PC Drivers & Software

[anoriel]

Describe your system:

AMD Graphics Card: Sapphire R9 270X 2GB OC with 1070Mhz base clock and 1400Mhz memory clock

Desktop

Windows 10 64 bit upgraded from Windows 7 64 bit earlier this month

Radeon 18.5.1-may23, 18.6.1, 18.7.1, 18

Samsung S22B150 (1920x1080@60Hz) monitor

Motherboard: MSI Z170A PC MATE (MS-7971) (U3E1)

CPU: Intel i5 6500

Power Supply: Corsair 650W

RAM: 8GB

Problem:

For most of 2017 and 2018 I have been getting crashes in these games: Diablo 3, Heroes of the Storm, Dota 2, Warframe, Far Cry 4 Himalayas levels.

All of these have been crashing ever since the Crimson drivers were released and continue with Adrenaline drivers.

Overwatch crashed for me between 17.7.1 and 17.12.1

The only temporary solution is to use MSI Afterburner to downclock my computer. I do not want to go further back than the 16 series crimson drivers, but when I tested on the 15 series drivers I got the same crashes unless I downclocked.

Heroes of the storm and Dota will crash 10 or more times in a single game unless I downclock my graphics card all the way to 902Mhz core clock and 1100Mhz memory clock. The other games in my list above will run on 999 Mhz core clock and 1200 Mhz memory clock.

When I run Far Cry 4 at normal clock speeds the game runs fine until I go to the Himalayas at which point I get hard lockups. Other than that the game runs perfectly. When I downclock my card I have to run it on medium settings because on high I get small blue square artifacts on the mountains in the background. Diablo 3 and Warframe do not crash at all at 999 core, 1100 memory. I have not tested Dota 2 yet.

Reinstalling Windows or changing drivers seems to have little to no effect on the games that crash. Many games do not crash at all and run fine and many of them are heavy duty current generation games like Witcher 3, Metal Gear Solid 5, the new Doom, Shadow of the Tomb Raider, Shadow of War and lots of other graphically demanding games like GTA 5, Arkham Knight, Dishonored 2, Shadow Warrior 2, Subnautica, Alien Isolation, Dragon Age Inquisition, Assassin's Creed Black Flag, Fallout 4. My card is not overheating and I often take my card out, clean out the dust and reseat it properly.

On Windows Event log I see - Display driver amdkmdap stopped responding and has successfully recovered.

On the AMD settings app I can see - Default Radeon WattMan settings have been restored due to unexpected system failure

With some crashes the Blizzard App will also crash, and for other crashes that does not happen.

Does this mean that my card is dying? Does AMD not support overclocked editions of older cards? Is it a power supply problem which is solved by downclocking? I am seeing many other users getting TDR and black screen crashes with the R9 cards.

[pokester]

Increase the power limit slider to it's maximum. No they don't account for overclocked cards. Only the people that made your card can do that. The vanilla AMD drivers are set to reference card standards. So slide that power limit slider all the way to the right, it should accomplish the same result as underclocking, without the performance loss.

[anoriel]

Is this overvolting or undervolting? I am not very confident in my PSU to try and get more power out of it. I have a 650W corsair power supply but I have many drives connected to my PC. I don't want to do something that will fry my graphics card or PSU. Is there a guide somewhere around here on doing this?

AMD drivers used to work in the past on my graphics card and now they still work on most games. But on some games the R9 270X just crashes whenever there is a lot happening on screen in some multiplayer games. I am not sure what leads to a graphics card using more power but these games cause TDR errors when I do not underclock. It is unusual that more demaning games like Doom or Witcher 3 do not give TDR errors at all. If I crank up the graphics settings I get low FPS, but no TDR errors. If I downclock to 900Mhz core clock and 1100 Mhz memory clock no game crashes, and 999Mhz core and 1200Mhz feels safe for most games that were crashing.

[pokester]

Neither really as it is just allowing the gpu to access the upper range of what is considered normal for that card. So many of the overclock additions of these cards should really be set there by default but are not. This is not the same as a voltage increase, but yes it does allow the card to receive a bit more power under load. If that doesn't get you stable you can do a combination of both, leave the power limit set at max and down clock the card a little bit till it is stable, this typically should not be needed though. Also remember that the power used is dynamic so raising that power limit doesn't mean it uses that power all the time. It is just giving it more headroom when needed. A 650W corsair should be plenty to run that card. It certainly is not impossible the PS could be faulty, it happens. Run OCCT from OCBASE_COM and test your power supply and gpu.

