• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Radeon RX 580 Idle GPU 100% Usage

Athlonite said:
If it's dropping frames then you have a network or browser problem I play 1080p youtube videos all the time and never suffer dropped frames ever even in peak times and i never turn off HWA in my browser
Click to expand...

I should mention my tests are 4K videos. Probably would be fine at 1080. But I do blame the VP9 hardware acceration because it magically disappears every time I turn it off. Can't be browser/network then, can it?

Edit: I should be clear and mention the drop frames are only a little bit here and there for a 60 FPS 4K video. Still, I prefer zero drops that my CPU is capable of.
 
Last edited:
Curious about the upgrade, cost wise was it worth it ? .... looking at TPUs testing...

The Sapphire 580 LE scored 100% @ 1080p to the Reference 970 which got 85%.
The Sapphire 580 LE OC'd 4.4%% compared to a random AIB 970 which managed 17.1% over reference.

580 = 100 x 1.044 = 104.4
970 = 85 x 1.171 = 99.5

What did it cost (cost of 580 vs sale of 970) for the upgrade for the anticipated 4.9% performance increase ... i ask as have a few users looking to upgrade and can't find pricing to make anything worthwhile.

EDIT: Saw later in the thread you were working at 4k which changes things a bit as you could expect a 14% performance increase. Don't have any users 4k in which case anything under a 1080 Ti would be off the table. Still curious tho about what it takes cost wise to get a 15% improvement.
 
Last edited:
@John Naylor my 970 died, so I didn't sell it, and I had to loan one until I got a new card. Long story short, my PC as a whole died, and insurance only paid out enough for motherboard, CPU and RAM. Luckily a friend had the same card as my old 970 spare, so I used that til I could afford a new card. I didn't want to blow my entire bonus last year, which would have happened if I got a 1070 / Vega56, so I settled for this 580 which meant I didn't really look to get a "normal" upgrade.

Honestly, I find it faster in some things (pubg, BF1, Deus Ex Mankind Divided) and about the same in others (BF4). When its faster its a lot faster in my experience, but of course I don't have numbers for you. I was pretty skeptical at first, I honestly thought it was a sidegrade at most, but I was pleasantly surprised.

Other factors are at play, for sure, since most of my experience with my original 970 was at 1080p, very overclocked (I ran it at 1500MHz core) with my 3570K @ 4.5GHz, and now I'm on a Ryzen 7 1700 at stock, though I still play at 1080p. I also think driver optimizations between release and now have helped a bit, too, but as I say I was very pleasantly surprised.
 
hey guys, i'm using rx 560 and notice the same problem. but only when watching youtube on firefox gpu keep running at 100%, game is normal when i quit. and radeon software is 18.2.1.
is chrome has this issue. i've test with edge and everything is fine.
 
Yes this bug is still there in the newest drivers. I think the only fix is to do what was done above and to delete amddvr.exe in safe mode.

Hopefully AMD fixes this soon.
 
Just kill amddvr.exe in task manager, then delete or rename that file and the problem stops. There is also another bug related to this process: in-game clocks remain high, even at low load scenarios and with power efficiency enabled (on my rx 480)
 
I had the exact same issue as you. I have an AMD rx 580 gigabyte aorus 8gb edition. The core clock locks up at max 1365 mhz and the memory clock stays at idle 300mhz, and gpu usage stays at 100% usage after running specific application, and then stopping these applciation. I have not realized this in firefox because I dont use firefox, but in google chrome I did not have this issue. I had this issue specifically issue after running the etherum miner, when i would stop the mining after about 20min I would have this issue (This is not a mining virus I am sure, I ran it with another gpu no problem) (The only way I was able to resolve issue was to restart pc, or just switch gpu mode from graphics to compute or compute to graphics). Another application I realized same issue is Aida 64 gpu stress test, if I stress test gpu for around 30min and stop, then core clock issue comes up. I thought there was a big issues with gpu so I send it to repair, and gigabyte has tested it with their multiple benchmarks and they have not seen any issue relating to core clock so they are returning it to me, it should arrive in my home in Thursday. Can you please test the Aida 64 gpu stress test for around 1 hour with your gpu and tell me if you have this issue when you stop stress test? I have not tested the latest drivers yet because my gpu has been in repair for a long time, but maby it is resolved with new drivers. The last driver I was to use before gpu was in repair was the 18.2.1, now we have the 18.3.2. This issue is very weird. (Just letting you know if you use the amd block chain driver that is beta, issue is resolved. But I dont recommend to use old blockchain driver, especially if you game alot and need stabability). Other then that I am going to contact AMD and see if they can do something about it. So anybody having this issue please try the aida64 gpu stress test for around 1 hour and stop the test and tell me if you are experiencing the core clock and gpu usage stuck at max. Thank you very much if you read the whole thing, there is alot to read!!!!!!!!
 
MaorRX said:
I had the exact same issue as you. I have an AMD rx 580 gigabyte aorus 8gb edition. The core clock locks up at max 1365 mhz and the memory clock stays at idle 300mhz, and gpu usage stays at 100% usage after running specific application, and then stopping these applciation. I have not realized this in firefox because I dont use firefox, but in google chrome I did not have this issue. I had this issue specifically issue after running the etherum miner, when i would stop the mining after about 20min I would have this issue (This is not a mining virus I am sure, I ran it with another gpu no problem) (The only way I was able to resolve issue was to restart pc, or just switch gpu mode from graphics to compute or compute to graphics). Another application I realized same issue is Aida 64 gpu stress test, if I stress test gpu for around 30min and stop, then core clock issue comes up. I thought there was a big issues with gpu so I send it to repair, and gigabyte has tested it with their multiple benchmarks and they have not seen any issue relating to core clock so they are returning it to me, it should arrive in my home in Thursday. Can you please test the Aida 64 gpu stress test for around 1 hour with your gpu and tell me if you have this issue when you stop stress test? I have not tested the latest drivers yet because my gpu has been in repair for a long time, but maby it is resolved with new drivers. The last driver I was to use before gpu was in repair was the 18.2.1, now we have the 18.3.2. This issue is very weird. (Just letting you know if you use the amd block chain driver that is beta, issue is resolved. But I dont recommend to use old blockchain driver, especially if you game alot and need stabability). Other then that I am going to contact AMD and see if they can do something about it. So anybody having this issue please try the aida64 gpu stress test for around 1 hour and stop the test and tell me if you are experiencing the core clock and gpu usage stuck at max. Thank you very much if you read the whole thing, there is alot to read!!!!!!!!
Click to expand...

