AM4 B350/X370 black screen crash thread

gavinh87 · 9 Jun 2017 at 18:35

Xyz said:
Has anyone tried a board RMA or a different AM4 board recently?

I did some testing last night and have some weird settings 3.7ghz and 3200 ram. 5 hour memtest to 2000% coverage 0 errors. 20 back to back runs of cinebench and instant crashes occt. Still more tweaking to do...

Performance bias was causing crashes for me. Plus avx is really hard on a CPU you may need a little more voltage, what are you on now?

Xyz · 9 Jun 2017 at 18:42

gavinh87 said:
Performance bias was causing crashes for me. Plus avx is really hard on a CPU you may need a little more voltage, what are you on now?

Auto performance bias, I've never changed it.

My piece of paper is at home, but i think vcore was 1.17, LLC2, and vdsoc 0.95. The weirdest thing that I got more cinebench runs as I decreased vcore and vdsoc. but then if occt needs more voltage, it seems it would be a trade off of passing one test over the other.

gavinh87 · 9 Jun 2017 at 18:45

Xyz said:
Auto performance bias, I've never changed it.

My piece of paper is at home, but i think vcore was 1.17, LLC2, and vdsoc 0.95. The weirdest thing that I got more cinebench runs as I decreased vcore and vdsoc. but then if occt needs more voltage, it seems it would be a trade off of passing one test over the other.

That voltage is quite low tbh. It is also advised to not use any LLC due to overshoot. That 1.17 could well be spiking to 1.3 when coming off a high load.
My 3.8 chip is at 1.35v and 3.9 is at 1.4v these are the lowest I can use to pass OCCT.

Xyz · 9 Jun 2017 at 18:48

gavinh87 said:
That voltage is quite low tbh. It is also advised to not use any LLC due to overshoot. That 1.17 could well be spiking to 1.3 when coming off a high load.
My 3.8 chip is at 1.35v and 3.9 is at 1.4v these are the lowest I can use to pass OCCT.

And here I was thinking LLC was supposed to give a more consistent voltage, ugh. Tomorrow is when I'll have time to really sit down and grind something out. Will look into your advice for voltage.

gavinh87 · 9 Jun 2017 at 19:11

Xyz said:
And here I was thinking LLC was supposed to give a more consistent voltage, ugh. Tomorrow is when I'll have time to really sit down and grind something out. Will look into your advice for voltage.

It will be more consistent but only from a software point of view. There will be spikes that a multimeter and software cannot detect. It is advised to use more voltage than needed to account for vdroop. Hence my 1.35v

eyetrip · 9 Jun 2017 at 19:58

gavinh87 said:
It does sound like RAM instability mate. I know you've had it running at 3000 and it might have tested ok at first but I'm sure you aware they need to be left overnight tbh.
I can do 3600 and its gaming/bench stable but it will not pass memory tests.
Give this a go, get your final mem speed and then increase your CPU. Many users are finding that their CPU cannot do both a max OC's at the same time.

https://www.techpowerup.com/memtest64/

Thing is, had it passing hours and hours of memory tests before.

stressing memory never caused black screens before.

The goal posts jsut keep shifting with this system

I don't do two minutes of stressing and call it stable

And now it gives same crap with cpu stock and ram at lower speeds.

it's just a ******* mess.

gavinh87 · 9 Jun 2017 at 20:01

eyetrip said:
Thing is, had it passing hours and hours of memory tests before.

stressing memory never caused black screens before.

The goal posts jsut keep shifting with this system

I don't do two minutes of stressing and call it stable

And now it gives same crap with cpu stock and ram at lower speeds.

it's just a ******* mess.

Are you doing the old dos based memtest?
I heard that it doesn't test at the correct speeds or something similar.
Use that tool I linked to, it finds errors quite quickly. I have had memory cause black screens, my hynix stuff would black screen at anything over 2400.

eyetrip · 9 Jun 2017 at 20:51

I am using that tool

gavinh87 · 9 Jun 2017 at 21:08

eyetrip said:
I am using that tool

I'm struggling with what to suggest other than go from stock and work your way up but you have already done that.
Try playing with secondary timings or the performance bias options?

eyetrip · 9 Jun 2017 at 21:21

I'm just back to testing everything from stock. Again.

******* ****** system

gavinh87 · 9 Jun 2017 at 22:00

I'd suggest once you have found your max, just dialling it back a touch.

eyetrip · 9 Jun 2017 at 23:56

gavinh87 said:
I'd suggest once you have found your max, just dialling it back a touch.

Don't see that as even an option. It's not an oc problem I am having now.

It's black screening at stock now running a ram test. That's new.

Going to flash back to 1201 tomorrow.

gavinh87 · 10 Jun 2017 at 09:10

eyetrip said:
Don't see that as even an option. It's not an oc problem I am having now.

It's black screening at stock now running a ram test. That's new.

Going to flash back to 1201 tomorrow.

Is it black screening with the ram at 2133? For me the crashing is either CPU ram or that performance bias.

eyetrip · 10 Jun 2017 at 09:57

Think it's down to the gpu software again - which only used to be a problem with occt.

Though I did re-seat gpu and ram as well.