[pokester]

You can reply back with a  DX diag report too from a game crash and we can see exactly what the game is crashing from. Worth trying too is disabling the AMD user experience program, in Radeon Settings, we have recently seen it linked to crashes in DX diag reports.

[anoriel]

I ran OCCT for a few minutes and it seemed normal. I just ran heroes of the storm on low settings with 953Mhz Core clock and 1200Mhz Memory clock with power limit % at +10. Temperatures were normal and I didn't crash. The base clock and memory frequency of this card is 1000Mhz and 1400Mhz. I'll try the game again on medium later on. I am hesitant to increase the power limit all the way if this has a risk of doing damage.

Should I be touching the memory clock at all?

I am sure the black screen and TDR problems on older cards with the new drivers are because of clock speeds and power. I do hope AMD fixes this problem because not everyone will be able to use something like Afterburner or Wattman.

[kingfish]

AMD graphics performance

[anoriel]

The card is well past it's warranty so I don't really need to worry about that. But I do not want to fry my card on purpose.

I'm using Afterburner to downclock my card to make it more stable. If increasing the power limit helps with that I will try it. I don't like erratic black screen and TDR crashes and they seem to be linked to clock and power settings. Much of the advice online is to either reinstall windows, upgrade to windows 10 or use a very old driver. None of them have any effect on R9 card stability issues and resulting black screen crashes. The only thing that helped getting it to work was downclocking.

[kingfish]

Here's a new one for you....and I won't post the dozens of identical posts.

Get rid of Afterburner..or any other third party graphics controller. They conflict with AMD Wattman settings. Then reinstall the AMD driver of your choice Clean Install AMD Graphics Drivers

RX480 FPS drop in all games

Could anyone make the R9 Fury work using Oculus Rift CV1 with any driver newer than 16.11.xx? Please...

[anoriel]

Afterburner is what is making my graphics card usable with downclocking. I've tried many other drivers and I do clean installs every time and on the games I listed my graphics driver crashes with TDR errors.

Afterburner isn't the problem, it's the solution to most TDR errors on R9 and other older cards.

[pokester]

Kingfish's advice is not wrong. Please allow me to explain in a bit more detail.

Afterburner will be an issue if you use any settings in Wattman. It is a one or the other thing. Only recent versions of Afterburner are again really compatible with the Adrenaline drivers and only if you don't change any settings in Wattman. Personally I find no reason to use Afterburner anymore as Wattman does everything it can and much more, but does have a bit of a learning curve.

You are not wrong, many people have had to down clock to get stable results. My guess is they got chips that should have not passed QC or more likely the card maker made an overclock edition with a weak overclocking GPU, that happens a lot, not AMD's fault. Lowering your clock may absolutely be required for playable results for your card. Raising the power limit may reduce how much you need to down clock the card if at all. Unfortunately on top of what is likely a hardware issue for you, the Adrenaline drivers tend to starve the cards of their maximum power on default settings. This is why raising the power limit is probably the the biggest piece of advice given in these forums for stability issues.

Please remember all of us here are users like you trying to help each other and are not AMD employees. Just nice people volunteering our time to help others as we were likely helped when we first came here.  Feel free to kindly disregard any advice you don't feel helps you. We are all only trying to help.

[anoriel]

Is Wattman on Global Overdrive in the Gaming section of the AMD settings app for 18.7.1? Or do I have to download it separately?

If Wattman is a better way to overclock I can use that but right now I'm not facing any TDR crashes or hard lockups using a recent Afterburner version. I did try to play around with power levels and clock speeds and I'm getting good results with the power limit raised to +5-10% and the core clock at 999Mhz and memory at 1200Mhz. The Core clock is just 1Mhz below the base factory setting of the R9 270, but my OC card came with a 1070 default core clock and a 1400 memory clock. I'm trying to see if it's the core clock or memory clock that's the problem but I also don't want to force crashes and hard lockups on my computer for no good reason.

