ASRock.com Homepage
Forum Home Forum Home > Technical Support > Intel Motherboards
  New Posts New Posts RSS Feed - Samsung 950 pro m.2 raid 0 sleep problem
  FAQ FAQ  Forum Search   Events   Register Register  Login Login

Samsung 950 pro m.2 raid 0 sleep problem

 Post Reply Post Reply Page  <1234>
Author
Message
parsec View Drop Down
Moderator Group
Moderator Group
Avatar

Joined: 04 May 2015
Location: USA
Status: Offline
Points: 4996
Post Options Post Options   Thanks (0) Thanks(0)   Quote parsec Quote  Post ReplyReply Direct Link To This Post Posted: 12 Jul 2016 at 6:03pm
Originally posted by clubfoot clubfoot wrote:

There is an option to lock the buses at GEN3,...but I'm thinking that is only for the PCI graphics bus as changing it to GEN3 lowers the graphics scores :( Is there a UEFI option to lock the DMI for the m.2s?

I've always enabled the remapping for my two m.2s,...are you saying I do not need to?

In my "normal" setup i.e. o/c at 4.5Ghz with C States disabled, Speed Step disabled,...my comp will not sleep :) Tomorrow I can retest the cpu speed drop from sleep to see if that is what is causing it.


OMGosh, NO!

When the UEFI/BIOS is cleared, or after an update, the PCIe Remapping options are Disabled.

My theory, and it is just a theory, is the failure of a PCIe NVMe RAID 0 array after a UEFI/BIOS clear or update is caused by the PCIe Remapping options being disabled. I don't know if that is true, but it is the only thing I can think of that causes this to happen.

If C States and Speed Step are disabled, your PC will not Sleep? Interesting, never tried that since I normally have those options enabled.
Back to Top
clubfoot View Drop Down
Newbie
Newbie
Avatar

Joined: 28 Mar 2016
Location: Canada
Status: Offline
Points: 246
Post Options Post Options   Thanks (0) Thanks(0)   Quote clubfoot Quote  Post ReplyReply Direct Link To This Post Posted: 13 Jul 2016 at 5:42am
I don't think it is the remapping being disabled as I can F9, save and reboot, and enable raid and remap and not lose my array. I can also flash between 2.6? and 3.0 without losing my array,.... The earlier BIOSES,... No!

Magician doesn't give any vital info of a raid array, unlike IRST where you can see the individual attached drives.

Clear CMOS is allot "harsher" on an m.2 raid setup,... It must be the way it totally disconnects power from it and may be a "flaw" in the m.2 design,.... Re sleep mode.


Back to Top
parsec View Drop Down
Moderator Group
Moderator Group
Avatar

Joined: 04 May 2015
Location: USA
Status: Offline
Points: 4996
Post Options Post Options   Thanks (0) Thanks(0)   Quote parsec Quote  Post ReplyReply Direct Link To This Post Posted: 13 Jul 2016 at 10:37am
Originally posted by clubfoot clubfoot wrote:

I don't think it is the remapping being disabled as I can F9, save and reboot, and enable raid and remap and not lose my array. I can also flash between 2.6? and 3.0 without losing my array,.... The earlier BIOSES,... No!

Magician doesn't give any vital info of a raid array, unlike IRST where you can see the individual attached drives.

Clear CMOS is allot "harsher" on an m.2 raid setup,... It must be the way it totally disconnects power from it and may be a "flaw" in the m.2 design,.... Re sleep mode.




Interesting about the new UEFI versions that don't loose the 950 Pro RAID 0 array, I was not aware of that. Or is that the non-loss of power during an F9 and restart that is the key?

If the RAID 0 arrays are not lost ONLY in the scenario that you described, then there is a difference between that and a restart after a shutdown. POST is not always the same on a cold (after shutdown) startup, then it is on a restart of the PC. That may be a factor too.

Waking from Sleep issues that are common include loosing over clocks, and core multipliers being set to throttled values (such as 8 or 12). That is caused by the internal CPU registers that contain that data, not being reset correctly, or other registers used to signal thermal and power throttling being set to the throttle value, and the CPU's internal logic sees that and sets the core multipliers accordingly. The same thing may be happening with the DMI3 buss in the Z170 chipset. We don't know if that is happening.

