7970 Radeon freeze; Pixels issues

[BlueJhins]

Hi Guys ,

Just unpacked my old PC after 1 year in the closet. I'm mostly a player of Warframe and LoL. When I'm playing League of Legends there's no issue but when I'm playing Warframe the game freeze and impossible to do anything. However I can still hear the sound in game when I jump, enter menus , ... . To me it sounds like the graphic drriver crash or something like that ...

This force me to ALT+F4 everytime in game ... When I'm back in the desktop, AMD notify me that Wattman settings have been restored to default.

While being in game (Warframe) the CPU has a temp of 50 - 60 degrees and the GPU 70 - 90, also according to MSI Afterburner th GPU is used at 90%.

Also the few times I could have done something in the game I had some weird pixels issues (like flashy rectangles, idk what is it)
Just played for 40 minutes Risk Of Rain 2, had no issues, works like a charm

 

My specs :
CPU : Phenom II x6 1090T
GPU : XFX 7970 Black Edition
MOBO : GA-990XA-UD3
PSU : Corsair TX 650
RAM : 4x2 DDR3 Corsair XMS3 @ 667 MHz
SSD : Crucial 1TB MX500

I'm using the latest drivers and just did a fresh install of windows. Tried a DDU with latest drivers, nothing changed

Thanks for any help

[taiiat]

it sounds like Warframe is the most GPU intensive game that you have based on the others mentioned, and by quite a significant margin. that could explain why the others don't express any issues.

so, i'll corroborate of try removing the GPU and slotting it back in, and also perform a Clean Install of your Video Drivers of the last version that was updated for that GPU.
acquire Display Driver Uninstaller from one of these loctions:
https://www.guru3d.com/files-details/display-driver-uninstaller-download.html
https://www.wagnardsoft.com/forums/viewtopic.php?f=5&t=2421

simple to operate, clean and restart, then install Video Drivers, restart again.
it's preferable to have the Installer for the Video Drivers you're going to install on your Desktop to make this process simpler.

 

i'll agree that thermals should be taken very close care of, in part because you're using a Tahiti series GPU, in part because of the age of the GPU (your GPU is about 8 years old now), and in part because Thermals can cause things to break. while the GPU edge die Sensor reads a particular number, some other part of the GPU could be warmer, and excessive heat can cause Memory to have artifacting issues (actually just more artifacting issues, all types of RAM have artifacting issues but it gets much worse as Temperatures get very high) and even a couple other Component types as well.
so you want to make sure that your Case Fans are bringing as much fresh air as possible to the GPU, as well as pulling waste heat out. because of the era of GPU that you have.
ideally the GPUs' Heatsink should be spotless in terms of dust, as it very much needs all of the surface area of that Heatsink.

 

 

 

though i might have found some unhappy discussion about VRM cooling on the XFX black model in the process of looking for some images of the PCB & Cooler. which if it's the case (never had one, so i can't verify or refute), then.... there's a reasonable chance that the VRM is getting to overheating Temperatures but apparently the Heatsink on the Card has a solid floor and doesn't make physical contact with the Mosfets so they essentially don't get cooled or very, very limited cooling.
but, we can give some benefit of the doubt and surmise that perhaps the Thermal Pad strip isn't making great contact with the Mosfets. that could be just screws loosening over time, or perhaps the contact even when screwed down just wasn't right to begin with.

this is assuming you're using the XFX model with the custom Cooler (if you're not sure, the Cooler with two Fans rather than one).
while we're here, it can't hurt to set the Fans to 100% and make sure both are spinning, and appear to be spinning at roughly the same speed (so one doesn't look massively slower than the other or something, which could indicate a Fan that could use replacement, replacement Parts might still exist around on the internet).

