AMD GPU gamers only - your opinion on AMD drivers?

[kapone32]

Almost my path exactly. . I’m just waiting for the 6700XT as my next upgrade

Whatever the card is it's mine. I remember when I first got Vega after my Polaris crossfire and getting the same feeling. I can't wait for Big Navi.

 

[INSTG8R]

Whatever the card is it's mine. I remember when I first got Vega after my Polaris crossfire and getting the same feeling. I can't wait for Big Navi.

i just know I’ll be on 1440 indefinitely so I no longer feel the need for the flagship,

 

[HD64G]

I had some instability issues (CTD, not waking from sleep, losing profiles on reboot, etc) due to a few specific driver versions on my RX5700 and my previous RX580.

 

[ratirt]

Oh I'm going balls out 4k have a 4k screen, all I need to get is a proper graphics and I will be OK for some time

 

[BoboOOZ]

I've had a 5700XT for about 9 months. In the beginning, I had lots of crashes, took me about 2 weeks to understand that it was due to interactions between the driver and other tuning software (MSI Dragon Center, MSI AfterBurner). When I realized that, I uninstalled the rest of the software and I kept only the driver, and crashes disappeared. I still got black screens from time to time (without crashes, just the image disappearing) until I understood that my HDMI cable was unable to keep up with the data throughput of the videocard and I replaced it. Since then, smooth sailing for me.

About the driver itself, it's awesome, plenty of nice features, the only 2 defects are:

1. No memory underclocking possible (would be nice, since I'm running the card at 120W, 1700MHz is probably a bit overkill)
2. Whenever the computer crashes for whatever reason (memory overclock, virtualization gone wrong, etc.), the driver detects that and resets to the default settings, so I have to reload the saved settings (minor annoyance). This is a useful feature, but I wish there was a tickable box to disable that.

 

[kapone32]

I've had a 5700XT for about 9 months. In the beginning, I had lots of crashes, took me about 2 weeks to understand that it was due to interactions between the driver and other tuning software (MSI Dragon Center, MSI AfterBurner). When I realized that, I uninstalled the rest of the software and I kept only the driver, and crashes disappeared. I still got black screens from time to time (without crashes, just the image disappearing) until I understood that my HDMI cable was unable to keep up with the data throughput of the videocard and I replaced it. Since then, smooth sailing for me.

About the driver itself, it's awesome, plenty of nice features, the only 2 defects are:

1. No memory underclocking possible (would be nice, since I'm running the card at 120W, 1700MHz is probably a bit overkill)
2. Whenever the computer crashes for whatever reason (memory overclock, virtualization gone wrong, etc.), the driver detects that and resets to the default settings, so I have to reload the saved settings (minor annoyance). This is a useful feature, but I wish there was a tickable box to disable that.

There are tons of posts on the web about MSI program woes. I saw a 10 to 15% increase in performance once I removed Dragon Center. I also no longer use afterburner to log usage when Gaming and that gave me another 5 to 10% increase in performance.

 

[Nater]

Been great since not long after launch. I bought a 5700XT to essentially play Black Ops 4, and they had a few driver releases there where the game was totally broken with stuttering and frame drops. Haven't had issues with anything else since. There was a few weeks there that was a real kick in the nuts though. (new power supply $170, new video card $425, new monitor $800, AND paid $60 for a game I couldn't play, because - drivers)

My kid's rig each had an RX580 for about 2 years (down to 1 now 2+ years)...absolutely ZERO issues.

 

[BoboOOZ]

There are tons of posts on the web about MSI program woes. I saw a 10 to 15% increase in performance once I removed Dragon Center. I also no longer use afterburner to log usage when Gaming and that gave me another 5 to 10% increase in performance.

Yes, it's visible if you look at the MSI software that the development is not super tidy. I think AfterBurner is very useful for Nvidia cards owners, but Radeon users should stick with the driver and that would make most of the "driver" issues go away.

 

[kapone32]

Yes, it's visible if you look at the MSI software that the development is not super tidy. I think AfterBurner is very useful for Nvidia cards owners, but Radeon users should stick with the driver and that would make most of the "driver" issues go away.

The thing is I was a user of MSI afterburner since Tahiti.

 

[BoboOOZ]

The thing is I was a user of MSI afterburner since Tahiti.

