Z170 Extreme7+ random Dr. Debug codes

Yes, matched DIMMs, particularly in large capacities like you have (16GB) are less likely to have differences that can cause problems. The more memory chips on each DIMM, the more important matching will be. Your current configuration, one set in the '2' slots, and the other set in the '1' slots, is mixing unmatched DIMMs in each channel. You may have better luck with them using one set in the A slots/channel, and the other set in the B slots/channel.

Memory matching is not a marketing scheme, but can be cheated at times. I have experienced both, using mixed sets working and not working. Pure luck or bad luck.

jnmanocchio wrote: The M.2 SSDs are installed in the onboard M.2 slots (not using PCI-E adapters) using PCIe remapping in RAID0 thru the Intel Rapid Storage Technology configuration. Is it true that the remapping to PCIe does not affect the SATA and SATA_EXP ports on the MB, i.e., "M2_1, SATA3_0, SATA3_1 and SATA_EXP0 share lanes. If either one of them is in use, the others will be disabled." Or does the remapping for the M.2 ports still share the PCIe lanes with the associated SATA ports?

The Z170 chipset provides the only resources used for the M.2 slots, SATA Express, and SATA ports. PCIe Remapping does NOT change that. A SSD in one M.2 slot uses the resources of two SATA ports, as one SATA Express connection does as well, PCIe Remapping enabled or not. That option allows PCIe RAID to work, and does nothing to free up resources or use other resources. That is hardware limitation of the Intel Skylake platform, all Skylake boards are the same.

The previous generation Z97 boards used the PCIe lanes from the CPU for the M.2 slots. But that takes away resources from video cards. Use one PCIe M.2 SSD in a Z97 board, and a video card will only have 8 PCIe 3.0 lanes for itself.

Why not use the Z97 chipset resources? That is possible, but the resources in the Z97 and all earlier Intel chipsets are equivalent to PCIe 2.0 in speed. Our Intel 750s and Samsung 950s need PCIe 3.0 bandwidth to reach their full potential. The Z170 chipset is the first from Intel with PCIe 3.0 level resources.

So which way do we compromise? Use the video cards resources, or the storage IO resources? We don't get to completely choose, but gamers did not like losing the 16 PCIe 3.0 lanes from the CPU to a SSD. More users will accept losing SATA ports. The best we can do is use M.2 to PCIe adapter cards or PCIe form factor Intel 750s, but I don't know if the PCIe RAID would still work mixing drives on the CPU and chipset resources. Actually it would for data drives, but may not for bootable drives. Regardless, our choice is made purely with hardware, not software switching which would be great.

jnmanocchio wrote: The main reason for me purchasing this MB was to run the M.2 SSDs in a RAID0 configuration but if I update the UEFI or clear BIOS the RAID will be lost? Should I look into using another MB then?

The PCIe SSD RAID support is supplied purely and only by the Intel Z170 chipset, and the appropriate Intel RAID software, including EFI Option ROMs that are part of the UEFI/BIOS file. Of course a mother board manufacture must supply those with the board to allow PCIe RAID to work, but a different mother board manufacture cannot change or fix the way the PCIe RAID works by themselves.

The situation with the fragile RAID arrays (my terminology) is real, I've experienced it with Samsung SM951s and 950 Pros. Other users in this forum have also experienced it themselves, it is a sad reality.

Try it yourself: create a RAID 0 array (or any type, like RAID 1) of 950 Pros with an OS installation on another drive. The RAID 0 array will just be a storage drive, right? Don't put anything on it you don't want to lose, keep it empty. Then shutdown the PC, cut power, and clear the UEFI/BIOS. Start the PC again, go into the UEFI and check the status of the RAID array. It will be failed, sorry to say. Same thing for a UEFI update of course.

On one hand this is a bad situation. On the other, we have the ability to create RAID arrays on drives using two completely different storage protocols, SATA and NVMe, using the same underlying hardware resources. That is an amazing achievement IMO. When have we had multiple protocols supported with the same hardware? USB is separate, for example.

In reality, I have never had an OS drive of any type or configuration, boot an OS in less than 2 - 3 seconds. That is from the POST beep (boot begins) to the Windows desktop. It's the same for one good SATA SSD, multiple SATA SSDs in RAID 0, or two 950 Pros in RAID 0.

Yes I'm being apologetic for Intel, but remember this is the first time we've had RAID in PCIe NVMe SSDs that were bootable. IRST version 14 gives us that, along with the Skylake platform. AMD has nothing like this. Yes IRST needs work and is not perfect, and I'm just trying to give you all a warning FYI. If any other mobo manufacture is not affected by this, I am not aware of it.

Hopefully Intel will improve the IRST driver, or a UEFI/BIOS fix will be found.


Edited by parsec - 07 Mar 2016 at 1:43pm