AM4 B350/X370 black screen crash thread

eyetrip · 12 Mar 2017 at 14:36

gavinh87 said:
That took a whole 4 seconds

Can you try de-selecting one of those options at a time and see if any allows you to run longer?

I would leave out the disk stress bit in geenral

Have jsut started with GPU included. Just memory left to try in a bit

Will give it about 20/30 min

eyetrip · 12 Mar 2017 at 14:45

I'm going to let it run a bit with just gpu added.

eyetrip · 12 Mar 2017 at 15:19

gavinh87 said:
Its the system memory. Everything else was left selected and it ran for 10 minutes fine. I selected the memory and within 9 seconds the CPU usage dropped from 100 to 20 and gave me the failure warning.
This memory test passed 4 hours of memtest yesterday...... Incompatibility with my board maybe?

Just did half an hour with all but disk stress enabled, no problems.

eyetrip · 12 Mar 2017 at 15:28

So, hour of prime, fine.

Few half hour runs of Aida with different things selected, fine.

Quite a few runs of RB benchmark, fine.

2 mins in RB stress, black screen.

eyetrip · 12 Mar 2017 at 16:10

gavinh87 said:
Yours is even weirder than mine. Have you tried furmark to test GPU only?

No.

Tend to only run things that I consider more indicative of a normal work load. Usually. I tended to find realbench ok as encoding is a fairly typical thing to run.

eyetrip · 12 Mar 2017 at 16:29

gavinh87 said:
I'm running out of things to suggest no eyetrip :/ Did AIDA go ok for you?
Might be worth leaving it overnight and if it passes agree to ignore realbench?

aida seems fine for the limited time runs I have given it.

eyetrip · 12 Mar 2017 at 16:37

easyrider said:
Wouldn't other people be having issues with realbench aswell?

I'm not entirely sure a whole lot of other people are running the test for long enough to know. Some are though...

eyetrip · 12 Mar 2017 at 16:38

mmj_uk said:
I'd have thought so but it might be some sort of obscure bug or compatiblity issue that only affects certain configurations, although I'd not happy with Realbench failing myself. TBH I'd just return everything if possible, wait a couple of months for the platform and try again.

Unfortunately, stupidly, I have already sold my Z97 stuff.

So now I NEED to get something...

eyetrip · 12 Mar 2017 at 16:48

Yeah, i'll do that shortly. Just installing new nvidia drivers on the off chance...

Man, I can't afford to have two stressful tiring weekends dealing with this kind of crap...

It can only be CPU or GPU related. Or Asus related, I guess.

I don't want anotehr quad core - the whole point of going this route, so that means, spending far more ******* money than I wanted if I return this.

eyetrip · 12 Mar 2017 at 16:49

gavinh87 said:
Stock is now 1 hour stable in AIDA as oppose to 4 seconds
I have literally swapped sticks around :/

that's good.

Have tried all that myself.

Tried the crucial RAM again, same ********.

eyetrip · 12 Mar 2017 at 16:51

I would just ******* ignore Realbench, but I'm worried it's indicative of an issue I will run into later in actual usage, when it's too late to return.

eyetrip · 12 Mar 2017 at 17:05

So, if I put the Gt 730 in, and it passes, does that eliminate the cpu as a culprit? Could it still be down to a controller on cpu?

I know I won't get to bottom of this without spending money, but don't know whether to try new cpu, new gpu, or a non-asus board...

eyetrip · 12 Mar 2017 at 17:14

mmj_uk said:
Not necessarily I'd use your 980Ti, the trouble with using your GT730 is that it won't push the motherboard/cpu as hard. Although ideally you need to test your 980Ti in another system too.

Have you tried underclocking your 980Ti?

not under-clocked, but have pushed voltage for it up and power limit.

eyetrip · 12 Mar 2017 at 17:27

mmj_uk said:
Try just underclocking it, I don't think you answered before if it was factory overclocked?

Trying now, but MSI afterburner doesn't give me much downward movement. - while I was typing it crashed with underclock.

It's a reference card

eyetrip · 12 Mar 2017 at 17:40

Man I neeed to stop typing these while testing as I never manage to finish before a crash.

So, I have done a total reset to stock - including removing all my upped voltages/power stuff.

While I was in there, I noticed the GPU was runnign at 8X - I know I saw it running at 16x before but can't remember when. So I re-seated.

restarted and fired up RB stress, and this time fiale dwith blue screen kmode_exception_not_handled with ntoskrnl.exe + nvlddmkm.sys flagged. Not sure if this is relevant to anything yet...

eyetrip · 12 Mar 2017 at 17:54

mmj_uk said:
So it's still unstable even with GPU underclock? did you do a fresh Windows install when you upgraded btw?

I did with the B350.

Didn't when put the x370 in, but it didn't even seem like Windows noticed - didn't even have to re-activate.

But before reading your post I had set Windows 10 to re-install.

Going to run with nothing 'extra' installed. No nvidia drivers, nothing.

eyetrip · 12 Mar 2017 at 18:34

Ergh... So can't do this absolutely clean, as the stress test won't run without opencl installed.

Only way to install opencl is through installing the nvidia drivers.

So essentially, realbench offers you no clear way of testing individual items to gain a better idea of what may be the failure point.

eyetrip · 12 Mar 2017 at 19:15

mmj_uk said:
Have you tried with NVidia drivers from a few months ago? if you can run Heaven and Prime95 then I find it hard to believe it's a hardware problem.

Would unigine Valley and prime be sufficient?

Though I suppose can just download heaven

eyetrip · 12 Mar 2017 at 19:36

Currently on 33 minutes.

Had a moment of anguish when I went in to see a black screen - thankfully it was just the screensaver!

eyetrip · 12 Mar 2017 at 19:44