I understand, I imagine there was a time when it was super-useful.
The fact that dumb interactions between tuning software can be problematic is obvious only once you understand it. All I had done is open Afterburner once and move a memory slider and forgot about it and it took me 3 weeks to find the root of the problem. And i only understood it because the DragonCenter crashed more obviously

 

[kapone32]

I understand, I imagine there was a time when it was super-useful.
The fact that dumb interactions between tuning software can be problematic is obvious only once you understand it. All I had done is open Afterburner once and move a memory slider and forgot about it and it took me 3 weeks to find the root of the problem. And i only understood it because the DragonCenter crashed more obviously

Exactly Dragon Center is a dog's breakfast and is not smart enough to remember ARGB settings.

 

[TheoneandonlyMrK]

Exactly Dragon Center is a dog's breakfast and is not smart enough to remember ARGB settings.

To be fair, in my opinion, almost, no every single piece of clock and fan control software I have ever used has convinced me that it was shit given enough time, due to janky behaviour.
Latest would be Icue Corsair's, I loved it initially but.
Most are just about capable of their initial design inspiration just not without some annoying shit too, no Asus, Gigabyte, MSI board software I ever tried lasted more then a month before being uninstalled here.

 

[BoboOOZ]

To be fair, in my opinion, almost, no every single piece of clock and fan control software I have ever used has convinced me that it was shit given enough time, due to janky behaviour.
Latest would be Icue Corsair's, I loved it initially but.
Most are just about capable of their initial design inspiration just not without some annoying shit too, no Asus, Gigabyte, MSI board software I ever tried lasted more then a month before being uninstalled here.

Yes, if you look a bit around it's obvious they tend to be developed a bit quick and dirty style.

I know I'm repeating myself with this software interaction thing, but I imagine that that is the main problem behind Navi 21 "driver issues" and I have never seen any of the bigger sites or YouTubers warn against this problem.

 

[fec32a4de]

Never had problems with AMD drivers on my Vega 64, no glitches, problems or crashes - but ReLive or video recording was always a problem. Sometimes recording works, most often doesn't. Switching from h264/avc to h264/hevc doesn't help. Sometimes recording shows that its working, only to end up with a green screen or a tiny 262byte file.

I wonder if AMD drivers work better in regards to recording with RDNA series, but in terms of stability, no issues or problems. No problems in games or software (encoding such as in Magix Vegas Pro).
Just recording issues.

 

[EarthDog]

During reviews, I never saw this problem.

I let one of my kids use a review card on his system with the same driver... black screens. Move it back to mine, none. Try again with new drivers in a couple of months (after fresh OS install for other reasons, mind you) same issue. Not sure what the deal is on that system (B450 + Ryzen 2K CPU), but that is what's happening as of a few months ago. We gave up and he has an NV card.

I know I'm repeating myself with this software interaction thing, but I imagine that that is the main problem behind Navi 21 "driver issues" and I have never seen any of the bigger sites or YouTubers warn against this problem.

It could be software, but that said, why can't it work nice will all software? We don't see this issue on the other side, do we? I don't follow this logic considering we've seen people simply swap to NV cards and it works. So while it may be software causing the issue (doubt it), the competition can figure out how to play nice with it. Root cause analysis, not symptom treatment.

 

[INSTG8R]

During reviews, I never saw this problem.

I let one of my kids use a review card on his system with the same driver... black screens. Move it back to mine, none. Try again with new drivers in a couple of months (after fresh OS install for other reasons, mind you) same issue. Not sure what the deal is on that system (B450 + Ryzen 2K CPU), but that is what's happening as of a few months ago. We gave up and he has an NV card.
It could be software, but that said, why can't it work nice will all software? We don't see this issue on the other side, do we? I don't follow this logic considering we've seen people simply swap to NV cards and it works. So while it may be software causing the issue (doubt it), the competition can figure out how to play nice with it. Root cause analysis, not symptom treatment.

Sadly despite catching users with black screen issues it won’t produce a useable dump file or event that can be used to actually catch the cause to be able to fix it. It’s literally so random it can be a TDR or something completely unknown for lack of data. It’s+literally the most elusive issue

 

[BoboOOZ]

It could be software, but that said, why can't it work nice will all software? We don't see this issue on the other side, do we? I don't follow this logic considering we've seen people simply swap to NV cards and it works. So while it may be software causing the issue (doubt it), the competition can figure out how to play nice with it. Root cause analysis, not symptom treatment.

The important thing for me is to give users that have issues workarounds for their issues.

The less important thing is who's to blame for the issues. It could be AMD, it could be the other software.

I'll give you a personal example. I develop mostly in Java, my code works on Windows and Linux platforms. However it tends to have way more bugs in Linux, because I work in Windows, so I can spot many of the Windows bugs during development.

