The RX Vega 64 Owners Thread

[stuartharrison]

So I've since found that cranking the power limit to +50% and setting memory State 1 to min has since stopped these "Random" black screen crashes. Time will tell.
I've had 3 monitors running video or some streaming or another. Run a Superposition benchmark and temps not above 62c on the card.

We'll see...

I found the 19.3 drivers cured those irritating hang hangs and black screens when switching to full screen watching videos.
For things that only use the card lightly I found locking P1 or P2 as min stopped random black screen crashes - I think they're down to the mem speed jumping up and down.
How do you lock mem to state 1 btw?

 

[stuartharrison]

That's brilliant thank you. Going to have a play with it today. Can I ask why you reduced the p6 frequency as well as voltage, is it to get into p7 quicker? I'll probably have to shave some fan noise off as I sit quite close to the case. Just go myself 3dMark to do some stress testing.

If you want to shave some fan noise off - try using the power saving bios (switch to the left) - set P7@1672/1050, P6@1632/1000, mem @1080/1000, powerslider +50%, fans on auto.
That gave me a steady 1610mhz clock speed (with dips to 1600) and a quiet fan speed of 1130rpm, temps under 70deg, and fast FPS.

 

[stuartharrison]

Good day all, been checking this thread here and there and had to create and account for this issue I believe Thiago is also experiencing.

So I have been pulling my hair out dealing with the Vega 64 instabilities and I think it has to be AMD tinkering with the drivers that is causing it. So I have a DIY AIO water cooled Reference Vega 64, I basically won the dollar scratch off ticket silicon lottery with this one, it is not a great undervolter. So I thought I had a stable profile with the last drivers "19.3.1" - P6 1532 @1125 P7 1652 @1150 Mem 1080 @1100 PL 18% Would run 1630+ mhz all day GPU around 55C, hot spot in the 80's. (fan rpm low and silent)

So I test stability with Valley Benchmark, windowed at 2560x1440 max settings, set camera to free and crank the environment settings (cloud, wind, rain) to max, move around a bit to find a spot in the environment with the lowest FPS and that puts a 99%+ load on gpu. It will usually sit there if stationary running fine, but if its unstable, it will soon crash with "Default Radeon WattMan settings have been restored due to unexpected system failure" when I move around the environment a bit.

Sometimes it will crash and Valley will keep running, but back at stock Vega settings, what I have found out is that after a crash, If I go and reload my profile, the clocks do not go as high as before, and thus doing the same thing with Valley, it no longer crashes. Same thing with Witcher 3, except it always quits running after a crash. I believe this "no longer crashing when reloading the profile" is due to the card not clocking as high as it was after a fresh reboot or having not crashed previously. Glitch with the drivers / software? Now if I reboot first, then clocks are back to normal higher clocks and will crash. (if unstable)

So after recently updating to "19.3.2", I started crashing in Witcher 3 again, I rebooted and ran Valley to check stability there and noticed I was hitting 1650+ mhz! It has never hit that at a 1652 P7. Usually in the 1630's Of course it crashed a short time later when moving around the environment. That time a hard crash, black screen. I rebooted and tried again, and again same higher 1650+ clocks, crashed after moving around a bit, benchmark still running this time. For ***** and giggles I reloaded my profile, (after killing and restarting the Radeon Settings process as it always craps out on a crash) this time the clocks were in the 1630's and no Valley crash.

So there seems to be some software / driver glitch that after a crash, (if not first rebooting PC) and reloading your same profile, is will not running correctly and run at lower max clock speeds.

Also these 19.3.2 drivers for me at least are bumping the MHZ somehow, the same profile is no longer stable as it was with 19.3.1.

About my (default) bios - 016.001.001.000.008892 it seems to clock closer the the actual P state that is set. For instance, P7 at 1652, I'll get close to 1640. This does not seem to be the case with other bios's correct? With the 19.3.2 drivers, P7 @ 1652 was giving 1650's+mhz.


**Update:**

So why this is effecting me? (and possibly Thiago), I think having a profile that runs the card right on the edge of stability. The profile was fine with 19.3.1, now with 19.3.2 and 19.3.3, for some reason my card is clocking around 15 - 20mhz higher, just enough to push it to instability.

I now have to back off P7 to 1642 to get around 1630mhz and keep it stable.

I discovered that when putting the PC in standby, which I do often, then upon waking. The damn clocks are also lower by around 15mhz than they are after a fresh reboot. WTF?

