New computer - LinX instability

przemo1109

Junior Member
Jan 22, 2016
7
1
36
Hello, I was wondering if You could help me with my problem.

Problem: New computer (made by me), is restarting or getting BSOD Machine Check Exception after 20-40 minutes testing stability by LinX.

Workaround: Changed default RAM frequency from 2666 to 2133 Mhz

Spec:
i5 9600k (not overclocked)
gigabyte z390 gaming x (default settings, bios version upgraded to F5)
be quiet! Pure Power L8 CM 630W
adata 16gb (2*8) 2666mhz xpg z1 red cl16, 1.2V
Gtx 1070

Some tests:

- LinX with 2666Mhz ADATA memory: computer gets BSOD after 30-50minutes
- LinX with 2133MhZ ADATA memory: 3 hours test completed without error
- Gaming session Battlefield V multiplayer: 2 hours no problem with ADATA 2666Mhz/2133Mhz
- MEMTEST86 with 2666Mhz ADATA memory: 3 hours test completed without error
- Windows memory diagnostic with 2666Mhz ADATA memory: completed without error
- Intel Processor Diagnostic Tool with 2666Mhz ADATA memory: completed without error

I borrowed from a friend other memory: 2 x 8 GB patriot white led ddr4 3200Mhz and I have other results:

- LinX with 3200Mhz PATRIOT memory: LinX test stopped after 30-50minutes, no BSOD but LinX error appear
- LinX with changed PATRIOT 3200Mhz to 2666Mhz: LinX test stopped after 30-50minutes, no BSOD but LinX error appear
- LinX with changed PATRIOT 3200Mhz to 2400Mhz: LinX test stopped after 30-50minutes, no BSOD but LinX error appear
- LinX with changed PATRIOT 3200Mhz to 2133Mhz: 3 hours test completed without error

Additional: problem appear in F4 bios version and the newsest one F5 that should improved RAM stability :p

Some voltages and temp:
Motherboard: https://imgur.com/a/wCxrgpE
CPU: https://imgur.com/a/O9FvdkQ
Disk and GPU: https://imgur.com/a/flPzqmA

Conclusion: as you can see (cpu - voltages and temp), CPU power package is much smaller in gaming than in LinX (61.50W vs 91.65W) and also when I am using 2133Mhz is smaller than with 2666Mhz (89.89W vs 91.65W). Maybe here is a problem. I will add that my motherboard has 2 power sections for CPU:

1) ATX 12V 8pin
2) ATX 12V 4pin

Motherboard power sections photo: https://imgur.com/a/I9J03nK

Because my PSU has only ATX 12V 8pin my motherboard does not use additional ATX 12V 4pin. I had read that this additional ATX 12V 4pin is for extreme CPU overclocking I did not do it. What you think? Maybe my CPU needs this additional ATX 12V 4pin durning LinX test?
 

coercitiv

Diamond Member
Jan 24, 2014
6,187
11,859
136
What are your current VCCIO and VCCSA voltage values? You can find the displayed in HWInfo64 sensor window, motherboard section (togheter with temp sensors and fan speed values). VCCIO may be named CPU I/O.

These values influence memory controller performance and they usually go up when enabling XMP. Even if their values are normal (VCCIO around 0.95V, VCCSA around 1.05V), you can still attempt some stability tests with both these voltages increased around 1.15V. Don't increase one without another, they need to stay somewhat close in value, so just use the same value for both.

You can also increase memory voltage to 1.25-1.35V.

In your place I would increase memory voltage to 1.35V, VCCIO and VCCSA to 1.15V, then I would test the ADATA memeory at 2666Mhz. If it becomes stable with these settings, then I woudl gradually revert back voltages: first decrease memory voltage by 0.05V and see if it's table all the way back to 1.2V, then gradually decrease VCCIO and VCCSA by 0.01V at a time (togheter) until system is unstable again. Once you reach the limit, go back by 0.02V and leave it there.

Further BIOS updates may help, unless your CPU is actually the culprit by having a very weak memory controller (I doubt this though, probably the BIOS isn't mature enough yet).
 
  • Like
Reactions: przemo1109

Charlie22911

Senior member
Mar 19, 2005
614
228
116
Make sure that your board isn’t enabling multi core turbo or extending your cpu power limit by default. These are technically over clocking and it may be enough to be the source of your instability in AVX workloads.
Next I would run memtest86 for a few passes, I actually had a stick of ram spontaneously die in my server after running stock for over a year (never over clocked).
 
  • Like
Reactions: przemo1109

przemo1109

Junior Member
Jan 22, 2016
7
1
36
Hey, thank for fast answer! VCCSA right now (in dekstop without gaming/linx etc) is 1.200V, I can not find VCCIO in HWInfo64 :/
In attachment you can find my default bios settings. Enhanced Multi-Core Performance and CPU power limit is auto. How about additional ATX 12V 4pin? My PSU does not have it (only has 8pin) should i go Molex to 4Pin and then I will have 8 pin + 4 pin?
 

