The Radeon VII Owners Thread

[northun_munky]

@northun_munky let me tell you something.. Until two days ago, since I started using 20.3.1, I've got zero crashes. I used all 2020 drivers before including 20.3.1 and I was having crashes within 5-10 minutes of BF5 and CODMW as well. I was stuck with 19.5.2|19.7.5|19.11.3 all the time. I had a reason to be swapping in between them, I just don't remember now, but I can say they were stable at BF5, 3DMark, and other games, and mainly stable with CODMW - 19.5.2 was rock solid all over.
And as I said before, any 2020 driver was giving me crashes, freezes, system hang ups. Until I decided to run ramtest from Karhu, why not? BAM, around 1 hour and 11 minutes of test I got an error. But wait, HCI, TM5, AIDA, and some other stress tests, all passed. HCI is really reliable, but did not catch anything. So I've increased my ram voltage (from 1.450V to 1.465), and plim-plim, the error at ramtest was gone, and straight away I connected that to all my ramdom freezes using 2020 drivers, as unstable RAM ruins any heavy demanding game. So I DDUed all, installed 20.3.1 two days ago, and so far, not-a-single-crash.
So, might be worth DDUing all there, reinstalling a driver (20.4.1 has just been released), and upping your ram voltage by a bit, even if you are sure the ram is 100% stable, as I thought mine was. If you are on AMD cpu, and overclocked, try that with stock settings. Or if you are stock on the cpu, make sure is on all auto(no negative offset etc, pbo/xfr off etc), to conduct this test. I hope this helps, every case is different, but looks like my problems was my RAM was not 100% stable.
Let's see now if I'm right, I just installed 20.4.1. I'll report back.

EDIT

For my surprise, and to reassure my instability was caused by unstable ram, I've played two whole rounds of bloody money and no crash. After dinner will try BF5 and will report back.

EDIT 2

Yep, even BF5 is running no issues, two conquest 64 played, butter smooth

@Thomasck interesting... my first thought was a dodgy stick of RAM as I had the same issue with a stick of DDR3 back in my 2500k days. Memtest picked it up then so it was the first thing I tested this time around. But it passed with no errors so I moved on.

I've had 2 days now of no issues what so ever since rolling back to 19.12.1, even with my Undervolt applied so I had just put it down to Drivers. Let me know in a few days time if you are still issue free and I might give it a go. Would be nice to get on the latest drivers. I'd even started t like the 2020 layout and UI lol!

 

[Dave2150]

I've had my Radeon VII since release (feb 2019). Perfectly stable at 970mv, no crashes/black screens, until 2020 drivers came around. These will black screen hard crash (reboot required) if I use chrome with hardware acceleration enabled.

Last (pre adrenaline 2020) driver is 19.12.1, use that if you can't live without chrome hardware acceleration. I'm using the latest 2020 driver due to improvements with HL Alyx on my Index, though sucks keeping chrome hw acceleration disabled.

Was looking forward to buying big Navi, but will sadly have to go NVIDIA 3000 series, unless AMD completely resolve their hardware acceleration driver issues before release. I hope they do, as I'm a big AMD fanboy and have bought their cards since the 9800pro.

 

[LtMatt]

I've had my Radeon VII since release (feb 2019). Perfectly stable at 970mv, no crashes/black screens, until 2020 drivers came around. These will black screen hard crash (reboot required) if I use chrome with hardware acceleration enabled.

Last (pre adrenaline 2020) driver is 19.12.1, use that if you can't live without chrome hardware acceleration. I'm using the latest 2020 driver due to improvements with HL Alyx on my Index, though sucks keeping chrome hw acceleration disabled.

Was looking forward to buying big Navi, but will sadly have to go NVIDIA 3000 series, unless AMD completely resolve their hardware acceleration driver issues before release. I hope they do, as I'm a big AMD fanboy and have bought their cards since the 9800pro.

Do you have any specific repro steps for that black screen crash? Something that triggers the black screen within 30 minutes or less? Or is it completely random and could happen after minutes hours or days?

 

[Dave2150]

Do you have any specific repro steps for that black screen crash? Something that triggers the black screen within 30 minutes or less? Or is it completely random and could happen after minutes hours or days?

Seems to happen mostly when playing back youtube, or other media in chrome on my second monitor, while gaming/browsing/working on my main monitor. Playing back media (typically MKV files) in Potplayer/Media Player classic, and skipping through the timeline seems to also trigger it (while chrome running).

