PC Graphics

[shadowrulerx]

I have a Radeon VII that i bought off AMD website. At first everything was ok, but then i started to see these quick flashes of black lines on my desktop. Then i started getting hard crashes in destiny 2, division 2 and anthem. It even happens randomly while browsing Chrome. Its at stock settings. It can pass Heaven and Superposition on extreme settings for stability. My driver is up to date. Power limit is highest it can go. Idk what to do..

[shadowrulerx]

I have already done DDU for fresh install of drivers for my gpu as well. Im at a loss at what to do. Ive had the card less than a month.

[khun_doug]

There are a few things to take into consideration. First is the PSU. You didn't mention system specs. These GPUs really do use upwards of the 300 watts specified. If you factor in the system load from the CPU being busy, the PSU can get overloaded. I use an EVGA Supernova 850 and in heavy game stress I see my UPS reporting near 600 watts of power. And I am not sure that there are not peaks that the UPS is not reporting. I probably wouldn't use anything under a 750 watt PSU.

The Radeon VII can run hot depending on the load. I've noticed that some games, such Assetto Corsa and Project Cars 2, really stress the card. Even an older game, Two Worlds II, seem to want to thrash the GPU. I had to lower the frame rate on Project Cars and Assetto to 100 FPS to keep the GPU from hitting 100C and higher. It appeared that once the temps were 109C and more the GPU became unstable. On the other hand, Forza Horizon 4 runs at a high FPS and is not as hard on the GPU. I attribute it to game design. FH4 has an option for "unlocked" frame rate. The FPS value varies constantly without the GPU running at 100C or higher. Check your games to see what they have to offer, YMMV.

In Adrenalin there is an option to lower the frame rate per game. You can try lowering your problematic games to 60 FPS and see if the system becomes stable. If a lower FPS resolves the problem then you either have a heat issue or a power issue. If you do get a stable system, try increasing the frame in increments of 10 to try and find a sweet spot.

I use the manual frequency/voltage option to get the best frequency at the lowest voltage. The Auto Undervolt option did not work so well for me. I was able to get a clock of 1870 Mhz with a voltage of 1093 mV. Again, many people are reporting different frequencies and voltages so YMMV. But you really can get some heat savings by finding the right frequency/voltage combo.

[shadowrulerx]

So what would be symptoms of a bad undervolt overclock? Would it cause system lock ups? What was your stock voltage?

[khun_doug]

The first question is more easily answered than the second. What I experience on an under volt that is too low is having both monitors go black and within seconds the sound from the game stops. No way to do a soft recovery from this. Use either the reset button or a power-cycle. So the answer is yes, it was a system lockup but with black displays. I have not had much success going with a clock greater than 1870. I thought I was okay at 1901 but the GPU eventually crashed and I would need to raise the voltage too high, causing more heat. I do see a gain from the default (stock) settings versus the 1870 clock using Unigine benchmark. Incidentally, I leave the memory and fan on automatic. I have never needed to manually raise the memory speed.

As to the stock voltage, that I am guessing at. Upon doing an install from a clean uninstall using DDU or using the Adrenalin Clean option, the card always has Manual for the clock/voltage. Without going through the process again, I know it was at least 1120 mV and a clock of 1801.

If you suspect you are having a clock/voltage issue, use the Reset feature on Global Wattman. That should put you back to the defaults. If not, try using DDU and then a clean install and then check Global Wattman to see what the clock/voltage is. If I was having frequent crashes I would go to the default and then check the game FPS. Try using Wattman to lower the maximum FPS for the game. That also brings up something I discovered by accident. I one of the games there is a FPS limiter. I set the limiter to 100 and the game FPS display showed 100, but the GPU was still running hot, well into the red. Then I used the FPS control in Wattman for that same game and got a much cooler GPU, and did not notice any visual loss, and the game FPS display still showed 100. Crazy, right!