So I also found out I can use this tool restart-1.3-src.zip to restart the display driver instead of rebooting. Which then bumps the clocks back to "normal" after a crash or coming from standby. Its something I've been using with another tool (CRU) for monitor over clocking and Free Sync rate min/max tweaks. All found here - https://www.monitortests.com/forum/Thread-Custom-Resolution-Utility-CRU

Damn fun and pita card this is...

Also posted this in the drivers section, since it pertains to both.

People in the V56 forum, who had flashed their bios to the LC ones, were having problems with the new drivers.
They were fixed by flashing back to the correct bios for the card - which for some resulted in higher clocks.
Did you flash your bios to liquid cool, or is it a liquid cool card?
An air bios would probably allow you to run at lower volts (but connect the fan to the header on the card) - or is that the default bios you mention?
Can you back off the volts on P7, leaving it at 1652, to give you lower and more stable clocks?

 

[AS_Platinum]

I found the 19.3 drivers cured those irritating hang hangs and black screens when switching to full screen watching videos.
For things that only use the card lightly I found locking P1 or P2 as min stopped random black screen crashes - I think they're down to the mem speed jumping up and down.
How do you lock mem to state 1 btw?

Go into wattman and goto the memory section and on the State 1 header, click it and a dropdown will give you the option. And from what I have read, it is to do with the memory being downclocked to 167Mhz. Setting State 1 as the minimum will stop it downclocking to 167Mhz. It will be set as 500Mhz minimum

 

[AS_Platinum]

Had another black screen crash this evening while watching Youtube on one monitor and some CCTV footage on another. So I've updated to 19.3.3.

However, what I did notice was Wattman was setting the P7 State to 1632 which would boost the reported frequency to 1693Mhz. My Card is a Gigabyte RX Vega 64 Gaming OC which the boost clock is 1560. Is this something to be concerned about?

I've since set the P7 state to 1560 and the reported speed is now 1630Mhz.

I've also since run a 3dmark benchmark a couple of times and it's not crashed.

 

[Thiago Diniz]

Good day all, been checking this thread here and there and had to create and account for this issue I believe Thiago is also experiencing.

So I have been pulling my hair out dealing with the Vega 64 instabilities and I think it has to be AMD tinkering with the drivers that is causing it. So I have a DIY AIO water cooled Reference Vega 64, I basically won the dollar scratch off ticket silicon lottery with this one, it is not a great undervolter. So I thought I had a stable profile with the last drivers "19.3.1" - P6 1532 @1125 P7 1652 @1150 Mem 1080 @1100 PL 18% Would run 1630+ mhz all day GPU around 55C, hot spot in the 80's. (fan rpm low and silent)

So I test stability with Valley Benchmark, windowed at 2560x1440 max settings, set camera to free and crank the environment settings (cloud, wind, rain) to max, move around a bit to find a spot in the environment with the lowest FPS and that puts a 99%+ load on gpu. It will usually sit there if stationary running fine, but if its unstable, it will soon crash with "Default Radeon WattMan settings have been restored due to unexpected system failure" when I move around the environment a bit.

Sometimes it will crash and Valley will keep running, but back at stock Vega settings, what I have found out is that after a crash, If I go and reload my profile, the clocks do not go as high as before, and thus doing the same thing with Valley, it no longer crashes. Same thing with Witcher 3, except it always quits running after a crash. I believe this "no longer crashing when reloading the profile" is due to the card not clocking as high as it was after a fresh reboot or having not crashed previously. Glitch with the drivers / software? Now if I reboot first, then clocks are back to normal higher clocks and will crash. (if unstable)

So after recently updating to "19.3.2", I started crashing in Witcher 3 again, I rebooted and ran Valley to check stability there and noticed I was hitting 1650+ mhz! It has never hit that at a 1652 P7. Usually in the 1630's Of course it crashed a short time later when moving around the environment. That time a hard crash, black screen. I rebooted and tried again, and again same higher 1650+ clocks, crashed after moving around a bit, benchmark still running this time. For ***** and giggles I reloaded my profile, (after killing and restarting the Radeon Settings process as it always craps out on a crash) this time the clocks were in the 1630's and no Valley crash.

So there seems to be some software / driver glitch that after a crash, (if not first rebooting PC) and reloading your same profile, is will not running correctly and run at lower max clock speeds.

Also these 19.3.2 drivers for me at least are bumping the MHZ somehow, the same profile is no longer stable as it was with 19.3.1.

About my (default) bios - 016.001.001.000.008892 it seems to clock closer the the actual P state that is set. For instance, P7 at 1652, I'll get close to 1640. This does not seem to be the case with other bios's correct? With the 19.3.2 drivers, P7 @ 1652 was giving 1650's+mhz.