Doesn't always happen within 30 minutes, sometimes works all day, other times I'll get two crashes within a 2 hour period. I appreciate this makes it tough to fix, I've given up and just disabled hardware acceleration in chrome, as that's the magic bullet fix for me, while still using 2020 drivers.

 

[LtMatt]

Seems to happen mostly when playing back youtube, or other media in chrome on my second monitor, while gaming/browsing/working on my main monitor. Playing back media (typically MKV files) in Potplayer/Media Player classic, and skipping through the timeline seems to also trigger it (while chrome running).

Doesn't always happen within 30 minutes, sometimes works all day, other times I'll get two crashes within a 2 hour period. I appreciate this makes it tough to fix, I've given up and just disabled hardware acceleration in chrome, as that's the magic bullet fix for me, while still using 2020 drivers.

Thanks, that's something to go on.

 

[LtMatt]

Seems to happen mostly when playing back youtube, or other media in chrome on my second monitor, while gaming/browsing/working on my main monitor. Playing back media (typically MKV files) in Potplayer/Media Player classic, and skipping through the timeline seems to also trigger it (while chrome running).

Doesn't always happen within 30 minutes, sometimes works all day, other times I'll get two crashes within a 2 hour period. I appreciate this makes it tough to fix, I've given up and just disabled hardware acceleration in chrome, as that's the magic bullet fix for me, while still using 2020 drivers.

Can you provide a DXDIAG report from your system?

 

[Dave2150]

Can you provide a DXDIAG report from your system?

Sure, PM'ing over a link.

 

[derlabzs]

@Thomasck interesting... my first thought was a dodgy stick of RAM as I had the same issue with a stick of DDR3 back in my 2500k days. Memtest picked it up then so it was the first thing I tested this time around. But it passed with no errors so I moved on.

I've had 2 days now of no issues what so ever since rolling back to 19.12.1, even with my Undervolt applied so I had just put it down to Drivers. Let me know in a few days time if you are still issue free and I might give it a go. Would be nice to get on the latest drivers. I'd even started t like the 2020 layout and UI lol!

Might be worth checking on that, at least is something to start with. Or try with only one ram stick instead of two, try lower clock for the ram with increased voltage. Is your cpu AMD? If so, how's CADBUS/procODT/RTT options? To check for ram issues I'd say to use a bunch of tests, if the memtest your mentioned is memtest86, well, it never actually helped me, only just with really gross/big (or if is fiscally damaged) mistakes when configuring the ram, but once you have "all more or less set" it does not catch anything, I could leave it running in loop for days without a single error and then games would crash, then I'd wrongly blame the GPU LOL.
TM5 is a good really quick test, if that passes, I carry on the next testing software. If does not passes, I'd go back to bios and check what's possibly wrong. HCI is good at catching errors, but takes tooooooo looooong, basic version is free but you need to divide the whole amount of free ram by the number of threads your cpu has, and input manually the amount in every HCI you open, and you need one HCI opened per cpu thread you have. Let's say, 12 cores 24 threads and 11gb free of ram. You need to open 24 instances of HCI and input 458MB in each instance, makes sense? - English is not my mother tongue
Last, but not least, there's https://www.karhusoftware.com/ramtest/ it's not free - like HCI Pro or Deluxe - (but it is worthy having it, I've paid for both cause as I did not know the existence of ramtest/harhu when I purchased HCI Deluxe), works similarly to HCI, and is waaaaaaay quicker. In about 10 minutes you can have an idea of what's going on which means around 83% of whole free ram tested, and around 60 minutes you will have around 98.43% of your ram tested. And that was the only piece of software that detected I got a issue with the ram, around 1 hour and 10 minutes of test. As the same way is stable for you, 19.12.1 was stable for me as well, 3dmark doing fine, bf5 doing fine, all ramtests going great even HCI for over 6000% of coverage, but codmw was freezing. Weird, isn't? I was about to RMA the card, although not really sure if possible cause of quarantine. So ramtest/harhu saved me a good time in this matter.

And so far, 20.4.1 has playing just nice, no problems whatsoever and I hope I won't come back here to say I started having problems again with 2020 drivers. - I've been playing all day long literally, so far my wife did not shout at me, yet

Right now I decreased procODT from 34.4 to 32 ohms, the less the better, and the test is going strong for the whole time,

Spoiler

Right before starting this test I've tried 28ohms and I got an error around 13 minutes. So yes, I trust ramtest/harhu is checking ram, at least in my case is has been very useful.

I hope any of this would help you. If you have a Intel CPU, well, I can't help much besides telling you to check your ram timings/voltage and making sure you cpu is stable with multiple runs of blender, cb15/20 and LinpackExtreme.

