Z170 Extreme4, m.2, etc. troubles |
Post Reply | Page <1 234 |
Author | |||||
parsec
Moderator Group Joined: 04 May 2015 Location: USA Status: Offline Points: 4996 |
Post Options
Thanks(0)
|
||||
Given my personal experience with the performance of PCIe SSDs in RAID 0, and that of others in several threads that are months old now, the results are consistently disappointing. The only thing I found that helps the performance is setting the RAID 0 stripe size to 128K. The default stripe size of 16K gives results that are barely if at all better than a single PCIe SSD. The RAID 0 scaling of performance we get with SATA SSDs is not seen with PCIe SSDs, at least not yet. The Intel IRST driver is a factor in this, or the bandwidth of the PCIe interface is simply not up to providing the performance we expect to see. If you are using the Windows RAID capability, then its performance is another thing entirely, and is normally not as good as the Intel RAID. I would like to see a screenshot of your AS SSD results, as well as information about its configuration. |
|||||
Mushu
Newbie Joined: 08 Feb 2016 Status: Offline Points: 11 |
Post Options
Thanks(0)
|
||||
I recon you use XMP profile for your memory, not sure If you tried already but make sure your DDR voltage is set to 1.35V. Also double check that the "auto" values match to those recommeded by memory manufacturer.
I would advice against to cross using bioses... doubt it is even possible to flash by "normal" means. My advice would be to try to make these 3 setting changes as I did and see if it solves your remaining stability problems. BCLK to 100.000 (not auto) PCH +1.0 Voltage up by 0.050V => 1.150V VCCIO Voltage up by 0.050V => 1.200V
Great news! Thank you for posting Christer. Hope Asrock gets the 3.0 out for rest of the ppl as well. Heres hoping it is a more general fix for a lot of these m.2 related issues.
Glad to oblige, please note though that I only have the 2 Sandisk ultra ii 960gb, Revision X35510RL in RAID0. Z170 Extreme4 got only 1 m.2 port. RAID0 configuration; stripe size 16kB, write-cache buffer flushing: enabled, Cache mode: off, GPT partition, NTFS filesystem. So here are my results with the bios 2.90 (with updated intel RST(e) efi driver: 14.8.2.2397) and intel raid driver (OS) 14.8.1.1043. (ps. results similar with default 2.90 bios) In case someone is curious here are the M.2 samsung SM951 results: (samsung driver 1.4.7.17) |
|||||
domogled
Newbie Joined: 02 Mar 2016 Status: Offline Points: 4 |
Post Options
Thanks(0)
|
||||
Thanks a lot Mushu,
indeed i use the xmp profile, actually the profile for xmp-2666 It requires ther same 1.2v, as all the other profiles either be spd(jedec) or xmp. I set the memory manually to work at 2800 but I haven't changed anything else, i left them all on auto and it results in a slight voltage increase to 1.27v, done automatically by the auto setting in bios/uefy. Shouldn't just let it manage the amount of voltage it finds to be necessary, i mean it is not a problem to rise manually the voltage to 1.35v i just thought not to wear the modules too much if they do not require a higher than 1.27v. I wrote down the rest of the settings and i'll set them accordingly. Indeed I had the fast boot set on dispabled, yet enabling it the CSM section disappears completely. If i deactivate de boot loader the system hangs again at win logon image, so it is no go for me that way. I'll be back with opinions and hopefully improvements regarding my system in the next days. thank you ! |
|||||
Mushu
Newbie Joined: 08 Feb 2016 Status: Offline Points: 11 |
Post Options
Thanks(0)
|
||||
Ok, now given that you are running outside the XMP profile, it is not guaranteed that memory performs with standard DDR4 1.2V voltage (even 1.27V might not be enough). Typically the higher the frequency the more relaxed the timings must be and also it tends to draw more power -thus a bump in voltage is often required and maybe a drop in the timings too. ps. Auto voltage setting is a very conservative in its "guess". 1.27V might be enough to boot but not enough to assure proper functionality during heavy operations. Might make sense to run memtest to verify minimum required voltage as too low memory voltage can cause miss writes and corruption of the OS.
