• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

The AMD Driver Thread

TonyTurbo78 said:
Yep, just guesswork at this point.

Got some rough footage on my phone to show the frequencies and the drops on the LED lights.

Click to expand...
This has always happened to me Tony. Since the beginning this 3dmark's version. It will also show gpu usage dropping in the graph at the end of the run as well. It has always been that way and I've not found a reason why the gpu never stays at 100%.

It has been my experience that as soon as AMD release a driver that seems to improve of GPU usage in 3Dmark they release an update and we are right back were we started. I'm not implying any malice when I say that. It's just that, for me, it's been at least 2 years like this. I can hand count the number of times I saw a solid straight line at the end of my runs with gpu usage. And the next time there is an update be it AMD/3dMark/Windows it goes right back to what your video shows.

I stop using 3Dmark for a while now, because I know that it should be 100% gpu usage until the end of the run.

BTW, nice case I too have it in white :D

Option #1
When you run it again, like you did in your video, also open up task manager. See if you see an app/process called GameBarPresenceWriter
This process only seems to open up when you are in 3D mode (IE: games). This process has been known to cause problems.
 
EastCoastHandle said:
This has always happened to me Tony. Since the beginning this 3dmark's version. It will also show gpu usage dropping in the graph at the end of the run as well. It has always been that way and I've not found a reason why the gpu never stays at 100%.

It has been my experience that as soon as AMD release a driver that seems to improve of GPU usage in 3Dmark they release an update and we are right back were we started. I'm not implying any malice when I say that. It's just that, for me, it's been at least 2 years like this. I can hand count the number of times I saw a solid straight line at the end of my runs with gpu usage. And the next time there is an update be it AMD/3dMark/Windows it goes right back to what your video shows.

I stop using 3Dmark for a while now, because I know that it should be 100% gpu usage until the end of the run.

BTW, nice case I too have it in white :D

Option #1
When you run it again, like you did in your video, also open up task manager. See if you see an app/process called GameBarPresenceWriter
This process only seems to open up when you are in 3D mode (IE: games). This process has been known to cause problems.
Click to expand...

Yep I get that, the GPU dropping after each test is 100% normal as it's not being used between each test, but the core drops constantly each 5 or so seconds and frame tanking has never ever happened to me on any driver before, either when running 3dmark benchmarks or in gaming.

Also, when gaming, even when standing in one spot, the same thing happens. The core drops and the lights show a drop to P0 or P1. That is not right, that has never happened nor should it ever happen.

Now I'm back on 19.1.2 when I run the exact same test on 3dmark, I get absolutely 100% usage AT ALL TIMES and a solid line throughout the test. Same with games, absolute perfection, no stutter or frame drop anywhere to be seen.
 
Last edited:
TonyTurbo78 said:
Yep I get that, the GPU dropping after each test is 100% normal as it's not being used between each test, but the core drops constantly each 5 or so seconds and frame tanking has never ever happened to me on any driver before, either when running 3dmark benchmarks or in gaming.

Also, when gaming, even when standing in one spot, the same thing happens. The core drops and the lights show a drop to P0 or P1. That is not right, that has never happened nor should it ever happen.

Now I'm back on 19.1.2 when I run the exact same test on 3dmark, I get absolutely 100% usage AT ALL TIMES and a solid line throughout the test.
Click to expand...
Well it happens to me during the run is what I mean. It would be fine and the gpu usage drops. For me though, gpu core clock is the same but the gpu usage goes down while it's still testing in 3dmark.
Now for games I don't get that problem, that I've noticed.

From what you described it does appears to be an issue with the drivers if you go back and it's fine again. But there is one more thing I've seen happen that also causes this issue and it's Windows force update program. Here is what happens and its the most sneaky thing I've seen win10 do:

-AMD announces new update for gpu and you download it
-You uninstall your old drivers and reboot the PC back to desktop.
-Either A. Nothing happens at all and you update normally with no issue
-OR B. Once you reboot your PC your screen will flicker with no announcement why. This is usually win10 force updating to it's default AMD driver for your gpu. You right click on the desktop and see AMD Radeon Settings as an option you know that should not be there. Whatever though...at least you know win10 is working right :confused:
-OR C. nothing happens at all after uninstall of the old drivers and you reboot the PC. Once back to desktop you check Device Manager and confirm through Display Adapter that you have MS Basic Display Adapter. Which confirms there are no AMD drivers installed for your GPU. However, either during or after you install the new AMD Drivers win10 then replaces that driver with it's default one. As far as you know you've installed the new drivers. And have no reason to believe anything else. That's the sneaky part and I experienced this.

In this example, C, you have to confirm the correct driver via Device Manager and through AMD Radeon Settings > System > Software just to make sure you do indeed have the correct drivers. Now I will admit I've not read about this in a while now but I won't ever forget how win10 will auto "update" your gpu drivers once you've done it manually. And the only clue is that your screen will go black more then once during the install process. That's a sure way to know that more then 1 driver has been installed from my experience.
 
Last edited:
TonyTurbo78 said:
Yep, it's clearly not a widespread issue, I've only seen one other person in this forum with the same bug and they said they didn't raise the ticket (just rolled back to 19.1.2). I wonder how many more out there just roll back and wait for the next driver hoping the issue is fixed. That's the problem here, it's only affecting a very small number of people on certain configurations clearly, which means it's going to be a long wait and more luck that it accidentally gets fixed over time I bet.

I did put the ticket in anyway, also posted on the AMD support drivers forum with the video, no one answered or tried to help so meh not expecting any fix on this one for quite a while.