God, why can't this thing just be consistent?

currently at 45 min in tpu memtest

gavinh87 · 10 Jun 2017 at 10:07

eyetrip said:
Think it's down to the gpu software again - which only used to be a problem with occt.

Though I did re-seat gpu and ram as well.

God, why can't this thing just be consistent?

currently at 45 min in tpu memtest

You were using the palit software?

eyetrip · 10 Jun 2017 at 11:24

gavinh87 said:
You were using the palit software?

yeah. am just leaving all gpu software off now.

Can live with it boosting to 1936.

Just baffled more people aren't having this issue...

am also baffled as to why it was only an issue with occt before, but now when stressing ram too.

gavinh87 · 10 Jun 2017 at 11:34

eyetrip said:
yeah. am just leaving all gpu software off now.

Can live with it boosting to 1936.

Just baffled more people aren't having this issue...

am also baffled as to why it was only an issue with occt before, but now when stressing ram too.

I'm not using any GPU overclocking software so I cannot comment. It may just be that OCCT just found it earlier.
I see posts like I can run X but cannot run Y. There is a reason it cannot run Y and one day it will start to affect your ability to run X.

I also smile when I see prebinned CPU's with small print at the bottom *not tested with avx* There is a reason for this, they know it will not do it.

*this is not aimed at you

*

eyetrip · 10 Jun 2017 at 12:10

gavinh87 said:
I'm not using any GPU overclocking software so I cannot comment. It may just be that OCCT just found it earlier.
I see posts like I can run X but cannot run Y. There is a reason it cannot run Y and one day it will start to affect your ability to run X.

I also smile when I see prebinned CPU's with small print at the bottom *not tested with avx* There is a reason for this, they know it will not do it.

*this is not aimed at you *

Well, i use handbrake, which I believe uses avx, so it's worth me testing it.

Still, I can pass occt linpack avx, and then have handbrake crash encoding overnight, so even that isn't a fool proof test.

gavinh87 · 10 Jun 2017 at 12:30

eyetrip said:
Well, i use handbrake, which I believe uses avx, so it's worth me testing it.

Still, I can pass occt linpack avx, and then have handbrake crash encoding overnight, so even that isn't a fool proof test.

How long was OCCT running for compared to handbrake?

kekkomatic · 10 Jun 2017 at 19:14

Hi, this is my first post here, and probably not the right place for me since I never overclock anything, ahah! But the recent developments in this thread compelled me to talk about it, specifically to eyetrip.

I think you’re not alone in this at all, our builds are experiencing the same symptoms: black screen under W10 that sometimes recovers and sometimes doesn’t, keeping the machine in a state of “pre-POST”, with fans high and LEDs, keyboard and mouse off. Sometimes it even bluescreens with “VIDEO SCHEDULER INTERNAL ERROR” or “DCP WATCHDOG VIOLATION”.
And all of this happens extremely erratically, every log in Event Viewer seems to indicate video driver problems. Right? It can happen before login, right after login, opening or closing any program that requires hardware acceleration (and that means every program, since Desktop Window Manager is powered by directx), or just before or after a game has loaded. And every stress test and benchmark can either go absolutely fine or fail in the first seconds. And it can happen at stock settings, nothing overclocked, right?

I will spare you all the details, I'd talk for hours. I kept RAM at default settings, downclocked it to 1800MHz, used only one stick instead of two (btw, the motherboard needs to have CMOS cleared after changing or reseating RAM or it won’t boot, it’s like I’m back in the early 90s), replaced the GPU, the SSD, tested another power supply, and replaced the motherboard. Reinstalling Windows, changing power plan, disabling game mode, even installing the monitor drivers didn’t work. Nothing worked.

Another thing I can’t explain is the time strings in the minidump shown by Bluescreenview. Either Microsoft or AMD added “time travel” as an undocumented feature, or something is terribly wrong.

I really want to believe this is just a strange incompatibility, or a very unfortunate power-related issue, but I don't get why the computer works under stress for hours then crash while idle. Or crash immediately then work fine for 10 more minutes. Or work fine with simple programs then crash when I open a game. Or any combination of the above. I’m really starting to suspect the CPU was defective in the first place, but sending it back to Amazon will be a real pickle if that is the case (nevermind shipping it 2 weeks late despite being one of the first to preorder back in February). I don’t have money to test another set of RAM, but I spent almost 30 hours with Memtest86+ (the fancy shiny Memtest86 crashes with this motherboard) and Memtest64 with no errors. Will AGESA 1.0.0.6 fix this issue if it’s RAM related? Only time will tell, but at this point the machine I have is utterly unusable. Always has been.

CPU + MB: Ryzen 7 1700 + Asus Prime B350 Plus (BIOS 0613 as of June 10, 2017)
RAM: Corsair Vengeance LPX CMK16GX4M2A2400C14 (2x8GB, 2400 MHz - A “mystery meat” RAM, but listed in the QVL)
PSU: EVGA Supernova 650 G2 and Corsair CX600M (green version)
GPU: Asus STRIX-GTX960-DC2OC-4GD5 (GTX960 4GB), later replaced by a Sapphire Nitro+ RX 580 Limited Edition
Storage: Samsung 830 MZ-7PC128B/WW, later replaced by a Crucial MX200 250GB

Here’s another post of mine on Reddit, which tells more or less what I said above, if you’re into long-winded rants.