I could swear I am seeing a few options in the UEFI that are not reset after a UEFI update, that are not default values. Such as RAID for the "SATA" Mode. There are also "background" setting in a UEFI/BIOS, that we are not aware of, or don't match certain option settings. Turbo Boost can behave that way, being set to Disabled by some users for over clocking, but is really enabled, since it must be for any OC multiplier OC from Sandy Bridge processors forward. I'm hinting here that PCIe Remapping might remain enabled even when it seems it isn't.

The ASRock OC Formula boards have switches on the board that can turn the PCIe x16 lanes off and on. It would be nice if the M.2 slots had such a switch in the UEFI update/shutdown situation. But that is a unique situation, and would be like a switch to turn off a SATA port, which is not needed for any reason.

I've always wondered about the option to disable the SATA Controllers in the UEFI, Storage Configuration screen. Would that also disable the M.2 slots on a Z170 board?
Back to Top
clubfoot View Drop Down
Newbie
Newbie
Avatar

Joined: 28 Mar 2016
Location: Canada
Status: Offline
Points: 246
Post Options Post Options   Thanks (0) Thanks(0)   Quote clubfoot Quote  Post ReplyReply Direct Link To This Post Posted: 13 Jul 2016 at 11:23am
I came across another interesting discovery while testing power plans,...default balanced power plan produces higher disks scores >3200 than default high performance plan >2900! Using balanced plan with HDD turn off = never, Link state = off, and UEFI APCI suspend to ram = disabled,...C states and Speed step disabled.

balanced power plan:
CrystalDiskMark 5.1.2 x64 (C) 2007-2016 hiyohiyo
                           Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

   Sequential Read (Q= 32,T= 1) :  3207.355 MB/s
  Sequential Write (Q= 32,T= 1) :  1912.526 MB/s
  Random Read 4KiB (Q= 32,T= 1) :   860.171 MB/s [210002.7 IOPS]
 Random Write 4KiB (Q= 32,T= 1) :   552.720 MB/s [134941.4 IOPS]
         Sequential Read (T= 1) :  3057.911 MB/s
        Sequential Write (T= 1) :  1919.841 MB/s
   Random Read 4KiB (Q= 1,T= 1) :    56.030 MB/s [ 13679.2 IOPS]
  Random Write 4KiB (Q= 1,T= 1) :   188.692 MB/s [ 46067.4 IOPS]

  Test : 1024 MiB [C: 37.7% (162.2/430.5 GiB)] (x5)  [Interval=5 sec]
  Date : 2016/07/12 23:00:44
    OS : Windows 10 Professional [10.0 Build 10586] (x64)


Edited by clubfoot - 13 Jul 2016 at 11:25am
Back to Top
Grant Rogers View Drop Down
Newbie
Newbie


Joined: 07 Jul 2016
Status: Offline
Points: 7
Post Options Post Options   Thanks (0) Thanks(0)   Quote Grant Rogers Quote  Post ReplyReply Direct Link To This Post Posted: 14 Jul 2016 at 4:20am
Hi guys,

I have just noticed I didn't have notifications ticked when there was a reply so sorry for the late response.

@Parsec, yeah I know I am taking my chances with RAID 0, the reason I did it is so I don't have worry about where the data is going, using them as two separate drives mean I would have to start using Windows mount points.  It also has the added benefit of making the performance even better than a single m.2.

Regarding the check in the Samsung Magician software I checked, but it doesn't know anything about the RAID:




It is correct at least for me that I am using the m.2 sockets.

Originally posted by Parsec Parsec wrote:

That's an interesting event, but something is strange about it. storahci is the native Windows 10 AHCI driver. The Intel IRST driver is iaStorA.sys, which you can confirm in the Device Manager entry, under Storage Controllers.

Regardless, the reference to \Device\RaidPort0 seems to be related to this situation. You should check the Event Log for anything about iaStorA.sys, if you manually installed the IRST 14.6 or 14.8 version driver, which I believe you must have done.


Yeah I was wondering about this, but the drivers look correct so I don't think I am using the base Microsoft ones.



Originally posted by Parsec Parsec wrote:

We also have the Samsung NVMe driver to potentially add to the list of reasons for this issue. A question that I am unsure about is, when using these SSDs in RAID 0, is the Intel IRST driver in complete control, and the only one that is used? The Samsung NVMe driver not being used at all?


I don't think the samsung driver is being used by Windows which is why Samsung magician doesn't recognise them.  It is all hidden by the Intel drivers.