From memory in talking to Shanks when he got his card middle of last year, he was getting the same thing happen on his. His memory would tank all the time to 800mhz, it happened for a few drivers I think but eventually the issue seemed to resolve itself on a newer driver. At least I think that was what happened in talking to him several times.

It couldn't have anything to do with the Bios version that I flashed on to the card could it?
Click to expand...

Yep, I am the guy who have the issue as well. Looks like the 19.2.3 still can't fix it. I wonder if some settings in the driver cause the issue? HBCC On, Freesync On? or Window 10 game mode on? I have reported issue online from Radeon setting and who knows how long the tech support can handle it. Still keep the 19.1.2 and wait for next build.
 
heres evidence of the bug not present before the 19.2.1 driver and the subsequent 2 more versions since that one. This is rolled back to 19.1.2 again, everything is perfect. No frame dips, no throttling or dropping power states. LED's remain maxed out when running games or benching. Compare the two videos, it's absolutely night and day.


Volans_TW said:
Yep, I am the guy who have the issue as well. Looks like the 19.2.3 still can't fix it. I wonder if some settings in the driver cause the issue? HBCC On, Freesync On? or Window 10 game mode on? I have reported issue online from Radeon setting and who knows how long the tech support can handle it. Still keep the 19.1.2 and wait for next build.
Click to expand...

I've tried disabling HBCC, Freesync, not installing Relive etc, nothing solves it.
 
Last edited:
Okay, same here. Disable HBCC, Freesync, even down to 8-bit color and disable HDMI sound all got the same result: serious stuttering in all games.

What I understand is AMD telling me it's time to order the Radeon VII?:mad:
 
Sorry Tony, I tried to see if i could replicate the bug. I loaded your clock settings and used 19.1.1 and 19.2.3.
I'm seeing solid performance in firestrike windowed, without the hitching and clock drops.
 
Davedree said:
Sorry Tony, I tried to see if i could replicate the bug. I loaded your clock settings and used 19.1.1 and 19.2.3.
I'm seeing solid performance in firestrike windowed, without the hitching and clock drops.
Click to expand...

Yep its a strange one this mate, never had this before 19.2.1. Since then each driver has not fixed the issue, nothing I do stops it happening, it happens regardless of clock frequencies and voltages, default profiles...it seems to be a bug driver level with certain configurations I'm guessing. No idea if this will ever be fixed as I bet not enough people that are experiencing this are putting in a support ticket or making AMD aware of it on the AMD Support forums.

19.1.2 is the last driver I can use that works flawlessly, unfortunately I'm missing out on the latest drivers meant for latest game optimisations. It's one of the frustrating bugs man.

Thanks for having a play.
 
I doubt we custom water block on Vega leave the fan pin either unplugged or custom fans connected. Will it be the issue with the new driver? The reference fan profile maybe related in the new driver build or maybe amd has some ******** thing to throttle power when the tach is reading 0 rpm.

If this is true, AMD really give us custom water cooled a punishment.

It's more troublesome to connect the reference fan with the existing water block to give a test. Anyone?
 
Last edited:
Volans_TW said:
I doubt we custom water block on Vega leave the fan pin either unplugged or custom fans connected. Will it be the issue with the new driver? The reference fan profile maybe related in the new driver build or maybe amd has some ******** thing to throttle power when the tach is reading 0 rpm.

If this is true, AMD really give us custom water cooled a punishment.

It's more troublesome to connect the reference fan with the existing water block to give a test. Anyone?
Click to expand...


Thats a cracking idea. I've got a water block on mine and nothing is connected to the fan header. I'll have a look tonight to see if the fan header is visable and if it is do some testing. Hopefully we can sort this out. As you say this has something to do with the vega liquid firmware and a change to the new drivers checking for a fan signal.
 
Last edited:
What a pain I'm not going to be able to do this as its a alphacool completely enclosed cover and I'll have to empty my loop and dismantle the card. What I will do is flash a standard bios onto my card and try the new drivers. I'll report back with what I find.
 
eeii said:
What a pain I'm not going to be able to do this as its a alphacool completely enclosed cover and I'll have to empty my loop and dismantle the card. What I will do is flash a standard bios onto my card and try the new drivers. I'll report back with what I find.
Click to expand...

Was thinking of flashing mine back to stock bios to test also. Can't see it being something like that surely, but who knows.

Are you all running non LC cards but now running in custom loop with a flashed LC bios?
 
Last edited:
Yes, I also flashed to LC bios with air cooling Vega 64.

The difference is I attached three custom fans using Y splitter to the fan header on Vega itself.

So the Tach reading maybe reading error from the normal fan profile? The same thing is 0 rpm maybe.
 
Last edited:
TonyTurbo78 said:
Was thinking of flashing mine back to stock bios to test also. Can't see it being something like that surely, but who knows.

Are you all running non LC cards but now running in custom loop with a flashed LC bios?
Click to expand...

I have both, lc64 and a (P)easant64.
I think i'll try out the p64 and see if it's affected.
 
I've replicated it, I tried my p64 with stock 8706 bios, on 19.2.3 no problems at all, put the 8774 bios back onto the p64 which has the stock cooler on it and throttling and stuttering just like tony's.
So it's not fan header related, I think this is a p-state issue around p5 core, hbm p3.
Still testing ….

 
Last edited:
Davedree said:
I've replicated it, I tried my p64 with stock 8706 bios, on 19.2.3 no problems at all, put the 8774 bios back onto the p64 which has the stock cooler on it and throttling and stuttering just like tony's.
So it's not fan header related, I think this is a p-state issue around p5 core, hbm p3.
Still testing ….
Click to expand...

Interesting, fair play Dave for looking. I'll not be at my pc till about 8pm. @LtMatt interesting.
 
You must log in or register to reply here.
Back
Top Bottom