The CSM disappears when fast boot is enabled, but the settings seem to remain inside the CSM -thus I recommend to set them before disabling the CSM and enabling fast boot. Additionally, disabling the boot manager (under tools page in bios) should not affect normal booting. The boot manager is for setting custom names and order for different boot media. Hanging with it disabled would indicate that stability problem still persists.
|
|||||
domogled
Newbie Joined: 02 Mar 2016 Status: Offline Points: 4 |
Post Options
Thanks(0)
|
||||
Thanks a lot Mushu , the settings you've recommended indeed fixed all the issues i had.
The shutdown is done properly, the hibernation the same tho takes a bit longer to enter that state, longer that I was used to while using a Z77 platform, it is just fine, yet that hibernation issues i noticed while using various platforms persists. Actually the system wakes up out of the blue, no key or mouse button pressed. anyway that's not an issue of that new platform but something OS related i guess, or perhaps network (Wake on Lan) related as I have the network adapter connected directly to the ONT (fiber optics) equipment and perhaps it just gets woken up from GPON (ONT) device. So not a problem. Then the system no longer hangs while POST-ing, no freeze while accessing windows login page. Indeed i had one freeze but it was my mistake and had nothing to do i guess with the platform. I had an old Samsung ssd pluged on SATA 3_5 and something is wrong with its MBR. It used to be part of a raid0 array long ago, then it worked as a single ssd bootable for couple of years. Any OS could be installed yet if the ssd is connected as an additional device (not booting the os from it) the system hangs and gives all kind of errors, the ssd is detected as an array having its capacity seen as double. So the freeze I've got, the old Samsung 830 is the culprit. One thing I noticed tho, not an issue as it doesn't bother me at all... I noticed some sort of double start each time i power the system. It starts then quickly shuts down, the led's from power and reset buttons remains on, it shuts itself down then starts up again and the entire POST sequence is passing quickly and then boots up into Windows. Perhaps it is related to the slight over voltage settings I use for PCH 1.0V and VCCIO. I plan to run the memtest from a hiren;s boot disc and check if any memory errors occurs wil running both modules at 1.27v for 2800Mhz. I guess the timings should be lowered while using some poor quality IC's, I have no idea which IC's used Corsair for those vengeance modules but i guess they're just fine with those primary latencies made for 2666mhz profile. I've got no bsod or hanged program so far, actually the sole reason I upped a bit the frequency was that I thought it will be more stable using the memory divider for 2800 than the one for 2666. I gave a try on yesterday for a slight overclocking from 3700 mhz i upped the frequency to 4120, and as i use the stock cooler that wes as far as i ventured. It seems the 133mhz strap and the accroding memory dividers are not so great for these ram modules, and using the 100mhz strap the dividers were not that many. It was fun tho, yet the iGPU had to be turned off.As i plan to use the iGPU as graphics signal source the cpu overclocking of the i3 6100 will stay out of discussion. For the moment i am between monitors :) i sold the 24 inches planning to go for a 40 incher or 32, so i am back using some old dell 21'CRT, i have to use dedicated graphics solely for the vga output present on discrete graphics card but not present among the mainboard's outputs, so still curious about a slight overclocking for a little longer till i will remove completely the discrete card and swap back the iGPU and no more fun that way :) Thanks for your support Mushu ! |
|||||
Mushu
Newbie Joined: 08 Feb 2016 Status: Offline Points: 11 |
Post Options
Thanks(0)
|
||||
Great! So glad that you got your system stabilized.
I cannot comment on this since I disabled hibernation from windows. I only use sleep.
Indeed, the uefi/gpt boot scheme is quite touchy and I can see it misbehaving with MBR raid disk as you discribed.
Now, I am not sure but I think my system does this same dual boot after I have reset the CMOS. I believe it is some automatic process that adjusts voltages/memory timings etc. to facilitate a successful boot. It could be MB battery related, or maybe some memory setting is not sitting well with the MB.
As long as the system is stable and you are happy with it - all is good. |
|||||
Post Reply | Page <1 234 |
Tweet
|
Forum Jump | Forum Permissions You cannot post new topics in this forum You cannot reply to topics in this forum You cannot delete your posts in this forum You cannot edit your posts in this forum You cannot create polls in this forum You cannot vote in polls in this forum |