Thanks for finding those notes about the bugs I hadn't seem them before, it is interesting.

Is there any particular reason you chose RAID when you are not using RAID configuration?

@clubfoot, thanks for confirming it is useful to know how many people are affected by this.

It is a shame that version 15 of the driver doesn't fix this issue either.

I will let you know if I get anything back from Intel / ASRock, and thanks for all the research it has been interesting reading.

One thing I will say about Intel's bug list is plenty of people had a problem with hdds not spinning down (not my problem), and one year later it still hasn't been fixed.  It doesn't give me much confidence they will fix the known issues quickly.

Lastly regarding the sleep thing, I have used it a lot on my other machines as wake up time is near instant and I can resume my work (mainly programming) without having to open all of the apps up again, an added benefit is that dropbox doesn't keep indexing on startup.  It is an inconvenience rather than a show stopper, but annoying none the less.

Grant
Back to Top
clubfoot View Drop Down
Newbie
Newbie
Avatar

Joined: 28 Mar 2016
Location: Canada
Status: Offline
Points: 246
Post Options Post Options   Thanks (0) Thanks(0)   Quote clubfoot Quote  Post ReplyReply Direct Link To This Post Posted: 14 Jul 2016 at 9:19am
Have you found anyway to correct the half speed on wake?

Did you notice if your cpu is locking itself down at a lower speed?
Back to Top
parsec View Drop Down
Moderator Group
Moderator Group
Avatar

Joined: 04 May 2015
Location: USA
Status: Offline
Points: 4996
Post Options Post Options   Thanks (0) Thanks(0)   Quote parsec Quote  Post ReplyReply Direct Link To This Post Posted: 14 Jul 2016 at 11:40am
@Grant Rogers, You don't need to justify your use of 950 Pros in RAID 0 to me! LOL

I update and clear the UEFI too often, which is one reason I don't use RAID 0 for my 950 Pros. I have one not being used currently, which is ridiculous! Confused

About the Magician software, it cannot recognize RAID mode. Been that way since it was introduced.

Early on, it could not work with AMD SATA chipsets using AMD drivers, in AHCI mode!

I have RAID selected in the UEFI, but no RAID arrays created. This is what Magician shows me when I have my Samsung 840 Pro selected:



Notice the Storage controllers in Device Manager. Regardless of what it states above about AHCI mode, RAID mode has EVERY feature that is provided by AHCI mode.

Changing to my 950 Pro, I see this in Magician:



While the Intel RAID software (IRST) must provide the NVMe interface to the SSDs (Intel has their own NVMe driver for their NVMe SSDs), I don't see any reason why it would hide the Samsung NVMe driver.

Regarding the Microsoft storahci driver, your 950 Pro RAID 0 array is not using the storahci driver, but notice in your Device Manager picture, you have an IDE ATA/ATAPI controllers entry.

If I select my Samsung SM951 AHCI PCIe M.2 SSD in Magician, I see this:



In your PC, that Device Manager entry is for your BD drive. It isn't for your HDDs, since they would be controlled by the Intel RAID controller, even as single drives. That Event Log entry should be related to the BD drive.

The Samsung SM951 is unusual in that is has its own storage controller chip as part of the SSD. It uses the Windows AHCI driver by default.

I always use RAID mode, whether I'm using a RAID array or not. I prefer to use the Intel storage driver (which can be used in AHCI mode.) Intel calls that "RAID ready", and the Intel IRST driver only supports PCIe SSDs in RAID mode. Plus I create a RAID array on this PC occasionally, I did that to test something for someone in this forum a while ago, can't recall exactly what it was. That was a crazy one, a RAID 0 of a Samsung 840 Pro and an older Crucial M4 SSD. Did it work? It sure did.

Intel knows IRST has issues. That software is provided as is, without any guarantees. If you want perfect (or closer to perfect), you must use the Intel RSTe driver, Rapid Storage Enterprise. That driver is not supported on the Z170 chipset, you must use a server board with a server chipset. The first HEDT platform, X79, supported RSTe. It's performance was less than IRST (reduced benchmark performance)

Wake from Sleep work fine for me on this PC. clubfoot said if he has C States disabled in the UEFI, waking from Sleep fails. I have C States enabled.

