PC Drivers & Software

[charliedog]

Hello I have a windows 10 system with 8 gbs of ram an AMD FX-6300 6 core cpu a 500w PCU and a AMD Radeon R7 370 Series GPU. Im on driver version 18.6.1 and my temperatures are all normal, I havent had any spikes or anything, but as soon as I put graphic load on the system, mostly happens with games, whether its a brand new beautiful game or a game from 2006 my PC crashes, i get a million lines accross my screen and it just restarts, sometimes it takes a few hours of gaming til it crashes, and sometimes just loading up the title screen will crash it. This has been happening more and more frequently as time goes on, originally it would only crash every few days, now its several times a day. It will not crash in safe mode, ive tried stressing it and putting it under load for hours and it wont crash in safe mode, so my guess is it has something to do with the drivers that are loaded in regular boot up. I appreciate all the information and help you can give me .

Message was edited by: Matt B

Title Edit

[locodicegr]

Unstable power from PSU?

[charliedog]

How would i test this?

[elstaci]

Download a free diagnostic program called OCCT. Run the test for one hour with the "Errors" box check marked. If it finishes without crashing it will open up OCCT folder at "Documents" and show you all the statistics while running the test.

If it crashes while running the test, then go to "Documents" Folder and click on the OCCT folder and then click on the first image. then hit the side arrow to go through all the test images.

While running the GPU Test you will be able to see if your PSU outputs stay within tolerance. If one doesn't, may indicate a bad PSU.

You can also check your PSU. But note, if you have a cheap or weak PSU, running the test might damage it. Normally it will shut down immediately if the PSU is not strong enough or you will see the same image as when you ran the GPU Test.

The PSU Test basically runs both the CPU and GPU Stress test at the same time putting maximum Load on the PSU.

[charliedog]

It crashed immediately, black screen followed by no input then computer restarting, didnt even get a log or picture of what happened, it went from baseline idle to crashed, that was on the GPU test, any point in testing the PSU if the GPU crashed by itself?

[elstaci]

Go to Windows "DOCUMENTS" folder. CLick on the OCCT Folder . Then click on the first image. See if you see anything wrong with the data from the test then arrow through the rest of the images showing the various data from the test.

Check to see what the PSU outputs were before it crashed. Especially the 12 vdc, 5 vdc, and 3.3 vdc.

Since it crashed using the GPU Test it will most likely crash in the PSU test since the PSU Test consists of running the GPU and CPU test at the same time.

If you are able to, install your GPU Card in another computer and see if it works correctly. If it does, that indicates you may have a bad PSU or some other hardware that is failing. If it the exact same problem occurs in another computer, then it is a good indication your GPU card may be bad.

You can always install another GPU card even if it is older and see if it crashes under load. If it doesn't and the GPU Test works correctly without crashing this will probably indicate a bad GPU Card. But if it crashes also under load , then it is a good chance your PSU is bad and not providing enough power under load.

[charliedog]

I do not have that folder, i told you it crashes directly as soon as the test begins, doesnt even get to load the test all the way, and im noticing radeon settings pops up "reset wattman settings due to unexpected failure" or something like that, is it possible that newer drivers arent compatible with my GPU? and if so how would i figure out what the last compatible driver is?

[aliraza46514651]

usefull info

[pokester]

I don't think you are alone on this. I was trying to help another to no avail yet with I think a very similar setup. Read this thread too: A little help with computer crashes

amdmatt​  might be able to offer some advice or send you guys same issue to engineering.

[charliedog]

I used the driver uninstall tool from that post, but unsure if i should install the newest driver, or if theres an older more stable version i should be using instead

[pokester]

I would try the latest then regress if that doesn't work.

[colesdav]

Hello.

A few questions.

(1). Do you have anything overclocked on your PC at all. If you do, then turn off overclocking and set everything to default clocks including your graphics card.

(2). Regarding your AMD Radeon R7 370 Series GPU with the driver version 18.6.1. How did that driver get installed on your system?

Based on your description of constant crashing this sounds very similar to previous questions I have answered and tested myself.

Quick description of the problem.

(1). If you let Windows 10 automatically detect your AMD GPU  hardware on your PC it installs what it thinks is the correct driver for your PC.
I have found that what Windows thinks is the correct driver is frequently not what it should be installing.

For example - It installed the earliest (very first)  version of AMD Crimson (not Crimson ReLive, the ~ 2 year old Crimson driver. ) driver on my PC when I had HD7970 GPU connected. Windows should have installed latest version of AMD Adrenalin 18.6.1.

