[SOLVED] Sapphire NITRO+ RX Vega 64 Issue (Bottleneck?)

[Dale-L-]

Hi guys,

I recently obtained the above card, it's an absolute beast! Although i have a very concerning issue that i just can't sort or find the root cause of. Every now and again my system will crash, with 2 blue screens (no error info), just 2 blue screens and the speakers play a weird signal tone. Nothing responds until i hard reboot the PC. I've ran stress tests and health checks on all components and they are seem fine. The only thing i can think of is either maybe the card is faulty (literally brand new), or the CPU is bottlenecking the card although i don't know how that would cause the system to crash to abruptly. I also used AMD Fresh start and DDU to remove old drivers, tried to update to latest (and beta) drivers. It can run furmark gpu stress test at 4k and sit at stable 40fps, but crash playing Skyrim for 10 minutes in 4k? Temps tend to never go above 70 on full load as the card is very well cooled. CPU is the same. Also i've tried the BIOS switch on the card on both max and eco. Specs are below and any help at all is very much appreciated.

Thanks!

CPU: i5-6500 (Hydro H115i Cooling)
GPU: Sapphire NITRO+ Radeon RX Vega 64
Mobo: Asus Z170 Pro Gaming
RAM: 2x8GB HyperX DDR4 (2666Mhz)
PSU: Corsair 650W +80 GOLD (Modular)
Case: Corsair Graphite 780T

 

[Swarzenegger]

Did you try reinstalling skyrim?

Other reason might be that your card P4/P5/P6 states are not stable, try to lower the clocks for those (higher voltages maybe, put power limit at +50% also). Cause I am guessing it is not running at 100% while playing skyrim. Your PSU should be fine if you can run stress tests.

 

[tennis2]

How long did you run the test on your RAM? Did you have both 8GB sticks in while you tested?

Was this stuff happening at all before the GPU upgrade?

 

[Dale-L-]

Did you try reinstalling skyrim?

Other reason might be that your card P4/P5/P6 states are not stable, try to lower the clocks for those (higher voltages maybe, put power limit at +50% also). Cause I am guessing it is not running at 100% while playing skyrim. Your PSU should be fine if you can run stress tests.

Sorry i should have stated it happens with anything really. It's done it on Chrome, An N64 emu and several games. Card currently has stock clocks out the box. Card isn't at full load on most things but the CPU is the majority of the time. Do you suggest underclocking the card?

 

[Dale-L-]

How long did you run the test on your RAM? Did you have both 8GB sticks in while you tested?

Was this stuff happening at all before the GPU upgrade?

Issue started occurring as soon as the new card went in. I ran the Windows memory diagnostic tool and both sticks were present in dual channel mode.

 

[tennis2]

How long though? I've had bad RAM go for 19 hours on Prine95 without issue. Less sticks/GB takes less time.

 

[Dale-L-]

How long though? I've had bad RAM go for 19 hours on Prine95 without issue. Less sticks/GB takes less time.

It's hard to say as the Windows memory diagnostic tool doesn't run over time, it runs a preset amount of tests and then finishes when it completes. The basic test takes about an hour and the extended test can take up to 2-4 as far as i know.

 

[Swarzenegger]

Yeah the vega cards have issues running their stock clocks. But never had heard of it crashing if not under full load. Underclocking or at least tweaking the lower power states could help... At what MHz does your card run while playing?

If it is running at P7 (all the lights are lit up while playing), undervolt the P6/P7 state with power limit +50.

Could also be a cpu issue, because it might not be running at 100% in benchmarks.

Most of my in game crashes were also ram-related, but seems weird that it didn't happen before.

 

[Dale-L-]

Yeah the vega cards have issues running their stock clocks. But never had heard of it crashing if not under full load. Underclocking or at least tweaking the lower power states could help... At what MHz does your card run while playing?

If it is running at P7 (all the lights are lit up while playing), undervolt the P6/P7 state with power limit +50.

Could also be a cpu issue, because it might not be running at 100% in benchmarks.

Most of my in game crashes were also ram-related, but seems weird that it didn't happen before.

Thanks, i'm not too sure as i'm in work but i'll test it out when i'm in and get back to you on the MHz question. Also apologies but i'm not quite sure what P6/7 states are. Are these power profiles for the card or something?

 

[Swarzenegger]

Yes in the radeon performance settings you can see you 7 power states, going up in frequency and voltage the higher it goes. Vega 56 and 64 are notorious for not reaching their P7 state because they will power throttle before reaching it. So you will run games at a lower MHz than you should. Undervolting (-0,150V ish) the P6 and P7 state usually will make this problem go away, and let's you overclock your frequencies.

But because you are able to run benchmarks and stress tests, it might not be the issue here. Maybe do a full stress test of every component to see if you can get something to crash if undervolting does not work.

 

[Dale-L-]

Yes in the radeon performance settings you can see you 7 power states, going up in frequency and voltage the higher it goes. Vega 56 and 64 are notorious for not reaching their P7 state because they will power throttle before reaching it. So you will run games at a lower MHz than you should. Undervolting (-0,150V ish) the P6 and P7 state usually will make this problem go away, and let's you overclock your frequencies.

But because you are able to run benchmarks and stress tests, it might not be the issue here. Maybe do a full stress test of every component to see if you can get something to crash if undervolting does not work.

Thanks, i'll have another go at component stress tests. I fear i may not get to the bottom of this anytime soon. Might have to send the card back for a replacement just in case.