**Update:**

So why this is effecting me? (and possibly Thiago), I think having a profile that runs the card right on the edge of stability. The profile was fine with 19.3.1, now with 19.3.2 and 19.3.3, for some reason my card is clocking around 15 - 20mhz higher, just enough to push it to instability.

I now have to back off P7 to 1642 to get around 1630mhz and keep it stable.

I discovered that when putting the PC in standby, which I do often, then upon waking. The damn clocks are also lower by around 15mhz than they are after a fresh reboot. WTF?

So I also found out I can use this tool restart-1.3-src.zip to restart the display driver instead of rebooting. Which then bumps the clocks back to "normal" after a crash or coming from standby. Its something I've been using with another tool (CRU) for monitor over clocking and Free Sync rate min/max tweaks. All found here - https://www.monitortests.com/forum/Thread-Custom-Resolution-Utility-CRU

Damn fun and pita card this is...

Also posted this in the drivers section, since it pertains to both.

So, after going crazy with this I did my tests on my daily use:

for exemple, today I turned on my PC, run R6Siege and run its built-in Benchmark, almost instantly I got a crash, game closes, gpu driver reset, HWiNFO showed 1770Mhz on core clock (P7 set to 1727 - 1168mv | PL 24%). So I re-load wattman settings, run the game again and I was able to play for about 2 hours straight, with a very solid core clock (about 1680mhz-1690mhz)

Spoiler: HWiNFO - 2 Hours Gaming

After that, I just restarted Windows and I was able to run R6Siege built-in Benchmark for 3x, no crash, clocks close to 1700mhz..

So thats give me a hint that the problem only occurs on cold boot, but thats a pita, how come will I be able to run the card like that? everytime I turn on my PC, I'll have to first run some 3D benchmark to get driver crash and than re-load wattman settings and then I'll be able to use the PC... ??? thats sad! (which is what I'm doing for the last couple of days)

I dont see any performance drops (I'll do more testing later), I'm getting around 1680Mhz-1690Mhz with P7 set to 1727Mhz, so it`s quite acceptable, and Mem at 1100Mhz.. as the matter of fact, I can not even able to measure any performance because any 3D application give me a driver crash at very first run on a cold boot.

I don't use StandBy mode, only turn on/shutdown Windows.

 

[Jon_Snow]

[email protected]/4000 CL17 32GB/RX Vega 64 LC

P7 1737Mhz/1200mv
HBM 1100Mhz/950mv
PowerLimit+50%

 

[AS_Platinum]

Had another black screen crash this evening while watching Youtube on one monitor and some CCTV footage on another. So I've updated to 19.3.3.

However, what I did notice was Wattman was setting the P7 State to 1632 which would boost the reported frequency to 1693Mhz. My Card is a Gigabyte RX Vega 64 Gaming OC which the boost clock is 1560. Is this something to be concerned about?

I've since set the P7 state to 1560 and the reported speed is now 1630Mhz.

I've also since run a 3dmark benchmark a couple of times and it's not crashed.

Right, I've had enough. Old card is back in R9 390X. Which runs like a dog compared. Just gonna do what I do and see if the same issue occurs. If It don't this card is going back.

 

[Thiago Diniz]

I'm starting to think that the problem has to do with Power Limit (bug driver or something like that).

Stock Bios (220W) + PL 24% = 272W
With settings already said above (Clock/Voltage) I have that situation:
-> Cold Boot Windows, 3D Load on Gpu, Drive Crash Right after, after the Drive CrashRe-Load same Wattman settings, I'm good to go.

Now I applied SapphireRXVega64.8176.170811(LC) PowerPlan Table via Reg, so:
Stock Bios turned out to be 264W, so I left PL 0 % with the same Clock/Voltage settings
-> I don't have any problems at all with a Cold Boot + Settings applied on Wattman AND get basically the same performance as 220W+PL 24%

I played 2 full ranked in R6 Siege with no problem at all (264W PPT reg + PL 0), then I changed PL to -10% to see how it would perform and... CRASH! P7 spiked to 1760Mhz... definitely all of this had to do with PowerLimit.

 

[Grog]

Since one of the last few drivers my core voltage is messed up again. On any mode in Wattman the voltage works fine when I start Windows, but after playing a game or stress test the voltage will get stuck at 1050mv and randomly alternate between 1050mv and 1100mv.

 

[Sabrewulf...]

Are you using windows?

If you are then pressing win + ctrl + shift + b will restart the display driver. No need for a tool.

Tried that, does not do the same thing. Maybe just resets the driver only. The tool also reloads "Radeon Settings" it seems. It works, good enough for me.