(2). It is difficult to prevent Windows from automatically installing "the correct AMD Driver" at boot up unless you disable it using Local Group Policy.

(3). If you decided there was something wrong with your AMD Driver, and uninstalled it from control panel or even using DDU in safe mode, then reboot to reinstall, Windows will install "the correct version of the AMD Driver" unless Local Group Policy is used to prevent it. The User will not necessarily see that that this has been done unless they look at Control Panel Device Manager. Windows does not always install AMD Radeon Settings GUI, sometimes it just installs the Graphics Driver.

If the User then downloads and installs the AMD Adrenalin 18.6.1 driver, the AMD installer does not detect that there are already AMD Drivers installed and you end up with who knows what GPU drivers on your system.

I think you should investigate if this is what has happened in your case.

I will add a link to a previous case like this.

Note you could be facing multiple problems - for example R87 370 is getting old. It might be time to dismanle the GPU, give it a proper clean and replace the thermal paste. Also on some games, AMD Drivers simply crash when some graphics settings are used in the game versus Nvidia card runs fine on the same machine. I have found in the past that this can be due to Nvidia Gamesworks features being set in game. Recently some AA settings cause crashing with AMD Drivers.

Bye.

[charliedog]

I tried running OCCT several times but it just kept crashing before it even

loaded the test, I finally underclocked my GPU as far as i could and maxed

the power and fans and it finally allowed me to run the test, it ran for a

full hour and had no issues, I have included the data from OCCT as well as

a screen shot of my Overdrive settings

[colesdav]

RE: underclocked my GPU.

What are the GPU Specs for GPU CLK and GPU MEM CLK for your exact model of card. Please send a link to the manufacturers specs for the exact model of card you are using.
What were you running at when the GPU was crashing
What are you running with now.

Even if you managed to get through 1 hour of OCCT, please let me know how you installed your graphics driver and if Windows was automatically installing drivers or if you don't know then please say. I am searching fro the previous case I worked on where someone purchased an RX580 and it was constantly crashing.

Bye.

[colesdav]

Here is the link discussing the crashing issue in RX480 caused by Microsoft installed AMD Drivers:
GIGABYTE G1 AMD RX 480 4GB RED SCREEN
Bye.

[charliedog]

i was running it stock when it was crashing and now its not crashing im

running it at minimum speed (clock and memory) which is 300mhz/150mhz

respectively

[colesdav]

Ok

If you have to run OCCT at absolute minimum speed then there is something seriously wrong with your driver installation or your GPU.

First thing to debug is the Graphics driver installation.

1. Disable Windows Driver installation using Local Group Policy.
2. Uninstall AMD Driver in Control Panel.

3. Download DDU.
4. Download the driver installer for your card from AMD Driver Downloads page.

5. Boot into safe mode and run DDU and delete any traces of Nvidia and AMD Drivers.
6. Boot into Windows as normal.

7. Go to Control Panel - Device Manager  and check what it says for Display Adapters. It should say Microsoft Basic Display Adapter not R7 390.
8. Install the AMD Driver.
9. Reboot and test your GPU.

You can see some details of the above in one of my recent posts about Crossfire on RX580&480 Overwatch.
I will post that link.

Bye.

[charliedog]

I used the amd auto detect (64 bit) to install my drivers, windows auto

update is turned on so it may have installed what it thinks is the correct

drivers https://us.msi.com/Graphics-card/R7-370-GAMING-4G/Specification

[colesdav]

If Windows update is on then it is very possible you are getting corrupted driver installation.
Bye.

[charliedog]

going thru the proccess of disabling windows auto driver install and clean

install of amd drivers, i will get back to you once i have finished and

tested to see if it crashes when running OCCT on regular (stock) gpu speeds

[colesdav]

Hi,

Here is the link showing how to set Local Group Policy:
Hard Freeze with Crossfire Enabled
Page 7 of 8 has the detail.

I really have to go now.
I hope this helps.

Bye.

[charliedog]

Done all of that perfectly double checking the steps to make absolutely

sure and it still crashes immediately at test start of OCCT, worst part is

it crashes so soon that OCCT doesnt even get a reading of what happened, so

i dont know if its a lack of power or if its the newer drivers that cause

it, but the official specs state that the r7 370 only requires 110w of

power and my psu is a 500w. Seeing how i can only succesfully run OCCT with

minimum possible underclock would seem to suggest a power issue, but as

stated my power supply is more than enough for it. P.S the psu is the

newest part of my computer being installed shortly after the gpu about 2

years ago, and the whole system is again dust free and hardly pushed, 99%

of the time the gpus smart fan system doesnt even kick the fans on because

the gpu isnt hot enough to trigger it.