One final thought. My GPU was rock solid in Unigine, so I thought I had a good clock/voltage setting. But using some of the games proved that to be wrong. It could have been a heat issue, but I had to settle on a lower clock/voltage (the 1870/1118). That seems to be a winning combination so far. The machine played several hours of Forza Horizon 4 this evening and no lockups.

[k405nk0nfu510n]

It took me a while to work out my crashing issues, I believe it was Temperature related.  You need to undervolt the card.  I keep mine at 50mhz on gpu and memory over stock and undervolted to 1000mv.  Been working good since. 

[shadowrulerx]

I thought it was too. I was undervolted to 1065mv and crashing happened still (although was overclocked too). I have taken the overclock off and it seems stable. Have not had a crash in division 2 since even after like 5 hours playing. Temperatures dont seem that different though. Im ordering water block for it anyways so if its temp related ill be fine.

[k405nk0nfu510n]

I kept the Clocks Stock and undervolted to 1000mv in MSI Afterburner and it's working great with no more crashes.  

[shadowrulerx]

Well now im kind of curious if it might have been my cpu overclock. I was running 3.8ghz on my ryzen 7 1700x, and i was having absolutely no issue. Then i decided to push it up to 3.9ghz at 1.375v. I suppose it could be possible that i was crashing due to that. So i am currently testing with my cpu at stock settings and a simple undervolt on my Radeon VII. I will update with any information i have.

[k405nk0nfu510n]

I'm overclocking my Ryzen 7 1700 @ 3.9GHz on an AIO and I never have issues, I could do 3.7GHz Stable on the Wraith included.  I don't believe overclocking your CPU is the issue unless your Thermal Throttling.  Mine is 1.375V as well.  The 1700X should have no issue pulling that off.   

[shadowrulerx]

Hmm i have no idea then, thats the only thing that i could think of. Unless the gpu is just faulty

[k405nk0nfu510n]

Keep in mind every Radeon VII is different, they all clock differently and have different voltages stock.  Keep trying to under volt and see if it helps.  Keep the Clocks Stock, and undervol down to like 950mv, and go up incrementally.  Keep your Junction Temps under 90.  I use MSI Afterburner, but you may have something different.  Wattman is useless at this stage.  It'll get better but for now I would stay away from it. 

[shadowrulerx]

It seems if i even try to go under 1065 i crash immediately. Are you using LLC on your OC?

[k405nk0nfu510n]

For my CPU? No I keep it stock, Only thing I did was manually change the Voltage and the Clock.  I have never got LLC to work correctly.  

[shadowrulerx]

Well the strange thing is after i put my cpu to stock i have stress tested for 8 hours now and no crash. But it was weird because the hard lock/crash would happen randomly. The most recent was when i was just on the desktop and came back from bathroom and pc restarted. I check my event logs and it doesnt give any info on why it shut down. I had a GTX 1080 before this card and i was at 3.8ghz cpu OC and never once did i crash. Thats why im wondering if its my CPU OC or the GPU UV/OC. So right now im testing if it still crashes with cpu at stock settings

[k405nk0nfu510n]

I wouldn't believe it's your CPU Clock, I'm running higher than the Boost Clock overclocked, you're not.  So I doubt that is the issue.  Are you OCing your RAM?  I've had issues going over on my ram and too tight of timings.  but that's about it.  I never use Prime 95 to determine if my PC is stable, I use cinebench, If it locks up, back er down lol.  I would set your Radeon VII to 1800 Core and 1000 Memory at 1000mv and see if that helps, then back it down on the Voltage slowly until it crashes and then go up a notch.  

[shadowrulerx]

Well the only reason i say that is because coincidentally this all started after i got my randeon vii and bumped my OC up to 3.9ghz at 1.375v. So im just trying to cut some stuff out of the equation.

[khun_doug]

I haven't tried MSI Afterburner on my EVGA motherboard. I know I tried one software that was supposed to allow me to tweak the GPU but wouldn't even open. I have had success with HWInfo64 and GPU-Z to get information that isn't available in Wattman. But I'm not writing off Wattman. I find it a useful tool when dealing with different settings for different games and even some software that use OpenCL. Each game gets its own settings and they are applied automatically when the game opens. And for adjusting the clock/voltage and even the fan for Radeon VII, Wattman is a great tool.

