ZoranC
 
I'm sorry you are having hard time.
 
I'm not sure it's the mb. I'm running X299 Dark and I have observed discrepancies in voltage readouts but some variations are expected and readouts depended on which of the sensors were used to read them out (look for my posts on that topic). In the end I figured readouts don't matter as long as system is stable. It would help if you would share more info.
 
Intel's processor tests are not very thorough so I wouldn't say they are 100% guarantee CPU is OK. However if I had to suspect something it would be your memory. I use MemTest86 Pro and GSAT for memory testing. My G.Skill is 3200/16. Your 3600/16 might be too much. Have you tried using OCCT to stress test your CPU and memory?

Sajin
Sounds like you’re good then.

Sajin
Lol. You weren’t suppose to tell them that.

arestavo

Sajin
Sounds like you’re good then.

Intel is saying that enabling RAM XMP profiles above the CPU rating voids the warranty? The 10940X was only rated for 2933....
 
Guess I'll try RMAing the X299 FTW K with EVGA.

ZoranC
 
First, IME Intel's XTU doesn't stress system nowhere as much as OCCT and P95 do. So if I were you I wouldn't waste time on XTU, I would spend that time on OCCT's "power supply" test.
 
Second, you are saying exactly what I have been saying for quite a while now in my threads on my X299 Dark build, that constant load tests of single component are not the ultimate way to confirm final system stability, that best way is random variable loads simultaneously across multiple components.
 
Third, in other words, you are talking about same thing I have been talking about before, that I had issues with stability when I was allowing system to vary its power. That is why I locked in all cores to same multiplier, disabled turbo boost 3.0, disabled C states and rest of the stuff I mentioned above and am running Ultimate power plan so power fluctuations are minimal.
 
ALSO, I have configured driver for my 2080 to prefer maximum performance to minimize spike related instabilities with it too.
 
P.S. Is your CSM enabled or disabled? Is your above 4G decoding enabled or disabled? Are you trying to use ReBAR?

ZoranC
 
You might want to try with disabled above 4G and ReBAR. Few times I had odd behaviors they were enabled. I can't claim with 100% certainty they were the culprit because issue was very sporadic but ever since then I keep them disabled and issue hasn't happen repeated itself. Which is fine with me because I couldn't see any real world benefit when they were enabled.

Sajin
Lol.

ZoranC
Didn't you say you tested with XTU for 4 hours straight without any issues??? What has changed?
 
If I were you I would go back to basics: BIOS settings at default multipliers and values for voltages (no overclocking), AVX offset 3/5, and test that memory with XMP on extensively.

ZoranC

arestavo

ZoranC
Didn't you say you tested with XTU for 4 hours straight without any issues??? What has changed?

Oh no, that was several hours of testing OCCT and AIDA64.

This...

arestavo
One interesting note, as I sit here with Intel's XTU stress testing AVX (because Star Citizen uses AVX) for 4 hours straight with no issues ...

... is why I thought you said you already tested with XTU without issues.
 

arestavo
And yes, that was 100% stock settings with RAM speed set to 2933 where I got that BSOD in under a minute with Intel's XTU stress test.

 Manually setting RAM speed is not 100% stock setting. When I had problems with Corsair memory I tried manually setting memory speed and that resulted in even more frequent problems. Clear your CMOS, configure following:
 
{Boot}
Fast Boot : On -> Off
 
{CPU}
CPU Multiplier Control : Auto -> Manual - PerCore
Per Core OC Ratio : X -> 43 (CPU dependent, all core boost for i9-10900X/920X is 4.3GHz)
Core 1 to ‘n’ Ratio : X -> 43 for i9-10900X/920X (CPU dependent)
Mesh ratio: Auto -> 24
AVX2 Negative Offset : X -> 3 for i9-10900X/920X (CPU dependent)
AVX3 Negative Offset : X -> 5 for i9-10900X/920X (CPU dependent)
 
{Memory}
Memory Profiles : Automatic -> XMP Profile 1
Force Memory Retraining : Disabled -> Enabled
 
{CPU Configuration}
EIST : Enabled
EIST>Turbo Mode : Enabled
CPU C states : Auto -> Disabled (Enhanced C1 disabled)
MSR Lock Control : Enabled
Hyper-Threading : Enabled
Virtualization Technology : Enabled
Intel Virtualization Technology For Directed I/O : Disabled -> Enabled
CPU TjMax : Auto -> 86C (CPU dependent)
Intel Turbo Boost 3.0 Driver Support : Enabled -> Disabled
 
{PCIe Configuration}
Above 4G Decoding : Disabled
Re-Size BAR Support : Disabled
PE1-PE6 Speed : Auto -> Gen3
 
Leave everything else at -default- after CMOS clearing.
 
Test memory using PassMark's MemTest86 -PRO- booting from -USB flash drive-. Boot from flash is the only way to thorougly test practically all memory. Free version doesn't include "hammer" test and it was hammer test that revealed issues with my Corsair memory. Also do at least 4 passes, ideally 8 or more. At default settings first pass doesn't run everything, and errors sometimes wouldn't pop up until second or third pass.
 
Only once memory got clean bill of health at these settings move onto Windows based tests. Put Windows in Ultimate performance plan, and modify plan:
USB settings > USB selective suspend setting: Enabled -> Disabled
PCI Express > Link state power management: Off
 
Put Nvidia driver in prefer maximum performance and read up on debug mode. Consider putting your 3090 BIOS switch in different position.
 
Repeat your XTU tests testing one instruction set at the time. Start with regular moving onto AVX2 and then AVX3.
 
And consider trying out G.Skill memory model that has been certified for your mb.