Attachments

  • Advanced_CPU_Core_Settings.png
    Advanced_CPU_Core_Settings.png
    333.5 KB · Views: 6
  • Advanced_Frequency_Settings.png
    Advanced_Frequency_Settings.png
    312.2 KB · Views: 5
  • Advanced_Memory_Settings.png
    Advanced_Memory_Settings.png
    310.1 KB · Views: 3
  • Advanced_Power_Settings.png
    Advanced_Power_Settings.png
    277.9 KB · Views: 3
  • Chipset_Voltage_Control.png
    Chipset_Voltage_Control.png
    274 KB · Views: 3
  • CPU_Core_Voltage_Control.png
    CPU_Core_Voltage_Control.png
    307.6 KB · Views: 5
  • DRAM_Voltage_Control.png
    DRAM_Voltage_Control.png
    286 KB · Views: 4

coercitiv

Diamond Member
Jan 24, 2014
6,187
11,859
136
Hey, thank for fast answer! VCCSA right now (in dekstop without gaming/linx etc) is 1.200V, I can not find VCCIO in HWInfo64 :/
From what I can see in your BIOS screenshots, your VCCIO is 0.95V and VCSSA is 1.05V (they are named CPU VCCIO and CPU System Agent Votlage in CPU_Core_Voltage_Control.png)

It's a bit odd to see default values in BIOS and higher values in HWInfo64, but maybe this is what happens on your board when XMP is enabled. BTW, in HWInfo64 the VCCIO voltage may be named CPU I/O)
GxcS0BU.png

Don't mind the values in my screenshot, my RAM is overlocked.

If your VCSSA is aleady 1.2V in HWInfo and CPU I/O is also above 1V, then your don't really need to do anything more there. All you can try is to increase memory voltage a bit and see if it helps. Even 1.25V instead of 1.2V might make a difference.

How about additional ATX 12V 4pin? My PSU does not have it (only has 8pin) should i go Molex to 4Pin and then I will have 8 pin + 4 pin?
You don't need to do anything with the additional 4pin connector. The connector your PSU has is definitely enough for your current CPU.

From I can tell from your BIOS screenshots your CPU is configured to run at 95W TDP and stock clocks, from a power & frequency perspective everything should be fine.
 
Last edited:

przemo1109

Junior Member
Jan 22, 2016
7
1
36
It seems that "VCCIO" in my HWINFO64 is equal to "DDR", am I right?

1543786831586.png

All you can try is to increase memory voltage a bit and see if it helps. Even 1.25V instead of 1.2V might make a difference.

I did it, I increased memory voltage to 1.30V and 1.35V I had the same result if XMP 2666Mhz = BSOD after 30-50 minutes testing in LinX.

SAD
 

Charlie22911

Senior member
Mar 19, 2005
614
228
116
Have you tried memtest86 yet?

If you get errors running that try each stick individually until you find the bad one, or if you get errors on all of them individually then we can start looking elsewhere.
 

bononos

Diamond Member
Aug 21, 2011
3,886
156
106
Is Linx considered to be a 'safe' burn in tester? I thought its simple 'power virus' like app which is not suitable to run it for a long time if at all. Prime95 is a better utility for testing.
 

DrMrLordX

Lifer
Apr 27, 2000
21,620
10,830
136
LinX is not Furmark, and calling it a "power virus" doesn't really do it any justice.

It is at least as valid as Prime95 as a benchmarking/stability testing utility.
 

przemo1109

Junior Member
Jan 22, 2016
7
1
36
Hey,

- MEMTEST86 with 2666Mhz ADATA memory: 3 hours test completed without error
- Windows memory diagnostic with 2666Mhz ADATA memory: completed without error

Strange thinks for me is that with PATRIOT 3200Mhz durning LinX i get "LinX error" with ADATA 2666MhZ i get BSOD or freeze.

The main question: should I bother?

Disabling Enhanced Multi-Core Performance did not help :/
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
Hey,

- MEMTEST86 with 2666Mhz ADATA memory: 3 hours test completed without error
- Windows memory diagnostic with 2666Mhz ADATA memory: completed without error

Strange thinks for me is that with PATRIOT 3200Mhz durning LinX i get "LinX error" with ADATA 2666MhZ i get BSOD or freeze.

The main question: should I bother?

Disabling Enhanced Multi-Core Performance did not help :/
I know its different memory, but what speed are they running in bios ? If the 3200 is running@3200, then maybe you should set it to run@3066 or 2933.
 
  • Like
Reactions: Drazick

przemo1109

Junior Member
Jan 22, 2016
7
1
36
Patriot 3200/3000/2666/2400 -> LinX Error after 30-50 minutes LinX testing
Adata 2666/2400 -> BSOD after 30-50 minutes LinX testing
Adata/Patriot 2133 -> No BSOD, no Error after 3 hours LinX testing
 

ehume

Golden Member
Nov 6, 2009
1,511
73
91
Is Linx considered to be a 'safe' burn in tester? I thought its simple 'power virus' like app which is not suitable to run it for a long time if at all. Prime95 is a better utility for testing.
LinX is the front end for Linpack. Linpack is what the engineers at Intel use to test their own chips.
 

przemo1109

Junior Member
Jan 22, 2016
7
1
36
Hey, today I plugged additional ATX 12V 4pin so for CPU i have 8 + 4 and still after 30-40 minutes LinX BSOD with 2666Mhz Ram :(
 
  • Like
Reactions: ehume