Infamous R9 380/380X Crash Issues

[ultron]

Hi.

I am one of many people who have serious problems with R9 380/380X. Some of these GPUs have TDR crashes/black screen and lock ups/some weird issues that possibly graphics card related on some games. Interestingly a lot of users reported like me, most of the crashes have some patterns. For instance some users have TDR crashes on Arkham Knight when Batman grapples(and most of them resolve this with playing the game @ 2K VSR), some users says their games crash at certain locations etc. etc.

Do you have any opinion what is the real problem? Most of the Tongas faulty? Drivers mess up everything? Incompatibilty problems? Windows 10? Some PSUs can't run well these card(My gut tells me from when i bought this card it's somehow related with power management issues) AMD is aware of this problem and on their forums they only say "If you have these problems, revert back to 15.11.1 and wait for the fix"; but there's no fix and no explanation what cause this.

What do you think about that? Do you have any advice instead of RMAing the card(because it didn't work for me) or trying another driver(i tried every one of them)

 

[Bacon1]

I'd guess its either drivers (some extra particle effect or something when grappling?) or the game is breaking due to some issue. If its multiple games, I'd say driver issue.

Do you best to record the issue (maybe even use a camera / phone) and send all information you can about your hardware and software to AMD so they can figure it out and reproduce the issue.

 

[sandorski]

Infamous? Never heard of this issue.

 

[PeckingOrder]

AMD drivers are known to be less than ideal

sounds like standard stuff for their products, at least back from when I used them and anecdotal evidence




Threadcrapping not allowed.


esquared
Anandtech Forum Director

 

[Yakk]

Hmm, never heard of it either. Check your install.

 

[Elixer]

AMD drivers are known to be less than ideal

sounds like standard stuff for their products, at least back from when I used them and anecdotal evidence

That hasn't been the case in a long, long time.
Suffice to say, AMD & Nvidia drivers are pretty much on par with each other. They each have issues sometimes.

As for the OP, I have a 380X, and I am having 0 issues, so, sounds like your system is unstable.
Test the RAM with memtest86+ overnight. Then test the CPU with OCCT or prime95 for 4-5 hours. Each of these tests should have 0 errors.

 

[ultron]

Test the RAM with memtest86+ overnight. Then test the CPU with OCCT or prime95 for 4-5 hours. Each of these tests should have 0 errors.

I ran all of these test and i have never any error any other stress tests.

@sandorski many people have problems with these GPUs:

https://community.amd.com/thread/202337#2733900

https://www.reddit.com/r/Amd/comments/3ym7i5/address_the_r9_285r9_380_crashing_problem_plz/

https://www.reddit.com/r/Amd/comments/4f65db/r9_380_users_who_have_black_screen_crashes_help/

https://community.amd.com/thread/202543

 

[Elixer]

Well, obviously, it isn't all 380/380X cards, if it was, then you would be seeing tons of posts about it everywhere.
You see it with Arkham Knight when Batman grapples? Well, tried it, and it worked fine.
Could it be a bad batch of chips? Sure.
Could it be something else? Sure.
What PSU you using, and what are the voltages when you run whatever game has the most issues for you?

 

[PhonakV30]

Can you run VMT(video memory tester) and Test it ? I am waiting for result.Thanks.
Settings :
Test Set : small
Surfaces size : 256x256
Test type : DirectX
Ignore color bit mask : off
Use onscreen buffer : off

https://sourceforge.net/projects/vmst/files/?source=navbar

if you can't download :

http://www.majorgeeks.com/files/details/video_memory_stress_test.html

I had problem with HD5770.I finally noticed that I have 3 type of DDR2 800 with different timing & configuration ( Patriot , Kingston , zeppelin ) , so I reduces Clock to DDR2 666.now I don't have any problem even TDR crash.

 

[ultron]

@behrouz i ran them and they don't detect any defects.

@Elixer Certainly it's not all of them but it looks like R9 380/380X has high failure rate(actually DOA rate) or some compatibility problems with something. RMA don't work, the reseller claims that the card is OK(because they only run standard stress tests).

I'm using Corsair CX600 PSU. Voltages looks fine according to HWINFO64.

About grappling crash:

https://steamcommunity.com/app/208650/discussions/0/535152511376707837/

With 2K VSR game only crashes one time, but @1080p crashes frequently(most of crashes occur when bat is grappling)

@Bacon1: I reported these issues to AMD.

 

[ultron]

Any ideas? I can't RMA it, my first try failed. How can i be sure whether graphics card is faulty or not? If it's a software issues, what should i do? Some games run stable except some stutters but some of them crash frequently.

Here's my complete list

Far Cry 4 Over 12 hours play time with clockblocker(3-4 sessions) and no crash

Tomb Raider Over 13 hours playing(5-6 sessions) with clockblocker no TDR crash but one APPCRASH

Call of Duty Advanced Warfare Over 10 hours play at couple sessions, one TDR crash(while i didn't use clockblocker), then i activated it and the game didn't crash but some stutter and huge fps drops were very frustrating

TrackMania Turbo I played it many times and it crashed with TDR error/without TDR only black screen/only lock up without error message when i didn't use clockblocker. With clockblocker it crashed only 1 time.

Batman Arkham Knight 1080p crashed(tdr) many times(it couldn't last over 30 mins), 2K+clockblocker didn't crash too much

Black Ops 3 TDR crashed/black screened/frozen with even clockblocker. I couldn't play it

The Witness Crashed every time until game's update, after update i didn't play it too much but it crashed one time(and clockblocker wasn't running)

Injustice: Without clockblocker it crashed many times, with clockblocker it never crashed

GTA V: Crashed many times but clockblocker reduce the frequency of the crash. But the game had a lot of stutters.