People in the V56 forum, who had flashed their bios to the LC ones, were having problems with the new drivers.
They were fixed by flashing back to the correct bios for the card - which for some resulted in higher clocks.
Did you flash your bios to liquid cool, or is it a liquid cool card?
An air bios would probably allow you to run at lower volts (but connect the fan to the header on the card) - or is that the default bios you mention?
Can you back off the volts on P7, leaving it at 1652, to give you lower and more stable clocks?

As mentioned, I was on my shipping Bios. I had tried the Sapphire LC, didn't allow any better results. Also tried an older Sapphire Air Bios, also not better.

I just have a piece of **** chip it seems, figures, since I went through the effort of water cooling and can't do much with it. Should have tested it more before modding, could have returned it. I just couldn't stand the damn noise!

Best I can do with current drivers is stock P7 1632 @ 1150mv, PL 18% which gives me between 1619-1625mhz, any less voltage will crash Witcher 3 and eventually in Valley as well. Lower stress games will run 1125mv all day. Max memory is 1080, 1100 artifacts.

Now as I said, drivers "19.3.1" and below, I could get P7 1652 @1150 Mem 1080 @1100 PL 18% Would run 1630+ mhz all day. Not sure whats changed in these last few driver updates. Kinda sucks, but not like it really dropped by performance much, guess it just felt like I was getting some kind of boost over stock. I like to tinker / overclock and this chip is a letdown... Oh well, was $329, pretty good deal and many times better performance at stock than the Sapphire 290 vapor X (great card) I came from.

Hopefully Navi will provide!

 

[AngryFarmer]

If you want to shave some fan noise off - try using the power saving bios (switch to the left) - set P7@1672/1050, P6@1632/1000, mem @1080/1000, powerslider +50%, fans on auto.
That gave me a steady 1610mhz clock speed (with dips to 1600) and a quiet fan speed of 1130rpm, temps under 70deg, and fast FPS.

Thanks for the tip. Will it be a significant performance penalty in 1440p (which is what I game in). I've ordered some more case fans too to improve airflow; I'm hitting just over 90C in the Hotspot under significant load. I've ended up pushing the fan curve ever so slightly up to compensate but I've found even a 5% increase produces significantly more noise. Maximum temps I'm now getting are 74GPU/77HBM/93HS which I think is acceptable?

 

[Grog]

Hardware Unboxed did a Vega 64 vs GTX 1080 video, and he used an Asus Strix Vega 64. He tested with stock clocks (1590mhz/945mhz HBM) and OC clocks (1700mhz/1050mhz HBM) and the difference in most games was minimal. His Overclock was netting him anywhere from about 5-10 FPS more in most games.

IMO that small amount isn't worth the extra heat and power brought on by those overclock settings.

Vega 64's sweet spot seems to be around 1600mhz/1050 HBM, so if you can sustain that you should be fine.

 

[Grog]

I've been having those crashes caused by spikes in clock speed as well. Getting crashes in PUBG lately and checking HWinfo shows that the GPU boosted up a lot higher than it should have even with power limit set low.

 

[Nutella33]

definitely all of this had to do with PowerLimit.

Should get @LtMatt on it and post a bug report

 

[Grog]

So 19.4.1 has been pretty bad for me in terms of stability so I'm thinking of rolling back to an older driver.

Which one would you say was the best driver out of the drivers since 19.1.1?

 

[D1craig]

can anybody offer some advice with my memory overclock? it just wont stick. its staying at 944mhz. ive set it to 1050, 1075, 1100 and it just doesnt change.

EDIT - one thing i didnt think to check was upping the voltage. seems upping the mem volts to 1000 has made my mem overclock work...

 

[iamcll]

Any alphacool eiswolf 120rad users here ?, Would like to know your max temps at full load, And if your heatsink shrouding the gpu get's stupidly hot.

 

[-iudicium-]

Been a while, but reading the issues regarding frame drops and stutters you should revert back to drivers 19.2.1 , at least. After they introduced the mobile drivers, my stutters and crashes are so frequent. SO I went back to older drivers and no issues. Running heavy registry mods and powerdraw up to 380w with my LC bios. There is some sort of "overboosting" issue there again with latest drivers. Memory freq drops to 500Mhz etc. When i monitor the wattage, it drops more than 100w per stutter. Just weird. Fixed something for Radeon 7, but made it worse for Vega 64 I guess. Not sure.

 

[noj]

Had to rma my vega 64, started getting black screen crashes, random crashes with watt man saying its reverted to default settings due to "unexpected system failure" and worsening fps in games - all on stock settings as I've never ventured in over/under clocking gpu's.

As I bought a custom rig from overclockers they've said I won't have to wait for it to go back to the manufacturer so hopefully a quick turnaround for a new card.