As for clock/voltage, I was running 1870 at 1113 mV but eventually that rendered the black screens of death. Just bumping the voltage to 1119 stopped the crashes. There is a spreadsheet on another site where people are listing their clocks and voltages. The numbers are all over the place. It appears the Radeon VII is really not a one size fits everyone device. Some people went all the way to 2000 MHz. I haven't been able to get anything that will stay stable that is over 1870 MHz. But 1870 yields a better benchmark than the stock 1801 MHz so I am pleased.

[shadowrulerx]

Yea i thought i was stable at 1900mhz and 1100 on memory with 1095mV but not too sure anymore. Just so tired of all the crashes haha

[k405nk0nfu510n]

Honestly, I see no need to Overclock this card.  I know it can be done, but I don't see too much of a performance Gain.  I run everything maxed in 1440p no problem at stock, and some games at 4K no problem after turning things like DoF off and shadows down.  Those are the two most taxing things on a GPU.  I believe those are your issues, is the overclock hitting Junction Temp and it throttles, and then Black Screens.  That was my issue.  Once EK releases their water block and some other do thing will be entirely different.  As far as Wattman, yes it is a useful program, but even AMD acknowledges issues with it and are wokring on fixes.  

[shadowrulerx]

Well i have already placed an order for the Bykski waterblock. I heard that vega cards really love to be on water cooling. The junction temperature always gets up too high if you dont undervolt the card on air

[khun_doug]

I hear ya and feel your pain. Nothing more aggravating than being in the middle of something or even close to the end of and the monitors go black. GRRRRR!!! In your case, 1100 mV is close to the 1119 mV I needed. I would try dropping the clock in increments of 20. So try running at 1880 and 1100 and see if you if that stops the crashes. Ultimately, there is the heat factor to contend with.

Once you get a stable frequency and voltage please post it. These cards are new enough that people will find this information rather valuable as guides to their own fine-tuning.

[k405nk0nfu510n]

Once again the issue is that Each card is different, each one I have seen has different stock frequencies and voltages. you may not get the same result as the next person.  

[shadowrulerx]

So update here, with my cpu overclock pushed back to stock i havent had any instability so far. Going to test previous settings on card now and see if it was cpu or not

[oxworth]

Resurrecting this, ran into the same problem also under volted for several years, got fed up with it and did a deeper dive on the requirements etc... You need two PCI-e rails for most (maybe all?) PSU, if you used a splitter or double ended cable to power the card, switch to two separate cables. It's almost enough for the card, but weird stuff happens when the power spikes just above the max output of a single rail. It's now stable for me even overclocked and over volted.

[shadowrulerx]

So new update, i am currently stress testing my radeon vii at 1801mhz and 1001mV. So far no crashing and its 28min in. Its plausable that my cpu OC was causing random stability issues? Will update when i know more.

[k405nk0nfu510n]

Ok so when you're done, incrementally increase your CPU Clocks and go from there.  Once you get a crash back it down.  I use Cinebench and OCCT for stressing, I stay away from Prime95. 

[shadowrulerx]

So i can confirm that my cpu overclock is what was causing my instability. Out of curiosity i pushed my cpu back up to 3.9ghz and got a hard lock crash after 2 hours of gaming. When with it at 3.8ghz i can run games all day. My advice to anyone else having this sort of issue is to eliminate other variables and focus only on one item at a time. So far i am runing everythibg at 1801mhz and .993V

[k405nk0nfu510n]

That's interesting, My CPU Overclock is WAYYY over Boost Clock Max and I don't crash.  I'm on a Liquid AIO though.  

[shadowrulerx]

Maybe i just got bad silicone? With this new setup im running games with a max junction temp of 85c atm though, so im good with that for now haha