See how people say that drivers work fine under Linux, but not in Windows? Maybe because there's no afterburner or dragon center for Linux?

Anyways, it's impossible to know for sure, but the answer is pretty far from clear cut.

 

[EarthDog]

Sadly despite catching users with black screen issues it won’t produce a useable dump file or event that can be used to actually catch the cause to be able to fix it. It’s literally so random it can be a TDR or something completely unknown for lack of data. It’s+literally the most elusive issue

Reminds me of intermittent issues with a car.... they are SO TOUGH to pin down and resolve. The dealership doesn't want to throw parts at it (understandable), but the user still has the issue. It is frustrating.

I think either this or the last release AMD finally removed the generic black screen issue that has been noted for well over a calendar year in their release notes.

This won't stop me from getting an AMD card, but the product with the least amount of effort is what will go in my kid's system. I won't buy this card out of the gate until I see a couple of months of 'release notes' from users.

Anyways, it's impossible to know for sure, but the answer is pretty far from clear cut.

Fair enough.... but I again ask, why can team A get it right and team B cannot? Sites don't warn against this issue because they have no idea and reputable sites don't propogate guesses... it 'could' be software... well, it could be a lot of things...

 

[dragontamer5788]

Vega 64. I bought it pretty late into the cycle (only a few months before RDNA), but I can't say I've had any major issues with Vega64 drivers in games.

OpenCL on the other hand has had various issues. I don't really know if it's the driver or say Blenders code, but it's like OpenCL isn't as reliable to use in applications. I did take a look at Blenders OpenCL code and it's horrible, so maybe it's an application specific thing.

 

[INSTG8R]

Reminds me of intermittent issues with a car.... they are SO TOUGH to pin down and resolve. The dealership doesn't want to throw parts at it (understandable), but the user still has the issue. It is frustrating.

I think either this or the last release AMD finally removed the generic black screen issue that has been noted for well over a calendar year in their release notes.

This won't stop me from getting an AMD card, but the product with the least amount of effort is what will go in my kid's system. I won't buy this card out of the gate until I see a couple of months of 'release notes' from users.

Well I had a user on to AMD discord getting them with a degree of frequency I had him install the reg file to force dumps hasd him check event viewer for=relevant events but just couldn’t get anything conclusive to be able to attacch to the bug repport tool
For the most part it’s become a non issue, yet random untraceable instances still persis. I wish I could get one myself so I could hunt it down myself and maybe have better luck producing some viable info

 

[BoboOOZ]

Fair enough.... but I again ask, why can team A get it right and team B cannot? Sites don't warn against this issue because they have no idea and reputable sites don't propogate guesses... it 'could' be software... well, it could be a lot of things...

Because everybody is (or was) developing software mostly for team A? And because team A software is also much more limited and closed, you can't modify it much, and when you can't modify it you can't break it?

Anyways, I agree with you that AMD drivers should just work. As they are, they are much too powerful and they are hurting themselves with these issues. They should hide all the most powerful customizations under several layers of "Advanced" menus, maybe many normal clients would be way happier.

Well I had a user on to AMD discord getting them with a degree of frequency I had him install the reg file to force dumps hasd him check event viewer for=relevant events but just couldn’t get anything conclusive to be able to attacch to the bug repport tool
For the most part it’s become a non issue, yet random untraceable instances still persis. I wish I could get one myself so I could hunt it down myself and maybe have better luck producing some viable info

Have you gotten any PSU issues out of the way, just out of curiosity?

 

[INSTG8R]

Have you gotten any PSU issues out of the way, just out of curiosity

For me it’s a non issue but “dirty power“ has been proven and using separate pci connections rather than a paired set has actually sorted some users issues
Someone I interact with who’s staff of a hardware producer has used rhis solution to sort a few customers issues basically using this image.

 

[EarthDog]

Because everybody is (or was) developing software mostly for team A? And because team A software is also much more limited and closed, you can't modify it much, and when you can't modify it you can't break it?

Uh, no. Software is developed for Windows not to work with Team A/B's drivers...

 

[P4-630]

For me it’s a non issue but “dirty power“ has been proven and using separate pci connections rather than a paired set has actually sorted some users issues
Someone I interact with who’s staff of a hardware producer has used rhis solution to sort a few customers issues basically using this image. View attachment 172388

I'm doing the "don't do this" at the moment, no issues. No AMD card though. (2070 super)

 

[Ants]

My 5700 XT has served me well so far.

Had an RX 590 before that also served me well.

Waiting for RDNA 2 before I decide if I stay team red or go team green for next upgrade.