At this point its looking like a hardware failure, but what hardware? The

gpu? The psu? Motherboard? I dont even know what else to do anymore.

Thank you all for trying to help me im open to ideas that i can try still i

dont think its a hardware failure, but if the shoe fits....

[locodicegr]

PSU hardware problem i think.

[charliedog]

Is there some sort of logging software i can get that will show me what

happened to cause it to crash? It doesnt seem to be a build up but a switch

that once flipped instantly crashes the computer, so would need something

that logged which switches were going to be flipped before it happens so i

can see which one was the last to happen before crashing, sorry for the

metaphore but its the best way i can explain it

On Thu, Jul 12, 2018, 6:32 AM Charles Collins <charlie47025@gmail.com>

[colesdav]

Hi,

(1). OK, so if I understand it correctly you followed the steps to disable Windows Driver installation using Local Group Policy, uninstalled the AMD Driver using DDU in Safe Mode, and downloaded and installed the correct driver directly from AMD Drivers Download yet you still get crash in OCCT when you use Graphics Test?

(2). What version of Windows 10 are you using. Is it fully updated and clean? (By clean I mean chkdsk, chkdsk /scan sfc/scannow, DISM /online /Cleanup-Image /ScanHealth , CheckHealth, RestoreHealth results should be clean).

(3). You can use a tool called bluescreenview to check what is causing crashes: Blue screen of death (STOP error) information in dump files.

That should give you some information regarding what is actually causing the crashing.

(4). Regardijng "i get a million lines across my screen and it just restarts",
This sounds to me like Vram problem in the GPU or maybe the GPU Die is disconnecting from the PCB.
What happens if you try to run the GPU at default GPU Clock but set the GPU Mem clock to minimum?

(5). Your PSU should have enough Power. What PSU is it. Is it a Corsair 'i' series with Corsair Link?

Thanks.

[charliedog]

Do i need to run all of these checks or just one of them? just slap it into

elevated cmd? my PSU is made by coolermaster and its pretty new it has no

signs of aging whatsoever, and the cause of crash according to that program

is 0x00000124 ntoskrnl.exe ntoskrnl.exe+7d98f7

[colesdav]

(1). You should run at a minimum

chkdsk /scan
sfc/scannow
DISM /online /Cleanup-Image /ScanHealth
DISM /online /Cleanup-Image /RestoreHealth

Yes you open cmd with Administrator Privilege

Please let me know if you get any errors.
If you do not, then run Windows Update.

(2). OK I was concerned that maybe you were running with a Corsair 'i' series PSU because I know that they can start to trip out at start up and during gaming after time ... I have had to RMA them in the past.

(3). You have probably looked into this... BSOD Caused by ntoskrnl.exe - [Solved] - Windows 10

Indicates bad drivers or bad RAM or overclocking.

Thanks.

[charliedog]

sfc/scannow and memory check both came back with no errors, any ideas for

how I can test the gpu memory and see if its just a setting i need to fix

or if the memory is burned out?

On Thu, Jul 12, 2018 at 12:07 PM, Charles Collins <charlie47025@gmail.com>

[colesdav]

Hi,

(1). RE: sfc/scannow and memory check both came back with no errors.
Which memory check did you run?
The built in Windows memory check or this: MemTest86 - Official Site of the x86 Memory Testing Tool ?
I suggest you use Memtestx86 to check one RAM Stick at a time.

(2). OK... to test the GPU VRAM Memory exhaustively is difficult. 
OCCT does do a GPU memory Error Test but last time I looked it only checks the first 1GB of VRAM.
I was looking into other ways to run exhaustive GPU Vram testing last year.
Their were some programs available to do this and I did try them out.
Let me go back and check the situation later today.

Something I meant to revisit soon anyhow.

(3). To see if the 'memory is burned out' you will need to make a physical inspection of the card and then remove the fans and heatsink and any backplate to inspect the GPU PCB. Don't go there yet.

Bye.

[charliedog]

ran the built in windows mem check, and the RAM is 2 sticks of 4gb ddr4,

but as i said i got better results when reducing the gpu mem clock speed,

so i dont think its a problem with normal RAM

[charliedog]

Reduced memory clock to 300mhz and it ran occt for about 10 mins before

crashing, could this be a problem where the gpu isnt dumping the old memory

so it runs out and crashes?