- - - - - 

if any and all other possible options are exhausted, it'll be a worthwhile endeavor to disassemble the Cooler on this GPU and inspect the Thermal Pads, remount the Cooler to ensure it's as tight as it should be, Et Cetera.
note that to do this you would need:

• an appropriate Screwdriver, probably just a couple different sizes of Phillips type
• some Isopropyl/Rubbing Alcohol(75%+ purity preferable), something to use with the Alcohol (Coffee Filters are the least expensive ideal option for cleaning surfaces without residue, but realistically in a pinch yes you can use Paper Towels and other stuff of that sort)
• some Thermal Paste to replace what's on the GPU die now since you're taking the Cooler off (since GPU dies are a sensitive area i recommend a Paste such as Arctic MX-4, it's very user friendly and safe to use while still being both cost effective and having fairly good performance)
• some patience.

since the floor of the Heatsink is solid, i may even suggest a physical modification to the Heat Sink, that being if you have a Drill around, and appropriate Bits for it (i.e. metal cutting Bits, not wood cutting Bits)

Spoiler

if you do have a Drill and the appropriate Bits (trying to use a Wood Bit on Metal will basically just destroy the Bit, though the other way 'round is ok), then taking a Bit of say around 3-5mm, and poking holes randomly through the area of the Heatsink above the VRM region will dramatically increase the Airflow to that area under the Heatsink (as these small holes will still be a lot more than the zero that it currently gets).

visual aids:
https://www.overclock.net/forum/67-amd/1200211-stay-away-xfx-7970-black-edition-design-flaw-12.html#post20410765

Spoiler

so essentially i'd just be talking about removing the Heatsink from the Cooling shroud, then zinging a bunch of holes through it in random places over in the area above the VRM (the area around the long thin Thermal Pad strip).you want to keep the holes somewhat spaced, you don't want to cut chunks of the Heatsink off, you're just trying to add some holes through it so that the Fans can move some air to the components underneath.
don't Drill through the weld beads though, those smeared lines across the bottom of the Fins is what's holding the Fins together as one solid object. steer around those.
GPU Cooling didn't used to be that great, it's gotten a lot better in the almost decade since this was Manufactured.

 

 

 

definitely feel free to ask questions or look for other measures that can be taken before going to the disassembly and physical modification level. i'm happy to explore other options first though i think you may very well end up there. and even if not maybe you should maintenance the Cooler anyways because of it's age and it has probably never been maintaned before?

Edited October 13, 2019 by taiiat

[Gabbynaru]

Have you tried re-seating your video card? Or maybe changing the thermal paste on the GPU? Cause apparently anything past 85 degrees is dangerous for that GPU, and if yours goes all the way up to 90, you could end up with a dead video card. Artifacting like that could also be a sign of imminent failure.

Edited October 13, 2019 by Gabbynaru

[krc473]

15 minutes ago, BlueJhins said:

My specs :

Your OS+Bit might be useful to add too. It got changed so you must have Win 8 - 10, and 64-bit. The only time I have had a similar issue to this is when parts were getting too hot. I had left a PC to gather dust and it had got inside. The CPU and GPU were both overheating (100C+) and causing the screen to lock. Noises could still be heard, when it was just the CPU the game because playable after the CPU temp had dropped. The solution was to clean the dust out. It was working fine in other games I played, just not Warframe.

 

You might also want to Verify the Download Cache for WF - sometimes there is a corrupt file that causes problems. If you installed it both before and after reinstalling Windows I wouldn't bother.

Edited October 13, 2019 by krc473

[BlueJhins]

Thanks everyone for your replies 😀 !!

@krc473 My os is W10 v1903 (no special custom settings that would harm my PC)

@taiiat I have changed the termal paste 4 years ago for the GPU ( I change my CPU thermal paste every year in December)

Before operating any holes in the heatsink, I'll start with the thermal paste then I'll see what's going on ... 

Also something weird, I run Furmark (OpenCL) for like an hour and Heaven Benchmark (Unigine) for an hour aswell (with different settings) and had none of the issues I have in Warframe ...

Did all the DDU stuff with previous drivers and latests, was still the same problem. 

 

Also I'd like to add aint a big issues to be honest since next month, I'll buy a new PC :

Ryzen 5 3600, Mortar B450M, Noctua U12, 2070S nor 5700XT(not in blower) and with a Define R6 to wrap up everything.

So I can deprive myself of playing warframe for a month ( waiting the november $$$), but still I'll try to work on this thermal paste.