The problem with AMD is that there is no communication from them about problems with their drivers, and they have had a Windows 7 bug with driver settings mismatch for months. If current generation OC cards work fine with AMD drivers, I don't see why we need to blame card manufacturers. Nvidia supports cards that have been OC by card manufacturers to.

People online often suggest unrealistic and painful solutions like doing a clean install of windows when a far simpler solution like downclocking is rarely suggested. There seems to be a tendency to deflect the blame from AMD drivers and assume there is a hardware fault or an OS fault when that's not usually the case. Your advice of raising the power limit seems to be working, thank you for that advice. Nothing against kingfish or you, my complaint was about the general attitude of AMD enthusiasts online.

[pokester]

No unfortunately the full featured Wattman is only available on some of the later generation cards. Yet the older cards are stuck using the watered down version of Global Overdrive. However that still has the same basic features just a bit less customize-able .

Nobody is blaming the card maker only placing responsibility where it belongs. AMD only sells a GPU chip the card maker chooses if they build a reference design that will work with the AMD vanilla driver or if they build something else including OC designs that they need to customize their bios and or drivers to support. AMD can't accommodate for this. There are plenty of examples of this exact same situation on the Green Team too. It is no different aside from yes I think Nvidia currently has better drivers and definitely puts gamers first in a way AMD used to but apparently lost focus on. You warranty, RMA, get repairs and support from the company that made and marketed the card. Not the OEM that sold them the chip that is one part of a bigger product. I don't see where Nvidia's support model is any different other than their execution of the model and public willingness to own an issue and quickly resolve the if possible. They however have a number of long over due promises that have gone unfulfilled too!  This stuff really is always a grass is greener scenario. Things always look better from the outside looking in.

I agree AMD's communication and recognition of issues is poor to say they least. Even in direct contact they don't acknowledge an issue in my experience. It is why it is almost exciting on the seldom times I see them take ownership of an issue in these forums. For 1.5 years now the issues have been IMHO grossly negligent. They don't have claimed features working across the board or at all in some cases. They often fix an issue to break it yet again in the next release. It leaves you thinking the lights are on but nobody is home. I rarely see people in these forums offering advice like re-installing an OS without trying other things first. In my experience that is rarely needed and when it is needed it is because it is the path of least resistance for inexperienced end users. Yes some snob fan boys make it hard for those of us trying to help and even harder on those needing the help. I however was raised that just because someone does you wrong you don't reciprocate in kind. I am a big believer in leading by example and keeping a smile on my face, they are contagious. In my opinion current generation OC from AMD don't work fine. My RX580 requires I alter my default settings and max the power limit as well as various other driver tweaks and system setting changes to avoid constant driver crashes. This should not be the case with anyone's driver IMHO. I never delflected any claim nor said made any incorrect claim as to where support belongs. I absolutely think the driver should do a better job. I also think that the card makers should update their bios so that current driver can work better with them, but truth is I don't think they care about you and only want to sell you another card. Again that is not a fault of AMD completely, now where I personally find fault with this and think that Nvidia does way better is I think they keep a much closer eye on board partner compliance and control what there partners do compared to AMD. This is why in many case there partnerships and product stability are better. The downs side is that also comes at a price. There is a reason you are in many cases at least historically paying that premium for that level of support too.

[benman2785]

"from yes I think Nvidia currently has better drivers " NO THEY HAVE NOT - SINCE YEARS; yet that rumor is spread about AMD since early ATI times! - actually the 2 worst drivers ever are Nvidia!

Wattman for GCN1 cards since 17.4.1

"Even in direct contact they don't acknowledge an issue in my experience." - i did send 4 bugs to AMD since Crimson - all got fixed fast // i also suggested some fixed wor AMD AMF/VCE and the dev responsible for that talked to me several times within 24h - so what are you talking about?

[benman2785]

preowned gpu?
flashed bios yourself?
if (one of them answered with) yes - VGA Bios Collection | TechPowerUp  search YOUR GPU bios and flash

pls deinstall MSI Afterburner!

also: clean cooler with compressed air spray + remove cooler (and old grease) and use a nice thermal grease like Thermalgrizzly "Kryonaut"

do you see green dots etc? your videoram is damaged

[creatie]

I found it's MS issue. Keep you windows 10 version before 17134.81.

One of the patches (KB4284835 or KB4338548) causes my random black screen.