what Samsung PROs are getting.  I get 75 MB/s on 4K and 414 MB/sec on 4k-64 depth on Read.  Also, weird, I got a really really good score initially after a fresh install of Windows and installation of base drivers (174 MB/sec 4K, 2297 MB/sec on 4K-64depth on Read).  But after a few reboots, I get significantly lower scores, especially on the 4K 64 queue depth.  I did not install anything in between reboots.  It a bit odd for sure.   And, yes, I am running Intel's NVME driver which is supposed to be better than the Windows nvme one.   I expected better speeds.  If anyone has any ideas, I am all ears.

Your model of the 900P may use an M.2 interface adapter, but the 900P needs much more power than any M.2 SSD. There will never be an M.2 version of the 900P, its controller uses much more power and could not physically fit in the M.2 form factor. The Anandtech review of the 900P goes over this well:

https://www.anandtech.com/show/11953/the-intel-optane-ssd-900p-review

Regarding the apparent lower performance, or low benchmark scores, two things.

In general we have noticed with Ryzen AM4 boards and processors that the 4K random and high queue depth performance for SATA and NVMe SSDs in benchmarks is lower than on Intel systems. Since X399/Threadripper is generally based on the same Ryzen architecture, I'm not surprised that it acts in the same way.

SSD performance is affected by CPU and PCIe power saving options. If the CPU C States (C6 for Ryzen) is enabled, and Cool n Quiet is enabled, the benchmark scores will be lower. That is true for the equivalent options on Intel systems too. The Windows Power Plan being used also affects SSD performance. AMD recommends the Ryzen "Balanced" or Windows High Performance power plans. The UEFI/BIOS of newer Intel boards have options to enable or disable several PCIe power saving features, which we don't find in the UEFI's of Ryzen and apparently X399 boards too. The PCIe power saving options may be enabled by default for Ryzen/X399, but I don't know if that is the case or not. If they are, it would potentially explain at least some of the lower 4K performance.

You coincidentally have brought up something about the 900P that I have been wondering about. The Anandtech review in the link above, and in the followup review of the 480GB 900P, and the Intel DC P4800X review mention some kind of "data refresh" activity that these SSDs perform. That apparently happens when power to these SSDs has been "removed for a period of time", and can take a few hours to complete. A shutdown of the PC should qualify as a removal of power, right? The reviews just barely mention a difference in performance results at times, but never discuss it in detail. I was wondering if the "data refresh" action would cause a reduction in performance, purely a thought or theory I have.

Your statements are the first clues that this might be the case, but as I've learned being a moderator, never jump on a theory as being true with only one clue and a lot of unknowns as the only basis for a conclusion.

The lower 4K performance for Ryzen has been seen by myself and other people in the forum with SATA or NVMe SSDs. In my case, both with SATA and NVMe SSDs. Also I have seen what I would call erratic benchmark performance results in the 4K tests, sometimes lower than previous tests, other times higher. But I only have used one 500GB 960 EVO in my X370 PC, so no where near enough a sample size of NVMe drives. I tend to think it's more likely the 960 EVO, but again not enough data to reach a true conclusion.

Use the Windows High Performance Power Plan, disable C6 and Cool and Quiet in the UEFI, and run a benchmark again.

I am pretty happy with my SATA SSD performance;

Mushkin 1TB 2.5

Edited by TANWare - 30 Dec 2017 at 12:08am