On Thu, Jul 12, 2018, 12:48 PM Charles Collins <charlie47025@gmail.com>

[colesdav]

Interesting.

So if you reduce both the GPU CLK and GPU MEM CLK it runs.

If you reduce just the GPU MEM CLK  it runs.
What happens if you leave the GPU Memory Clock at stock but reduce the GPU Clock to minimum. - Does it crash almost instantly?

RE: could this be a problem where the gpu isnt dumping the old memoryso it runs out and crashes?

It could be a case where the Memory Chips have died because they have somehow been run at too high frequency or voltage.
I do not know exactly what is monitored on that card. You get a GPU Temp reported in the software but I think that is from a sensor on the main GPU.
It could be something has happened to destroy or detach the GPU Memory VRM circuits on the PCB. 
It could be that they or the main GPU are simply starting to detach from the GPU PCB.

You can attempt repair if you have to. But maybe not worth your time and you might be better with a new GPU.

Thanks.

[charliedog]

chkdisk /scan says its not a command or batch file etc.

[colesdav]

Should be chkdsk /scan
I may have made a typo I will check and change it...

[elstaci]

Instruction on running CHKDSK from elevated Command Prompt:

1. Click Run as administrator
2. Log in as an Administrator
3. When Command Prompt launches, type the command:chkdsk C: /f /r /x
4. The parameters for this command are:/f option will attempt to fix any found errors
5. /r option will locate for bad sectors and recovery any readable information
6. /x option will force the volume you’re about to check to be dismounted before the utility begins a scan
7. If the C: drive is in use, type Y to run a scan at your PC’s next restart. If so, exit Command Prompt and restart the computer.

Make sure you are on C:\ drive. type "CHKDSK /f /r

[charliedog]

so this is just confusing: stock clocks - crashes immediately

both clocks on minimum (underclocked) - no crash

clock on minimum memclock on stock - no crash

clock on stock memclock on minimum - no crash

memclock on 1/5th stock speed (300mhz) clock on stock - crashes after about

10 mins

all testing was done with OCCT with these settings

[elstaci]

Look at the OCCT image concerning the 12v. It shows 9 volts. That is why it is crashing. PSU is not producing 12 volt under load. 9 volts is way below the 12 VDC standard Tolerances. Under load the 12 volts should be a little less than 12 vdc or something is shorting out the 12 vdc output on the GPU.

It is possible since you underclock the GPU is doesn't require as much power. But with Stock clocks it needs the full 12 vdc to operate and it is not getting it.

This just just my opinion. colesdav can verify if the 9 vdc is normal for 12 vdc output.

EDIT: I just ran a FULL SCREEN WITH ERROR GPU TEST on OCCT. Just for a minuter to show you my 12 vdc when running under load:

I notice that my PSU 3.3 is showing 3.1 which is still under the 10% tolerance. My Corsair is around 7-8 year old. It been showing that for at least a couple of years. When it goes below 3.0 I would need to replace my PSU.

[charliedog]

well unfortunately i cant use OCCT to test at stock speeds to see if it

does push 12v, so if you have a way i can test that i would appreciate it

[charliedog]

if i can definitively say that its the psu i will be very happy, cause

psu's cost much less than gpu's lol

On Thu, Jul 12, 2018 at 3:19 PM, Charles Collins <charlie47025@gmail.com>

[elstaci]

You can underclock your GPU again so that it will not crash.

Then use OCCT PSU TEST to check out the PSU. BUT if your PSU is weak, running the test might burn out your PSU since it will be running the PSU at full load. At best, your computer will just shut down as soon as the PSU is under load and not damage your PSU, at worst, it can fry your PSU and blow the circuit breaker in your house.

The PSU TEST involves running the CPU and GPU tests at the same time.

I didn't realize until I was researching finding software that tested PSU. OCCT is just about the only diagnostic software that test PSUs. That is where I read that if your computer has a WEAK or CHEAP PSU it might burn it out while being tested.

I have a very High Quality Corsair PSU  850 watts (7-8 y/o) and I was able to run the OCCT PSU Test successfully. All PSU Outputs stayed within the 10% Tolerance. The test screen will be the same as the GPU test screen.

I would contact Cooler Master Support and show them this thread or the OCCT Image and let them decide if your PSU needs to be RMAed under Warranty. I presume the PSU has at least a three year Warranty.

EDIT: Just ran PSU TEST on my computer. 12 vdc hardly fluctuate during test. Here is my OCCT Screen shot after the test:

So it seems like locodicegr​ was correct in his assumption. That it is a PSU issue.