Back to Top
clubfoot View Drop Down
Newbie
Newbie
Avatar

Joined: 28 Mar 2016
Location: Canada
Status: Offline
Points: 246
Post Options Post Options   Thanks (0) Thanks(0)   Quote clubfoot Quote  Post ReplyReply Direct Link To This Post Posted: 15 Jul 2016 at 10:37am
In further testing of sleep IRST still shows PCIe link speed:4000MB/s and PCIe link width x4 for both m.2s after waking. So I don't think the driver is falling back to single drive.

However, the cpu on the other hand drops the core voltage to >.6xx and the multiplier is at 8x. If I run Crystaldisk the core increases to 1.3xx and the multiplier varies between 42 and 8 and everything in between as it should.

What is a little disturbing is that it never throttles up!? Stays at half speed as if has not fully awakened.

Parsec does this happen on your single m.2 with the raid driver? Because that would tell us if it is the IRST!
Back to Top
parsec View Drop Down
Moderator Group
Moderator Group
Avatar

Joined: 04 May 2015
Location: USA
Status: Offline
Points: 4996
Post Options Post Options   Thanks (0) Thanks(0)   Quote parsec Quote  Post ReplyReply Direct Link To This Post Posted: 15 Jul 2016 at 12:12pm
Originally posted by clubfoot clubfoot wrote:

In further testing of sleep IRST still shows PCIe link speed:4000MB/s and PCIe link width x4 for both m.2s after waking. So I don't think the driver is falling back to single drive.

However, the cpu on the other hand drops the core voltage to >.6xx and the multiplier is at 8x. If I run Crystaldisk the core increases to 1.3xx and the multiplier varies between 42 and 8 and everything in between as it should.

What is a little disturbing is that it never throttles up!? Stays at half speed as if has not fully awakened.

Parsec does this happen on your single m.2 with the raid driver? Because that would tell us if it is the IRST!

I want to be sure I understand what is happening in your description of your CPU before and after Sleep. I'll go through that in a bit.

First, CPU speed, an over clock, and CPU power saving options enabled and disabled (mainly SpeedStep and C States) will affect the benchmark results of a SSD. When I've seen that with SATA III SSDs, it's not that much of a difference, compared to the reduced 950 Pro RAID 0 results, which are reduced a lot more.

The SATA III SSD sequential read and write speeds will lose about 25MB/s from the usual ~500MB/s Read, and ~450MB/s Write speeds, when SpeedStep and C States are enabled. The benchmark test does not put enough of a load on the CPU to take it out of the SpeedStep CPU speed of 800MHz, for any Mainstream/Performance CPU (all Intel CPUs except the HEDT 6, 8, and now 10 core processors, whose SpeedStep speed is 1,200MHz.)

Add C States (C3, C6) to the situation, at idle the CPU will be in the C6 state (cores virtually off) ~90% of the time. When an IO request finally wakes up a core or two, it takes time for the core to get going again. That should be no time compared to the slow IO of our storage drives, but it adds up and the benchmark result difference is real, small but real for SATA III SSDs at least.

Back to your CPU description before and after waking from Sleep.

Your signature specs show an OC of 4.6GHz on your i7-6700K. You might use a manual, constant VCore. No SpeedStep or C States, both disabled?

Is that right, and what you used when you noticed your 950 Pro RAID 0 performance problem, after waking from Sleep?

You wrote this:

However, the cpu on the other hand drops the core voltage to >.6xx and the multiplier is at 8x. If I run Crystaldisk the core increases to 1.3xx and the multiplier varies between 42 and 8 and everything in between as it should.

Does the above happen after waking from Sleep? You loose your OC to 4.6, and the CPU is back to stock speeds, Turbo up to 4.2, and SpeedStep seems to be enabled, since you are down to 800MHz, depending on CPU load.

Then you wrote:

What is a little disturbing is that it never throttles up!? Stays at half speed as if has not fully awakened.

Is this referring to the 950 Pro RAID array? Or your CPU speed? I think you mean the RAID array.

Does that happen to my single 950 Pro? Check my first post on page 1 of this thread, third post from the top. Here it is again, after waking from Sleep, a 256GB 950 Pro:



Exactly what it should be given the specs. Plus this is the OS drive.

Does this really tell us if it is the IRST driver? It could be, but I'm not ready to say that yet.

Why? Because of what might be happening to your CPU speed after waking from Sleep. BTW, I do NOT loose my OC on my i5-6600K when I wake from Sleep. It's still at 4.4GHz, which I've been using lately.