Also if I can't wait any longer to shoot some Grineers, I'll play on my Matebook x Pro

 

I'll keep you updated about the situation of this dinosaur 😂

[taiiat]

12 hours ago, BlueJhins said:

I have changed the termal paste 4 years ago for the GPU ( I change my CPU thermal paste every year in December)

Also something weird, I run Furmark (OpenCL) for like an hour and Heaven Benchmark (Unigine) for an hour aswell (with different settings) and had none of the issues I have in Warframe ...

Also I'd like to add aint a big issues to be honest since next month, I'll buy a new PC

i wasn't suggesting the Thermal Paste on the GPU die was a problem, i assumed it was fine since the GPU is in functioning condition, just no idea what technical level i was working with and mentioning that if you removed the Cooler it would need replacement.
wouldn't be terrible to check the Thermal Pads ofcourse, though. could even be theoretically as simple as some of the Pads having a layer of dust on them or something.

alright, so it can't be issues of the VRM getting toasty then since Furmark will certainly push it harder than any game will. the design of that Cooler is still kinda deplorable but it's sufficient in your case so w/e.

we can continue looking for an explanation if you'd like.
it seems like we've ruled out VRM Thermals, so we can look elsewhere. a powervirus or stress test ofcourse doesn't do much with the GPUs' Memory so there's still some opening left for the possibility that it sounds like Warframe is the most vRAM intensive Application you use.
so i guess there's the possibility that the Memory or some of the Memory is able to get quite hot and the singular Application that uses much more vRAM than the rest could end up with artifacts from that while the others that do much less with vRAM don't. one would still expect some visible issues with Applications using less vRAM but there is... a possibility.
or, we try to come up with a Software only issue. which uh.... i assume you have already tried Verifying Warframe so then all i could suggest after that is making a temporary Partition and installing Windows on that and Warframe and see if it works fine there, idk.

[BlueJhins]

@taiiat I assume that If it didnt crashed with Furmark aint an issue of vRAM tho . 

However will look into the thermal paste and a full clean, might solve the issue tho . Tbh I really don't know where to look at since, bench softs works and not Warframe. Hopefully an issue of dust and paste ... 

I'll let you updated about this case. Will probably do this this w-e since got lots of work in week 😕 

[taiiat]

3 hours ago, BlueJhins said:

I assume that If it didnt crashed with Furmark aint an issue of vRAM tho . 

However will look into the thermal paste and a full clean, might solve the issue tho . Tbh I really don't know where to look at since, bench softs works and not Warframe. Hopefully an issue of dust and paste ... 

Furmark uses like 50MByte of vRAM, and makes no Memory calls to speak of after that. it's displaying a static loop that is complex, but is... static. the only Memory needed is the couple Resources for the Assets, and then the Frame Buffer. Memory is mostly idle in Furmark.

it certainlycan't hurt to do a thorough cleaning.

[BlueJhins]

Even after applying thermal paste on GPU (new one), game still freeze.... good thing no more weird artifacts.

By the way temps seems to be fine now after cleaning and thermal paste.

[sam686]

6 hours ago, BlueJhins said:

Even after applying thermal paste on GPU (new one), game still freeze....

Could maybe try to reduce the clocks of graphics card using overclock utility. There might be "WattMan" in AMD radeon Settings, Gaming, global, Global WattMan, can reduce clocks.

Does EE.log show something after a screen freeze? Note: Restarting game will delete previous EE.log and start with a clean log.

If it shows DXGI_ERROR_DEVICE_HUNG , this is probably the GPU failing.

If problem continues, this GPU (graphics processing unit) might be failing and requires different GPU. Use benchmark comparisons like userbenchmark.com to compare GPU. Either go with newer and more energy efficient graphics card like NVidia GTX 1650 / 1050, or maybe AMD Rx 570 / 580 or the newer, upcoming Radeon RX 5500. Or just cheap out with an old DirectX 11 card.

[BlueJhins]

@sam686 So what I did is just switching the GPU and works like a charm now. Switched to my Sapphire radeon 6870 and no issues after an hour of gaming. Yet having less performance but at least I can farm grineers now.

I think my 7970 is just dying tbh and nothing we can do ... it's ok now I can play with at least 100fps in missions and almost 60 fps in open areas (fortuna and those stuffs)