Bad Memory on Z170/Skylake build?

Your problem isnt primarily voltage.
A small bump above 1.2V may help once you sort out whatever it is, then again it might be fine without it.
Try relaxing some timings.
 
Nenu said:
Your problem isnt primarily voltage.
A small bump above 1.2V may help once you sort out whatever it is, then again it might be fine without it.
Try relaxing some timings.
Click to expand...

Ok thank you, it helps for RAM overclocking. but unfortunately it has no impact on memtest86 passing in my situation.
and very strange but the only one test couldn't be passed from memtest86 package, so very fast to check -run this 1 test and see the errors if memory speed is higher than 2500 (during memory testing I used 42x 100 as CPU speed).

I've already tried to increase the following voltage:
RAM 1,2 ..1,37 - no huge improvement in stability
after I increased:
VCCIO to 1.175v.
VCCSA to 1.225v.
one of my displays, which connected via displayport was blank, but other via hdmi -worked, I returned to Auto for VCCIO, VCCSAA and both displays again work.
so I think adding a separate videocard may allow to increase this mentioned voltage.

but for me - more important problem - why system (timing, memory bandwidth and some CPU tests ) become more slowly after 2 months of PC using? relaxing timing and increasing voltage unfortunately can't improve this.
for me changing in timing from 55ns to 66 ns with same speed and parameters-is a huge difference, I repeeted tests multiple times.

I excluded all software reasons (by using the Ghost image of first day after windows installation), hardware is completely the same as 2 months ago, I tried 3 versions of BIOS, cleared CMOS, changed the order of RAM modules. But it seems as start of hardware degradation or may be changing in some hidden parameters in BIOS.

P.S. the temperature of CPU is completely controlled by Noctua 15, and in November I carefully scalped (I esperienced with the procedure) the CPU and put liquid metal as a thermointerfase under plate.
With full workload under Prime - the temperature is up to 80Celcius, the maximum CPU voltage was up to 1,42 (Offset mode with - 0.01).
but usually (90% of PC usage) the CPUtemperature is 30-35Celcius and CPUvoltage up to 1,25.
I use a good case (Nanoxia) with 4 coolers at minimum speed 700rpm, which powered by PSU Corsair 850w (gold) with very stable voltage, the system connected to UPS.
 
Perhaps the problem is CPU stability.
If you are overclocked, lower it a bit and see what happens.
 
Nenu said:
Perhaps the problem is CPU stability.
If you are overclocked, lower it a bit and see what happens.
Click to expand...
thank you
If you mean CPU stability for memtest passing - I used standard for this CPU - 4200 MHz during testing,

Main question, why system (timing, memory bandwidth and some CPU tests ) become more slowly after 2 months of PC using? I excluded all software reasons via recovery from Ghost image.
and in case of decreasing frequency - test results even more lower.
 
I'm having similar issues with my G1 Gaming Z170 and 3000 V G.skill and I'm leaning on the mb being an issue. Stock, no problem, but xmp just doesn't work even after trying several bios version. I sold a RVE I bought extra and I really regret not keeping it. Thinking of switching to asus or msi doesn't do much as I read these issues are present with them too. So I'm may just get X99 again (WS or God)... Has the op updated his post rma/return experience?
 
I'd like to add on to the OP's experience about memory instability and the symptoms of it on my machine. I will also post my fix. I hope my post may help someone in their troubleshooting.

I bought an i7 6700K CPU and paired it with a Gigabyte Z170X UD5-TH (thunderbolt board).

My initial RAM was a G.Skill Ripjaw V DDR4 3200 single stick 16GB, part number: F4-3200C16S-16GVK. I have a 1TB 7200 Seagate Barracuda drive, a Samsung 830 SATA SSD, a new Samsung 950 Pro m.2 SSD.

TL;DR My fix was to use XMP settings to put my RAM to 3200MHz clock speed at 1.35v. I was not able to run my system stable at 2133MHz, 1.2v.

With the only peripheral connected a Logitech k830 AIO keyboard, I updated my BIOS to version F5, set my BIOS to default settings and attempted to install Windows 10 on my m.2 SSD.

Whilst installing, I noticed that the screen would repeatedly "Just a moment..." while I'm at the account creation screen trying to type in stuff into the text fields. Every time it "Just a moment"'s all my filled in fields would get wiped and I'd have to do it again. This would happen 3 or 4 more times before the machine would reboot and I'd have to do the Windows first set up all over again (Privacy settings and all that)

I would eventually get lucky, create an account and be able to boot into windows 10 but the symptoms only got worse.