Submit a bug report and contact a Vangard member
 
MaorRX said:
Can you please test the Aida 64 gpu stress test for around 1 hour with your gpu and tell me if you have this issue when you stop stress test?
Click to expand...

The issue appears to be pretty consistent, anything that clocks the GPU up can get stuck - even FireFox when using hardware acceleration. Following newer drivers from AMD, there have been improvements (PUBG and BF1 for example never leave the GPU core at max when closed anymore), but it sometimes still happens. I've noticed that monitoring software in general can make it worse - like if I leave HWinfo64 running in the background and quit a game, there is more chance the clock can get stuck.

There are mitigations mentioned in this thread though, some are completely removin or renaming the amddrv.exe file, I just end it from task mananger when I see the clocks stick (usually can tell when the gpu fans don't stop). I'm reasonably resigned to having to do this until its fixed for good.
 
PHaS3 said:
The issue appears to be pretty consistent, anything that clocks the GPU up can get stuck - even FireFox when using hardware acceleration. Following newer drivers from AMD, there have been improvements (PUBG and BF1 for example never leave the GPU core at max when closed anymore), but it sometimes still happens. I've noticed that monitoring software in general can make it worse - like if I leave HWinfo64 running in the background and quit a game, there is more chance the clock can get stuck.

There are mitigations mentioned in this thread though, some are completely removin or renaming the amddrv.exe file, I just end it from task mananger when I see the clocks stick (usually can tell when the gpu fans don't stop). I'm reasonably resigned to having to do this until its fixed for good.
Click to expand...

Furmark, and msi benchmark, and firestike does not do this
 
I understand some things do it and some don't. AIDA obviously causes it where Furmark etc. don't. I can understand why Furmark and MSI benchmark wouldnt, as they are stress testing tools and mitigation measures are in place to limit their power usage anyway.

All you can do is manage it the way we have mentioned in the thread, and if you'd like report the bug to AMD - www.amd.com/report
 
What I dont understand is why dont other people have this issue. Why dont youtubers talk about this issue and all of the people who mine with this gpu?
 
As was said earlier if you’re not using ReLive don’t install it. It’s a process related to it.
 
MaorRX said:
What I dont understand is why dont other people have this issue. Why dont youtubers talk about this issue and all of the people who mine with this gpu?
Click to expand...

Because that's a crappy function that most enthusiast or experienced users arent going to install. It's the same reason you don't hear a lot of requests for pants with pockets on the ankles, because no one wants that.

Express install is equal to unwanted trash often ;)
 
jboydgolfer said:
Because that's a crappy function that most enthusiast or experienced users arent going to install. It's the same reason you don't hear a lot of requests for pants with pockets on the ankles, because no one wants that.

Express install is equal to unwanted trash often ;)
Click to expand...
Ya but even when I did a custom install and uncheck the relive crap, there is still the same issue!!!!
 
MaorRX said:
Ya but even when I did a custom install and uncheck the relive crap, there is still the same issue!!!!
Click to expand...

oh, well that sux. It would seem if not installing relive doesnt work, that its not caused by relive.
 
if your using win10 try the minimal install driver. HERE

DO NOT flash the GPU to "fix" the issue.
 
For the same reason that changing the windshield wipers on your car wouldn't fix the issue. Your issue is with something you installed in the driver package, do a clean uninstall and then reinstall the minimal set up , it will be fine .
 
MaorRX said:
Apperantly the AMDVR.exe thing that is causing issues is running even when relive is not installed
Click to expand...

yes you're correct it does run even when relive is not installed as it's used for the gaming overlay that shows stats on the GPU while being used if you don't use this function then rename the AMDVR.exe reboot and hopefully the problem would/should have gone if not then there is some other problem causing your GPU's core clock remain high
 
jboydgolfer said:
For the same reason that changing the windshield wipers on your car wouldn't fix the issue. Your issue is with something you installed in the driver package, do a clean uninstall and then reinstall the minimal set up , it will be fine .
Click to expand...

It has sat in gigabyte repair center, and now it is being sent back to me (They said they have not found any issues in the benchmarks they run which was fire strike and fur mark, it will arive at my home on thursday) . When it arrives at my house I will try the minimal install. I thought that changing the bios might fix some bug in the bios that causes the bios to tell gpu to stay at max core clock and max usage. I will also try renaming or deleting the Amdvr.exe thing. Hopefully the issue will be resolved.

Is it possible that this issue is hardware related?

Is it possible that this issue is hardware related?
 
Athlonite said:
yes you're correct it does run even when relive is not installed as it's used for the gaming overlay that shows stats on the GPU while being used if you don't use this function then rename the AMDVR.exe reboot and hopefully the problem would/should have gone if not then there is some other problem causing your GPU's core clock remain high
Click to expand...

Typically better to use Services.msc to turn off unused services before msconfig.

Your issue is not hardware related, do not screw your card up with a bios flash.
 
You must log in or register to reply here.
Back
Top