AM4 B350/X370 black screen crash thread

[eyetrip]

Yeah, but now getting black screens with cpu at stock and trying to run ram at rated settings k2933 rather than previous 3066), when just testing with memtest...

It's just ******* ******** at this point.

Was running fine for days at 3. 975ghz and ram at 3066, and the only thing that would cause black screen was occt.

I'm just tired of it.

 

[eyetrip]

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.

 

[eyetrip]

I am using that tool

 

[eyetrip]

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

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

 

[eyetrip]

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.

 

[eyetrip]

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

 

[eyetrip]

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.

 

[eyetrip]

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.

 

[eyetrip]

How long was OCCT running for compared to handbrake?

Generally ran oct for 2 hours.

Have backed down my known to pass 3.975 to 3.95 with same votlages, and upped my ram voltage to 1.4v (which the board seems to put to about 1.425v)

Passed a two hour and a three hour occt, and several hours of tpu memtest - I will test that more overnight.

Will continue to leave off all gpu software and see how things go.

 

[eyetrip]

Ha!

Installed a fan, closed my pc, decide to run tpu memtest while I play on PS4 - minutes in to test, display cuts out!

 

[eyetrip]

You've had this on 2 mobos now and 2 kits of memory on 2 gpus.
What happens if you set cpu to 3.8?

Pretty sure I remember testing down to 3.7/3.8 and experiencing same.

But then, have had it at stock too.

 

[eyetrip]

I'm not even going to troubleshoot the display cut out any more.

So long as it doesn't happen under normal usage.

There's only so much time on can spend trying to get a ******* pc to run without issue

 

[eyetrip]

As I'm investing a lot of time in lords of the fallen on PS4, and was out and about, figured I could spare the time for more testing.

I found the display cut out could happen all stock - without any gpu software!

I upped the SOC voltage and supporting phase/llc, etc, and it went away.

So, back on 9945 (that was on 1202 btw), and currently pushing just ram at 3066 with cpu at stock.

Chucked the SOC voltage up to 1.2v, and had some issues still, but relaxing timings, etc. I have managed 2 hours occct, 4 hours tpu memtest, and an hour of realbench without display cut-out.

Am going to re-enable gpu software, and see what happens out of curiosity.

Will probably run a 3 hour occt this evening and hci memtest overnight.

Am not playing with cpu speeds at all until I have spent some time with jsut ram at these speeds.

But am thinking it's just down to it requiring a high voltage on soc - normally didn't go over 1.15v

 

[eyetrip]

You'd think it would translate more in memory test failures...

Part of me suspects it may still be nvidia driver related - that is responsible for the largest latency spikes I see in latencymon. so perhaps when it hits one of those spikes when it's under heavy load, it then takes too long and it 'falls over'

so far have done a number of tests, including up to 3 hour occt, 1 hr realbench, and overnight hci memtest , and now started off another tpu memtest, and not getting the display cut-out. This is with gpu software installed.

Not sure how safe it is running soc voltage that high though... but even 2400 required over 1.1V

 

[eyetrip]

Ah, no. The board's had it's day of behaving... tpu memtest now crashing, despite yesterday running for hours, and hci memtest passing many hours.

Crashed once, and though fine, will drop to 2933, and crashes still, just differently.

But at least the display isn't cutting out

 

[eyetrip]

Give the new bios a go. Was able to bring 3466 CL16 down to 14.
3600 will boot but not stable.

I am. Same thing with display cutting out.

But!

I ddu'd the nvidia drivers and have left it without any drivers, and haven't seen the issue since.

Now, not saying it's just nvidia drivers to blame. Think it's a case of the latency spikes I saw from nvidia drivers, combined with running tpu memtest or occt, which cause the system to pause for periods.

Still early to say with certainty though.

If this is the case, I can live with it, knowing what the issue is, as it only happens stress testing (I think)

Edit: but then I have been here before where thought I had got to the bottom of it, only for it to kick up again in a day or two.

 

[eyetrip]

Still no blackout, with chopping and changing stress tests.

Was actually testing with the ram at 3200mhz. Couldn't believe I could boot at that, let alone be stable enough for hours of testing!

In the end, memtest did throw up an issue, but still impressed.

Have dropped back to 3066 and letting that run memtest overnight

 

[eyetrip]

Can you adjust teh SOC votlage on your board?

If so, set it to 1.175v

Edit: never mind, if I took time to look at photos, I would see you can, so yeah, set it to 1.175v

It may be high, but want to be sure to give it enough.

That was the primary cause, especially combined with gpu software like afterburner

Edit 2:
oh, and drop your ram to 2400 for now while testing the soc voltage.

 

[eyetrip]

Alright, tried it out and it's running fine so far but it's hard to tell if it'll stay that way.
Should I avoid running HWMonitor or are the issues only coming up if you run multiple monitoring software at simultaneously?

Just disable all monitoring software for now.

Just try it for a while with the increased soc and lowered ram

 

[eyetrip]

Here's an image of my Ryzen Balanced energy plan.
The minimal performance of processor @ 80% seems a bit off, could you guys compare it with yours?
Thanks for the help by the way.

Oh, use high performance profile as well.

Have suspected that Ryzen Power plan of causing issues.

I just use high performance and set min processor state to 60%