[SOLVED] Crashes - MemTest86 Errors, Dumps Suggest Drivers

[oakside]

New build seems mostly fine, but now had a couple random crashes for a couple straight days. Decided to test RAM, resulted in errors. Maybe some wisdom can be shared with me to help figure this out. Otherwise, this post may at least help others with similar problems. Will update until resolved (feel free to PM or notify me if I forget).

Details:
Crashes result in black screen for few seconds, then straight to BIOS and restarts (no BSOD displayed). Luckily I do still receive a crash dump, points to graphics drivers ("VIDEO TDR FAILURE" in both, have now prolonged TDR delay to 6 sec in registry, maybe it can now recover). Crashes occurred during Firefox (possibly YouTube) and video encoding (x265 via StaxRip). AIDA64 was also running both times (it would be so nice if all this was just from little AIDA here, since it does poke and prod in many areas).

Decided to test and adjust RAM first, since OS and drivers are all fresh and updated. MemTest86 results in errors, but extremely consistent and pinpointed only in one bit or area. (Error Address always 33A2EDEF4 at 13218 MB. Bits in Error and Max Contiguous Errors is always 1. CPU is always number 11.) Partial test on different motherboard (same model, before RMA) gave very similar results, exact same other parts. Windows Memory Diagnostic gave no errors. Other stress testing software has not yet given crashes or errors (but needs more time). Have run tests with different RAM speeds with similar results (default 2133 mhz at 1.20v, XML 3200 at 1.35, XML underclock 2800 at 1.35). Nothing else is OC, all stock defaults.

Questions:
Why is the RAM giving such a pinpointed error? Would you RMA? Or perhaps this is just a driver issue as crash dump suggests? Or AIDA64 is poking in the wrong places? Where should I go from here?

Parts:
ASUS ROG Strix B450-F Gaming (MB) (latest 2501 BIOS)
AMD Ryzen 5 2600 (CPU)
Corsair Vengeance LPX DDR4 3200 16GB (RAM) (CMK16GX4M2B3200C16) (on AMD and ASUS QVL)
XFX Radeon RX 470 RS 4GB (GPU)
Corsair SF450M 450W (PSU)
Windows 10 Pro 1903 (OS) (fresh on 2019-07-20)

Photos:
Crash Dump
MemTest86 Results

Edits:
cleanup, OCD
added photos

 

[jimmysmitty]

Since it is two 8GB sticks, the best route to take is to test each stick individually. Put one in and run the memory test. If one tests bad and the other doesn't then we know you have a bad stick. If both test bad it might be the motherboard. If neither test bad then try the next slot with one and test in each slot just to be sure.

The most logical answer is you happened to have gotten memory with a bad chip. It can happen, although its very uncommon. I have used nothing but Corsair and when I did builds for a show we ran into maybe a handful of faulty sticks. However we used thousands a year so our chances were much higher than most.

 

[oakside]

Thanks jimmysmitty. I'm really hoping it's not the motherboard or CPU. (Just had to RMA mobo due to a bunch of oddities.) It the RAM had more than an error in one spot like this, I'd just RMA. Or if the crash dumps pointed to RAM. Will remove a stick and run tests, good idea to see how errors change as I move them around.

 

[jimmysmitty]

CPU I would doubt. In my lifetime I have seen 2 dead CPUs out of hundreds.

I would still go for RAM first then board.

 

[oakside]

Haven't had any crashes for a couple days. So I have that going for me. "It's been [02] days since the last accident." Applause.

Encoded a few things fine. A few RAM error tests and other stresses. (No games yet, but soon, that'll be fun.) Changes: AIDA64 has been off a lot more. Turned off a few settings in BIOS (temporarily I hope: auto-oc type of things, performance enhancer, boost). RAM is at defaults (2133 mhz at 1.2 v). If it crashes like this, as stock default as I can get it, then I know something is very wrong.

I've not moved around my sticks, because honestly, I don't want to be digging in there any more than necessary. And there was a lot of digging lately (with RMA, esp. reseating RAM). But if there are more crashes, it's going to be necessary. Since CPU11 is always getting the MemTest86 errors, may turn off that CPU with next test and see what happens.

 

[Calvin7]

Minimum Power Supply Requirement for XFX Radeon RX 470 RS 4GB is 500 watts. Once you get that taken care of, test memory again. If you get any errors, replace under warranty.

 

[oakside]

Minimum Power Supply Requirement for XFX Radeon RX 470 RS 4GB is 500 watts. Once you get that taken care of, test memory again. If you get any errors, replace under warranty.

Interesting. I was running the RX 470 with a Corsair CS450 and higher 95W TDP processor for years (tons of games, incredibly stable). My UPS never showed a peak above 350 watts usage (usually 60 idle, 250 gaming), and that's for everything plugged in (monitor, mini amp, sub, etc). I doubt it's the problem unless my SF450 is somehow worse than the CS450 or faulty.

 

[jimmysmitty]

The SF unit is a platinum rated unit while the CX (I would guess) is a bronze rated unit. It is in no way worse than the CX450 was.

 

[oakside]

It looks like these issues really might've been from AIDA64. Which is a huge relief, all things considered.

I've had zero crashes or issues in the last couple of weeks, while AIDA usage was zero or minimal. However, recently I've had a few very concerning events with AIDA. First: Fan speeds locked on high (about 90-100%). Luckily I was sitting right there, and a quick restart fixed it. Second: All fans shut off!(At least all case and CPU fans, connect to mobo headers). Dead eerie silence, extremely alarming to hear it all of the sudden! Didn't really know what was going, decided to check fan sensors but they were missing, and temps raising! Quickly restarted, "fixed", but I'm lucky I was there to catch it (again). Unbelievable. I had no idea this was even possible by reading some sensors. I will try to figure out WTF is going on there, reset settings (esp. many sensor-related under Stability).

Update: There are surprising numbers of reports of similar issues with AIDA, as well as the original crashes. Examples: aida64 crash tdr, aida64 fans stop. For clarity: It doesn't behave this way every time, only perhaps 10-20% of the time, under certain (unknown) conditions. I'm not running any other monitoring apps at same time (and no ASUS AI Suite), besides Task Manager.

Still have no idea why I have that error with CPU11 on MemTest86. It really threw off my search. No other stress tests (or games) show any signs of errors or stability issues. I've still only has the two original crashes, and now these strange fan issues, all with AIDA running. Other relevant events, possibly helped: Memory at 1.35V and 3000 MHz. Disabled some auto-OC type options in BIOS (performance enhancer, overdrive, boost). Updated drivers (graphics, chipset direct from AMD). Increased TDR delay to 6 sec.

Photos:
AIDA64 Log - Fan Off (seems like AIDA even crashed, created 2-3 logs shortly after)