 

[fizzy]

Switched off my XMP profile, which then ran my memory at 2400 - same issue. Crashing within a few minutes, in fact crashed as it just loaded the level on one occasion, as soon I pressed the trigger to continue.

I have noticed that on quite a few occasions when I reboot then reload the game, when I start again my position is a bit off from when it crashed. Sometimes right against walls etc. Wonder if it's positional readings going a bit haywire occasionally causing renderer to crash?

On the loading screen the map and feet position indicators varies in direction. Don't know why they wouldn't always be straight ahead?

Played borderands 3 for a while before getting Alyx - completed the main storyline without any graphics issues or crashes.

Not tried 20.4.1 drivers yet.

 

[northun_munky]

Might be worth checking on that, at least is something to start with. Or try with only one ram stick instead of two, try lower clock for the ram with increased voltage. Is your cpu AMD? If so, how's CADBUS/procODT/RTT options? To check for ram issues I'd say to use a bunch of tests, if the memtest your mentioned is memtest86, well, it never actually helped me, only just with really gross/big (or if is fiscally damaged) mistakes when configuring the ram, but once you have "all more or less set" it does not catch anything, I could leave it running in loop for days without a single error and then games would crash, then I'd wrongly blame the GPU LOL.
TM5 is a good really quick test, if that passes, I carry on the next testing software. If does not passes, I'd go back to bios and check what's possibly wrong. HCI is good at catching errors, but takes tooooooo looooong, basic version is free but you need to divide the whole amount of free ram by the number of threads your cpu has, and input manually the amount in every HCI you open, and you need one HCI opened per cpu thread you have. Let's say, 12 cores 24 threads and 11gb free of ram. You need to open 24 instances of HCI and input 458MB in each instance, makes sense? - English is not my mother tongue
Last, but not least, there's https://www.karhusoftware.com/ramtest/ it's not free - like HCI Pro or Deluxe - (but it is worthy having it, I've paid for both cause as I did not know the existence of ramtest/harhu when I purchased HCI Deluxe), works similarly to HCI, and is waaaaaaay quicker. In about 10 minutes you can have an idea of what's going on which means around 83% of whole free ram tested, and around 60 minutes you will have around 98.43% of your ram tested. And that was the only piece of software that detected I got a issue with the ram, around 1 hour and 10 minutes of test. As the same way is stable for you, 19.12.1 was stable for me as well, 3dmark doing fine, bf5 doing fine, all ramtests going great even HCI for over 6000% of coverage, but codmw was freezing. Weird, isn't? I was about to RMA the card, although not really sure if possible cause of quarantine. So ramtest/harhu saved me a good time in this matter.

And so far, 20.4.1 has playing just nice, no problems whatsoever and I hope I won't come back here to say I started having problems again with 2020 drivers. - I've been playing all day long literally, so far my wife did not shout at me, yet

Right now I decreased procODT from 34.4 to 32 ohms, the less the better, and the test is going strong for the whole time,

Spoiler

Right before starting this test I've tried 28ohms and I got an error around 13 minutes. So yes, I trust ramtest/harhu is checking ram, at least in my case is has been very useful.

I hope any of this would help you. If you have a Intel CPU, well, I can't help much besides telling you to check your ram timings/voltage and making sure you cpu is stable with multiple runs of blender, cb15/20 and LinpackExtreme.

Thanks for this mate. Having thought I was out of the woods with a days hassle free gaming, I booted up yesterday and found that clocks weren't going higher than 800mhz and games were unplayable. Went back to stock settings again and then had MW crash on me so something still isn't right.

I'm inclined to think that I'm getting the usual driver issues along with some underlying hardware issue. I'll try the ram again using the programs you mention, would be easy if it was a ram issue!

Otherwise, not sure what I'll do as I haven't got an alternative GPU to test, other than an archaic 4850x2 lol!

I have subsequently disabled hardware acceleration on Chrome, but I never used chrome anyway, forgot it was installed actually so doubt it was that.

Edit: just bought the karhu software so will give it a buzz shortly. Forgot how frustrating issues can be lol!

EDIT 2: 9150% no errors. It doesn't look like the RAM is the issue.

 

[derlabzs]

@northun_munky Hum, that's not great news. 9150% was around 2+ hours? I'd go further and leaving karhu overnight, just to be on the safe side. Is your PSU ok/enough? Is you Ryzen 1600 fully stable? I've had the 800MHz issue many times, a reboot would fix but once the driver crashes, it comes back. Do you mind sharing your ram model, and you procodt, cad_bus and cad_bus strength config?