If YOU are loosing your OC settings, and other weirdness like SpeedStep being enables, for the 800MHz CPU speed, then it is possible that OTHER settings somewhere are off too, that affect the RAID 0 array. Changes to CPU multipliers are a classic wake from Sleep failure of a BIOS. Why mine are fine and yours are not (if true) is even more weird. Pinch

That is why I need to understand what happens to your CPU speed when you wake from Sleep!! I could not be sure given what you wrote. Confused


Back to Top
clubfoot View Drop Down
Newbie
Newbie
Avatar

Joined: 28 Mar 2016
Location: Canada
Status: Offline
Points: 246
Post Options Post Options   Thanks (0) Thanks(0)   Quote clubfoot Quote  Post ReplyReply Direct Link To This Post Posted: 15 Jul 2016 at 11:45pm
>> Your signature specs show an OC of 4.6GHz on your i7-6700K. You might use a manual, constant VCore. No SpeedStep or C States, both disabled? - Yes.

>>Is that right, and what you used when you noticed your 950 Pro RAID 0 performance problem, after waking from Sleep? - No. To duplicate the OP's bug I F9 default BIOS settings, rebooted, set up UEFI RAID and remap,...no o/c, no other changes and tested sleep. From wake,...half speed :(

>>However, the cpu on the other hand drops the core voltage to >.6xx and the multiplier is at 8x. If I run Crystaldisk the core increases to 1.3xx and the multiplier varies between 42 and 8 and everything in between as it should.

Does the above happen after waking from Sleep? You loose your OC to 4.6, and the CPU is back to stock speeds, Turbo up to 4.2, and SpeedStep seems to be enabled, since you are down to 800MHz, depending on CPU load. - No,...everything BIOS defaults, no o/c.

>>What is a little disturbing is that it never throttles up!? Stays at half speed as if has not fully awakened.

Is this referring to the 950 Pro RAID array? Or your CPU speed? I think you mean the RAID array. - 950 Pro RAID array same as the OP,...after extended desktop use you would think everything would be fully awake and drives at full speed but no :(

>>Does this really tell us if it is the IRST driver? It could be, but I'm not ready to say that yet. - There is one way to find out,...use the default Microsoft raid driver, but we can't because we load the floppy IRST during install to get the m.2s recognized in RAID :( As your single drive is working full speed from wake I'm going to point my finger at the ASRock BIOS implementation of m.2 because the z170 chip set support for m.2 cannot be at fault. LOL :)

P.S. Like you I am also using the default 4,4GHz built in o/c with XMP for 24/7 daily use instead of 4.6GHz because I really can't tell the difference in gaming, I can turn my fans to silent, the cpu runs cooler and I really only benchmark to uncover issues and check to make sure everything is working as it should. All the Crystaldisk results previous posted are default C states & Speedstep enabled.

This is my result today at 4.4GHz F9 default BIOS settings with XMP, with write back cache enabled (for benching/testing only) and Link Power Management Disabled in IRST.  I have my PC plugged into a server UPS,...just to be safe.

CrystalDiskMark 5.1.2 x64 (C) 2007-2016 hiyohiyo
                           Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

   Sequential Read (Q= 32,T= 1) :  3312.495 MB/s
  Sequential Write (Q= 32,T= 1) :  1923.209 MB/s
  Random Read 4KiB (Q= 32,T= 1) :   908.839 MB/s [221884.5 IOPS]
 Random Write 4KiB (Q= 32,T= 1) :   665.851 MB/s [162561.3 IOPS]
         Sequential Read (T= 1) :  3295.103 MB/s
        Sequential Write (T= 1) :  1921.093 MB/s
   Random Read 4KiB (Q= 1,T= 1) :    53.841 MB/s [ 13144.8 IOPS]
  Random Write 4KiB (Q= 1,T= 1) :   399.944 MB/s [ 97642.6 IOPS]

  Test : 1024 MiB [C: 35.0% (150.8/430.5 GiB)] (x5)  [Interval=5 sec]
  Date : 2016/07/15 10:17:12
    OS : Windows 10 Professional [10.0 Build 10586] (x64)





Edited by clubfoot - 16 Jul 2016 at 2:19am
Back to Top
 Post Reply Post Reply Page  <1234>
  Share Topic   

Forum Jump Forum Permissions View Drop Down

Forum Software by Web Wiz Forums® version 11.06
Copyright ©2001-2018 Web Wiz Ltd.

This page was generated in 0.078 seconds.