Need help with an RX 6600 I might've completely messed up while overclocking (without messing with voltage)

[DoctorVoltabolt]

I tried overclocking my new RX 6600 following the simplest of OC guides (increment clock by small amounts in MSI Afterburner, stress test after every increment). I was doing this for all of today, and got from 2684mhz to 2810mhz (24mhz increments) with zero problems both in MSI Kombustor's and 3DMark's stress tests (Time Spy and Fire Strike Extreme). Then I had to use the computer for other stuff, so I stopped there and turned it back down temporarily. Temperatures during the tests peaked at 70C.

After that, I turned on a game, and immediately encountered heavy stuttering. Steam's FPS counter showed the usual FPS for that game, but I couldn't even turn the camera around without it semi-freezing (despite the FPS counter not reacting to that in any way). At this point, all my Afterburner settings were back to the way they were initially.
I tried another less demanding game. 99% normal, very slight stuttering sometimes.

I tried launching another game that is much, much less demanding than the previous two, I think it could easily run on an Intel HD graphics (osu!). It was literally unplayable. 10fps, and the cursor freezes every half a second. Of course I tried fiddling with the game's settings, to no avail.

What I tried so far, in order :

1. Restarting computer, obviously did not work
2. Reinstalling drivers the normal way, did not work
3. Deleting MSI Afterburner and then reinstalling drivers the normal way, did not work
4. Deleting my drivers via DDU in Safe Mode, then reinstalling drivers (22.5.2) through AMD's Adrenalin Software. This fixed all problems...for several hours. I changed absolutely nothing, and in ~6 hours everything went back to being exactly the way I described above. I'm completely stumped now.
5. Deleting my drivers via DDU in Safe Mode again, then reinstalling a different version of the drivers (tried both 22.5.1 and the beta drivers Windows installs automatically). This didn't work at all, not even temporarily.

To sum it up, DDU worked once temporarily, and doesn't work anymore.

I also thought about inserting the card into a different PC to check if it works fine. If not, then the card itself is somehow damaged, and if it does then the only thing I can think of doing is reinstalling Windows completely.

If anyone has any better ideas on what to do or what I could've messed up during the OC process, I'll be very thankful...my process was as described above, with power limit at +20% (maximum allowed by the card), temp limit unchanged (card doesn't allow tweaking it) and custom fan curve with fans at 100% after 75C, it never got that high though. The only thing I changed was the core clock, I haven't touched the voltage or even the memory clock. Could I have even messed up the hardware this way?

 

[GerKNG]

Could I have even messed up the hardware this way?

No.

 

[SpittinFax]

Have you run HWInfo64 and just observed what the GPU clocks are doing during the game? I've tried overclocking my RX 6600 and personally I've never even seen the clocks reach that high, so increasing clock speed never really made a difference.

 

[DoctorVoltabolt]

Have you run HWInfo64 and just observed what the GPU clocks are doing during the game? I've tried overclocking my RX 6600 and personally I've never even seen the clocks reach that high, so increasing clock speed never really made a difference.

It wasn't with HWInfo64, but I did look at the clocks through MSI Afterburner when running Kombustor which isn't full-screen. They indeed never went that high, max was 2480 or so if I remember right. Which is why I'm even more confused as to why the OC screwed it up...

 

[SpittinFax]

It wasn't with HWInfo64, but I did look at the clocks through MSI Afterburner when running Kombustor which isn't full-screen. They indeed never went that high, max was 2480 or so if I remember right. Which is why I'm even more confused as to why the OC screwed it up...

HWInfo64 sensor monitoring tends to be the most accurate. And if you're getting huge stutters during gaming, it might be showing up as big clock speed drops. It just helps to narrow down what the cause of the problem might be. If it's the GPU or the game itself.

 

[DoctorVoltabolt]

HWInfo64 sensor monitoring tends to be the most accurate. And if you're getting huge stutters during gaming, it might be showing up as big clock speed drops. It just helps to narrow down what the cause of the problem might be. If it's the GPU or the game itself.

Okay, this is extremely weird.

I tried logging the GPU clocks while opening the most problematic game (literally unplayable, feels like 5fps while being extremely easy to run), and this is what I got :

How is that even possible? I'm not sure what effective means in this context but either way it's at least 5 times less than it should be

 

[SpittinFax]

Okay, this is extremely weird.

I tried logging the GPU clocks while opening the most problematic game (literally unplayable, feels like 5fps while being extremely easy to run), and this is what I got :

How is that even possible? I'm not sure what effective means in this context but either way it's at least 5 times less than it should be

Yeah so the GPU core isn't even trying to boost. Even at 1080p in basic games you can expect to see higher clocks than that.

Can you grab a screenshot of the GPU sensors like this? Under a gaming load if you can. Just interested to see what the GPU is doing. (Edit: This is my own RX 6600 with externally-powered fans, it gives an idea of what a working card should look like)

 

[DoctorVoltabolt]

Yeah so the GPU core isn't even trying to boost. Even at 1080p in basic games you can expect to see higher clocks than that.

Can you grab a screenshot of the GPU sensors like this? Under a gaming load if you can. Just interested to see what the GPU is doing. (Edit: This is my own RX 6600 with externally-powered fans, it gives an idea of what a working card should look like)

View attachment 252083

Weird, the GPU clock seems to be fine...but there's no statistic for the Memory Clock.