 

[northun_munky]

@northun_munky Hum, that's not great news. 9150% was around 2+ hours? I'd go further and leaving karhu overnight, just to be on the safe side. Is your PSU ok/enough? Is you Ryzen 1600 fully stable? I've had the 800MHz issue many times, a reboot would fix but once the driver crashes, it comes back. Do you mind sharing your ram model, and you procodt, cad_bus and cad_bus strength config?

Might well do just to be sure, although I think it's alright. I don't know what the settings are you mentioned, I've never OC'd the RAM so would guess they would be stock, or set to auto. I can check for you. The 1600 has never been OC'd either. Never seemed great need for 4k gaming.

One thing I did find from trawling the release notes for the latest drivers (2.3.##) is this:

Radeon RX Vega series graphics products may experience a system crash or TDR when playing games with Instant Replay enabled. A workaround for users experiencing these issues is to disable Instant Replay

I would imagine this applies to the Vii as well and thinking about it, I started recording some rounds of Warzone and about the time my issues started, I enabled instant replay and set it to 10 mins so I could record good rounds after they happened.

I've now gone with 2.4.1 and have left instant Replay off, and vii settings at stock. Could it just be this?.....

Hope to god so! Lol! Thanks for all your help mate, I'll did out the settings you asked for.

 

[derlabzs]

@northun_munky the thing about the ram, even if stock is about signal integrity, which can cause instability at some games. Those parameters I've mentioned before can be taken from a screenshot from a updated ryzen master, it shows all of them at the same time. If possible, a print from Taiphoon burner as well if is not asking too much, so then I can compare what is really set, and what is "recommended".
So if the CPU is stock let's not focus in it and assume it's completely stable. I'm really outdated about the zen1 but I'd run some heavy loaded real world benchs (like cinebench) just to double check what is the vcore while on load, sometimes the boards apply a not enough or too harsh LLC to the cpu causing also instability.
Let's hope then the issue is the Replay, I never use these settings as I know they can cause trouble so the only two things I set at radeon software if freesync (which is already enabled I guess) and vsync I change from default to always off. All notifications, etc, I disable all. Oh, and I also delete all games profiles around. Tunning etc which is the tab to overclock etc, I don't even touch cause overclocking though there does not work, so I use MPT for that.
Within the game, graphics options, I set custom frame setting to match my screen, just cause other than that the card would be full on while on menus etc, texture resolution I set to normal (cause I've got crashes while on high in the past, so I kept normal set), shadow map resolution is set to high isntead of extra, particle lighting is set to high. Over post processing effect AA is SMAA T2X, world motion blur and wearpon motion blur are both off, and film grain if OFF. Why am I telling you these settings? Again, before when I was crashing like crazy I reduced all video settings to low, and every match I was increasing the settings one by on and taking note of what was happening. Turned out that texture resolution gave me a few crashes when on HIGH. That might have been just coincidence, or not. What I can tell is I haven't set it back to HIGH, and I still without any crash so far.

I know I'm asking a lot of things, and might not be any of them, but, without checking we would never know what is the culprit, can be the simple fact that you are using Relive.

Also, you are playing at 4K right? Have you tried downscaling to 2K just to check if the problem persists?

EDIT

Do you notice some kind of pixelation around the iron sights, or around the whole gun while playing CODMW?

Spoiler

 

[northun_munky]

@northun_munky the thing about the ram, even if stock is about signal integrity, which can cause instability at some games. Those parameters I've mentioned before can be taken from a screenshot from a updated ryzen master, it shows all of them at the same time. If possible, a print from Taiphoon burner as well if is not asking too much, so then I can compare what is really set, and what is "recommended".
So if the CPU is stock let's not focus in it and assume it's completely stable. I'm really outdated about the zen1 but I'd run some heavy loaded real world benchs (like cinebench) just to double check what is the vcore while on load, sometimes the boards apply a not enough or too harsh LLC to the cpu causing also instability.
Let's hope then the issue is the Replay, I never use these settings as I know they can cause trouble so the only two things I set at radeon software if freesync (which is already enabled I guess) and vsync I change from default to always off. All notifications, etc, I disable all. Oh, and I also delete all games profiles around. Tunning etc which is the tab to overclock etc, I don't even touch cause overclocking though there does not work, so I use MPT for that.
Within the game, graphics options, I set custom frame setting to match my screen, just cause other than that the card would be full on while on menus etc, texture resolution I set to normal (cause I've got crashes while on high in the past, so I kept normal set), shadow map resolution is set to high isntead of extra, particle lighting is set to high. Over post processing effect AA is SMAA T2X, world motion blur and wearpon motion blur are both off, and film grain if OFF. Why am I telling you these settings? Again, before when I was crashing like crazy I reduced all video settings to low, and every match I was increasing the settings one by on and taking note of what was happening. Turned out that texture resolution gave me a few crashes when on HIGH. That might have been just coincidence, or not. What I can tell is I haven't set it back to HIGH, and I still without any crash so far.