 

[IEC]

What are your system specs? I would check your card manufacturer's site for any updated video card BIOS, then do the same for your motherboard. Afterwards, if neither of those steps resolve it your next best bet would be to try using a tool like MSI Afterburner to give your graphics card +25mV extra voltage and see if that resolves the crashing.

The weirdest stutter/crashing issues I had previously with a 280X was actually due to the power supply failing. After power supply RMA I had no more stutter and crashing.

So unfortunately, you will have to use the process of elimination to determine what component is the issue.

 

[Ketchup]

Yeah, I am guessing either overheating or pulling too much power. If you downclock the card, do you have the same problem?

 

[ultron]

Yeah, I am guessing either overheating or pulling too much power. If you downclock the card, do you have the same problem?

I tried underclocking and yes same problems continued.

@IEC Both mobo and video card's BIOSes are updated, but didn't add extra voltage because not to void warranty but looks like RMA'll never work i can try this.

Does Clockblocker's help indicate something is wrong with my PSU? I read on an article once and it said "Some PSUs, especially some very powerful are not fast enough, dynamic power demands of graphics cards and that can be the reason some systems never fail on brutal stress tests but they have big problems on some games." I can't remember the source but is it real?

And the Arkham Knight problem that i mentioned earlier; does 2K resolve some people's problems because of it pushes GPU harder than 1080p and prevents downclocking?

 

[Elixer]

Any ideas? I can't RMA it, my first try failed. How can i be sure whether graphics card is faulty or not? If it's a software issues, what should i do? Some games run stable except some stutters but some of them crash frequently.

You RMA'ed the card, and they sent back the same one? Was there any explanation? Did you send it back to the OEM, or the place you bought it?

Here's my complete list
Far Cry 4 Over 12 hours play time with clockblocker(3-4 sessions) and no crash
Tomb Raider Over 13 hours playing(5-6 sessions) with clockblocker no TDR crash but one APPCRASH
Batman Arkham Knight 1080p crashed(tdr) many times(it couldn't last over 30 mins), 2K+clockblocker didn't crash too much
Injustice: Without clockblocker it crashed many times, with clockblocker it never crashed

These are the only games I got in common with you, and I never used--or heard of clockblocker and each of those games, I never had a crash with. I even downloaded Injustice Gods Among Us Ultimate Edition again just to see if it would crash or replicate your issue, and I can't get it to crash.
I have everything set to stock, nothing special going on.

I tried underclocking and yes same problems continued.

@IEC Both mobo and video card's BIOSes are updated, but didn't add extra voltage because not to void warranty but looks like RMA'll never work i can try this.

Does Clockblocker's help indicate something is wrong with my PSU? I read on an article once and it said "Some PSUs, especially some very powerful are not fast enough, dynamic power demands of graphics cards and that can be the reason some systems never fail on brutal stress tests but they have big problems on some games." I can't remember the source but is it real?

And the Arkham Knight problem that i mentioned earlier; does 2K resolve some people's problems because of it pushes GPU harder than 1080p and prevents downclocking?

A simple way to know if the PSU is up to the task is, get a utility that monitors the PSU rails, like CPUID HWMonitor, play your games, and look at the voltages and check min/max values. It will also show the GPU min/max temps as well, along with other stuff.
GPUz can monitor temps & voltages, as well.

It really is too bad you don't have access to another motherboard to try out.
The only other thing you might be willing to try is to install windows 10 (or...) and dual boot your system, then try those games again on the new OS, and see what happens.

 

[ultron]

You RMA'ed the card, and they sent back the same one? Was there any explanation? Did you send it back to the OEM, or the place you bought it?



A simple way to know if the PSU is up to the task is, get a utility that monitors the PSU rails, like CPUID HWMonitor, play your games, and look at the voltages and check min/max values. It will also show the GPU min/max temps as well, along with other stuff.


It really is too bad you don't have access to another motherboard to try out.
The only other thing you might be willing to try is to install windows 10 (or...) and dual boot your system, then try those games again on the new OS, and see what happens.

Reseller always test with standard test protocols and they couldn't detect any problems.

I use HWINFO64 and there wasn't any voltage problems on any rails, but this softwares can't show true readings sometimes. I'll try with voltmeter. GPU temps looks normal, but VRMs could be hot sometimes.

I use dual boot configuration. One Win 10 is on SSD, another is on HDD. I'm gonna install Windows 7 on HDD to see whether this issue relates to Win 10 or not.

I test memory sticks, removing one module, then plug back and try another one, then try different slots, then underclock them. This didn't resolve anything.

I'm planning to RMA CPU,MoBo and PSU respectively.

Clockblocker is a simple program that prevents downclocking on running 3D games. Most of the people use it to eliminate stutters on AMD cards but someone on reddit suggested to use it to solve the issue:

https://www.reddit.com/r/pcmasterrace/comments/4411bg/graphics_card_randomly_shuts_off/czmipvn

http://www.guru3d.com/files-details/clockblocker-download.html

Thank you for your help.

 

[Ketchup]

Reseller always test with standard test protocols and they couldn't detect any problems. ...

What do you mean by this? Who did you get this card from?

 

[ultron]

What do you mean by this? Who did you get this card from?

Local store from Japan.

 

[Ketchup]

Local store from Japan.

Yeah, if you are doing an RMA, you need to go through the card manufacturer, not the shop you bought it from.

 

[ultron]

Yeah, if you are doing an RMA, you need to go through the card manufacturer, not the shop you bought it from.

I sent them but they sent back and they claimed it's OK. It's under distrubitor's warranty.

 

[Ketchup]

Who exactly is this (manufacturer), that you sent it to? Did they ask you to run some tests before sending it back?

If they aren't finding a problem, and based on what you are seeing, it almost sounds like the card is overheating. It could be the power supply, but that wouldn't be my first guess.