The screenshot is from a much more GPU-intensive game which behaves very strange - as long as I don't spin the camera everything renders fine, but the moment I move the camera more than a few degrees it instantly freezes for a second, with the camera turning a lot further than I wanted after.

 

[SpittinFax]

Weird, the GPU clock seems to be fine...but there's no statistic for the Memory Clock.

The screenshot is from a much more GPU-intensive game which behaves very strange - as long as I don't spin the camera everything renders fine, but the moment I move the camera more than a few degrees it instantly freezes for a second, with the camera turning a lot further than I wanted after.

Yeah that looks more like it. At least in terms of maximum values it all seems fine, you've got higher power draw values but that's only because you're running the card under a heavier load.

Something that jumps out at me is that you don't have any entries for memory-related information in the lower-half of the window. You've got memory temperature and power draw, but nothing else. Is it showing any information further down for memory clocks, memory usage, any of that?

 

[DoctorVoltabolt]

These seem to be the only memory-related settings in the lower half. Could that mean the card's memory is faulty?

 

[SpittinFax]

These seem to be the only memory-related settings in the lower half. Could that mean the card's memory is faulty?

I wouldn't call it just yet without further testing, but it could definitely be that the memory is faulty. I haven't had memory-related issues myself so I'm not familiar with the symptoms, but maybe you could try limiting VRAM usage while gaming by dropping texture settings down. And see if that reduces the heavy stuttering.

I also notice that the maximum VRAM usage is 2085 MB, which is very low and awfully close to a rounded 2GB of usage. Maybe only 2GB is accessible or something.

 

[Tatty_Two]

It looks like a possible memory issue, when you did the original overclocking did you max out the power limit slider because even in the TPU review there were all kinds of issues relating to overclocking and power limits, many relating to negative performance?

 

[DoctorVoltabolt]

Yes, I did...every guide recommended to do that up to the maximum limit allowed by the card (in my case, +20%).

Could I have bricked the memory by doing so?

It looks like a possible memory issue, when you did the original overclocking did you max out the power limit slider because even in the TPU review there were all kinds of issues relating to overclocking and power limits, many relating to negative performance?

I wouldn't call it just yet without further testing, but it could definitely be that the memory is faulty. I haven't had memory-related issues myself so I'm not familiar with the symptoms, but maybe you could try limiting VRAM usage while gaming by dropping texture settings down. And see if that reduces the heavy stuttering.

I also notice that the maximum VRAM usage is 2085 MB, which is very low and awfully close to a rounded 2GB of usage. Maybe only 2GB is accessible or something.

Unfortunately the game I'm having the most trouble running (osu!) doesn't have any graphics options, it's an extremely simple game that even an Intel HD graphics should be able to handle, so it most likely doesn't even use that much VRAM. This is what's most confusing for me, there's no consistency in what runs and what doesn't.

 

[Tatty_Two]

Maxing the power limit should not cause this issue, board manufacturers set their power limits based on card model and therefore quality of components for each model, they will always set safety margins, if only because they don't want thousands of cards requiring RMA because of the cards failing within spec.

I think in our review there was the ability to set minimum and maximum boost clocks within the Radeon software, what does it show for you?

If not with your game you should be able to drop resolutions within the benchmarking apps you have been testing with.

As a matter of interest, can you let us know your specific card model plus other system specs such as CPU, motherboard, memory and PSU etc.

 

[R0H1T]

First of all you can't brick just the memory, if anything it will be the whole card. You could try a variety of workloads from 3dmark, games & even other synthetic benches to see if there really is a problem or not.

Memory corruption in my experience is pretty consistent, if it's gone really awry pretty much all workloads will show a little bit to a lot of issues on screen.

 

[Bomby569]

Most likely the card had some problem to begin with, what you did was making it manifest.

 

[SpittinFax]

Yes, I did...every guide recommended to do that up to the maximum limit allowed by the card (in my case, +20%).

Could I have bricked the memory by doing so?



Unfortunately the game I'm having the most trouble running (osu!) doesn't have any graphics options, it's an extremely simple game that even an Intel HD graphics should be able to handle, so it most likely doesn't even use that much VRAM. This is what's most confusing for me, there's no consistency in what runs and what doesn't.

Generally no, it's hard to kill or damage a card just by playing with overclock settings. On mine I've moved the memory clock slider all the way to 1900MHz because that's the hard limit AMD has bestowed upon this card to limit performance.

Have a look at Afterburner and grab a screenshot of it. If it's a memory issue, I wonder if the memory clock slider is even showing up on yours.

 

[Steevo]

I damaged a card once when running multiple software based OC utilities, it pushed the vcore to its maximum and damaged the memory controller when I had it set where I wanted but was trying to see values shown by Afterburner.

If a complete uninstall of the drivers and utilities and reinstall of the drivers doesn't do it it may be the same issue.

 

[kaisersolo]

Why not try using the official software for it - you can oc there. You can also reset back to default settings - I would try it and not use afterburner.

Afterburner is a great utility but if you want a better idea of what's going with your AMD GPU use Radeon Settings, along with HWinfo64