I know I'm asking a lot of things, and might not be any of them, but, without checking we would never know what is the culprit, can be the simple fact that you are using Relive.

Also, you are playing at 4K right? Have you tried downscaling to 2K just to check if the problem persists?

EDIT

Do you notice some kind of pixelation around the iron sights, or around the whole gun while playing CODMW?

Spoiler

Morning mate, honestly ask as many questions as you like! You've been really helpful and I appreciate your efforts!

I managed another hassle and crash free evening of gaming last night which is really promising.

I'll check the settings for you today for the memory and get a screenshot. Again for the cpu core.

I never used relive either until I started playing COD. Wanted to record a few rounds so o could see where to improve and kept forgetting to record from the start, hence why I very recently turned on instant Replay.

Your in game settings are almost identical, I run with the high textures though but other than that the same. can't stand film grain and motion blur! Need to play around with the the settings to find out which is responsible for making the shadows pixelated when further away. When dropping in or using a sniper scope it's quite annoying to see blocky shadows!

Yes, I get the annoying pixelation around the sights. Play with a few xbox players and they dont get it, just seems to be a PC thing.

 

[Kavalier]

So, hi guys, I disabled Hardware Acceleration from my browser (I use Brave Browser) and so far after 1 week, no issues (black screen or system hangs than black screen) were noted. Am running driver version 20.3.1 (20.4.1 is not available for me).

For testing purpose I did a modest under-volt to 995 but kept the same frequency.
The fan curve was max at 50/% fan speed.
Power Tuning and VRAM were kept to defaults.

 

[northun_munky]

So, hi guys, I disabled Hardware Acceleration from my browser (I use Brave Browser) and so far after 1 week, no issues (black screen or system hangs than black screen) were noted. Am running driver version 20.3.1 (20.4.1 is not available for me).

For testing purpose I did a modest under-volt to 995 but kept the same frequency.
The fan curve was max at 50/% fan speed.
Power Tuning and VRAM were kept to defaults.

Pleased to hear that mate. I also disabled acceleration for chrome, even though I dont use it. Hope that's it for you!

Quick question, did/do you use relive instant replay at all?

 

[killem]

I've had my Radeon VII since release (feb 2019). Perfectly stable at 970mv, no crashes/black screens, until 2020 drivers came around. These will black screen hard crash (reboot required) if I use chrome with hardware acceleration enabled.

Last (pre adrenaline 2020) driver is 19.12.1, use that if you can't live without chrome hardware acceleration. I'm using the latest 2020 driver due to improvements with HL Alyx on my Index, though sucks keeping chrome hw acceleration disabled.

Was looking forward to buying big Navi, but will sadly have to go NVIDIA 3000 series, unless AMD completely resolve their hardware acceleration driver issues before release. I hope they do, as I'm a big AMD fanboy and have bought their cards since the 9800pro.

I am having the same problem. Complete black screen lock ups on youtube using firefox and playing videos in VLC. Mostly occurs when gogin or coming out of fullscreen modes. only started happening when using the last 2 driver revisions.

 

[Dave2150]

I am having the same problem. Complete black screen lock ups on youtube using firefox and playing videos in VLC. Mostly occurs when gogin or coming out of fullscreen modes. only started happening when using the last 2 driver revisions.

Disabling hardware acceleration in chrome fixes it for me. Sucky solution, but better than random ass crashes all the time.

If still having problems, try disabling hw acceleration in other programs you use.

 

[derlabzs]

@northun_munky good afternoon, jezz, I just woke up , and the little bear (wife) stills sleeping lol. so, try recording a match (which seems to cause crashes to you) and lower the textures to normal to see if it will help. the pixelated thing is really annoying, can't play with any gun without using a attachment. Seems the majority people affected by this issue are Vega users. I could not find anything to fix. If you fancy playing warzone, add me later on, I'm playing a lot through the day, as I won't be able to work until this crazy time is gone, at least. My ID is Labizoni#2810

@Kavalier I'm glat that you solved the issue!

 

[Kavalier]

@Thomasck I am still monitoring it, but so far finger-crossed it looks ok. If it remains this way, next week will try a modest under-volt and move from there.