1.) Windows search would not work, I'd type in a few letters and the search would just close itself.

2.) Whilst plugged into the onboard audio and graphics (with all drivers installed) I would get graphical corruption issues like text on Right click menus suddenly becoming tiny or weird triangular shadows showing up in the background when I try to view the calendar. At times, Windows Explorer windows that I open would randomly close. I can also hear the audio glitching like not playing sounds fully or audio sounding crackly.

3) Whilst using only a single HDMI cable to plug into a discrete graphics card (Gigabyte HD 7970 1GHz) with latest Radeon drivers 16.15.2211 installed, I will still see the graphical glitching and hear audio crackling.

3.) Whilst doing absolutely nothing, the system will BSOD after 5 or more minutes with all kinds of errors like IRQ_NOT_LESS_OR_EQUAL or CRITICAL_SERVICE_FAILED. The BSODs will usually happen with the same few error messages.

4.) When i tried to use MS Edge, the browser would load up, including the news tiles on the first open browser tab...and they'd all disappear after 2 seconds. If I try to search for something using Edge, a page will appear briefly, and then I'll get an error about the page not being able to load. I will have no way to go back to the last page and Reloading the page will do nothing.

I even tried installing Windows 8.1 and the same symptoms could be observed. I even tried installing an older version of Win 10 (before the 1151 update) and it didn't work. I tried on all 3 drives and all gave the same issues.

I ran memtest using a USB boot drive on my single stick of RAM and it ran without errors for an 8 hour test. Regardless, I decided to buy another pair of G.Skill Ripjaw Vs DDR4 3200 2x8GB sticks, part number: F4-3200C16D-16GVK to test.

Upon installing the twin sticks running at 2133MHz at 1.2v and attempting a re-install of Windows 10 into the SATA SSD, I encountered the same issues during install and after first boot.

I read this thread and noticed the OP increasing the voltage on his RAM and even activating XMP (to no success).

I decided to activate the XMP profile for these twin sticks of RAM. The system would remain in a black screen state, restart and advised me of a RAM overclock failed, I would need to enter Setup and change the settings or Restart. I chose not to change anything and just restarted. This time the system was able to boot into Windows and all the earlier symptoms were gone.

I downloaded Prime95 and ran a stress test for 3 hours. There were no errors and no workers stopped during that time. I was able to watch videos, listen to music, browse the web, use Windows search with no problem. I installed a sound card with drivers and installed a game, played for 30 minutes with no issues.

I've restarted my machine several times, cloned the SATA SSD OS drive into the m.2 SSD and everything is still going fine. I've even tried swapping my 2 sticks for the single stick 16GB and was able to use the XMP profile to achieve the same 3200MHz at 1.35v and the system remained stable. I boot into windows without seeing any messages about RAM overclock failed.

It seems my instability issues were cured by
1) Increasing the RAM voltage
2) Increasing the RAM clock speed
3) Possibly both.
 
Maybe activating the xmp profile tweaked some setting we missed? When I activate it on my z170x gaming 7 it ups my vcore to 1.3 - and who knows what else. I am guessing for memory controller stability? hmmmm. I installed windows from USB to m.2 sata with xmp active and cpu at stock clocks, went smoothly and windows seemed solid. I am running bios f8d beta with a 6700k @ 4.6ghz now (testing that oc 1.36vcore) and 2x8gb 3200 g.skill trident.
 
I'm either going to get Gskill or Crucial leaning towards Crucial Ballistics
 
Here's the situation, have verified MSI and Asus, they can only support 16GB at 3200MHz, 4x4GB. For 32GB the clock has to be dropped, why these companies do not mention that there are limits to this usage in real-world environments is beyond annoying. Had Vengence LPX 4x8GB 3000MHz but the MSI motherboard only allowed 2666MHz so clocked it at this while waiting for the 3200MHz, everything was stable. When then moving to 3200MHz it all became very unstable, so either have to downgrade to 16GB or run at a lower clock. Very stupid problem that the manufacturers should have resolved a long time ago.
 
limits to this usage in real-world environments is beyond annoying
Click to expand...

Remember all of these frequencies are overclocking.

Very stupid problem that the manufacturers should have resolved a long time ago.
Click to expand...

Intel may play the biggest part of that.
 
Stick to the QVL. I'm running the 3600Mhz kit of G.Skill with Xmp, multi, vcore and LLC being my only other adjustments. LLC was a big one, huge vdroop measured without it.
 
You must log in or register to reply here.
Back
Top