GPU Usage 100% on BF1

iturnip

Commendable
Nov 7, 2016
15
0
1,510
Hi guys

I am getting constant frame spiking and weird green squares when playing battlefield 1.
My card is an Asus Radeon Strix R9 380 4GB and MSI afterburner is showing my GPU usage going all the way up to 100% and then down to like 5% then back up to 100%. I am only trying to play on medium settings.

MSI afterburner is also saying 0 mV of voltage above temperature on the right. Is this normal?

I am also using the card on a PCIE 2.0 board even though it is a 3.0 card. I also have it in the 3rd slot on my motherboard which is a Gigabyte GA-990FXA-UD3.

I am pretty sure my drivers are all up to date, and I'm also using DX12 with the option enabled in bf1.

Any help or suggestions are appreciated.
 
Solution


The R9 family of cards and subsequent drivers were...
Oh you bought a R9 380, sadly there seems to be a driver issue from AMD with the R9 3xx series, many ppl who bought that video card (including myself) have the same issue, the GPU usage goes like crazy from 0 to 100 while also reducing its core and memory clocks.

While I was looking for a solution some ppl mentioned that a pre-Crimsom driver didn't have that issue, around the first driver versions that supported the R9 3xx series, however I was unable to find it (silly me I threw away the cd that came with my gpu), maybe you'll have better luck than me.

For the time being I use "Clock Blocker" to at least keep my GPU core and memory clocks at max, but the GPU usage is still a mess.
 

iturnip

Commendable
Nov 7, 2016
15
0
1,510


Do you also get really bad frames in games? It's weird because overwatch works fine, but BF1 has all green squares over it sometimes and also stuttering
 


The highest he can safely have on that board is an FX 9590 which is terrible. The FX 8350 would be a better choice but still isn't much of a gamer.

No matter what CPU you have, the AM3+ platform is basically dead when it comes to gaming.
 
I haven't tried BF1 yet, but I had issue on several games dropping FPS due to the clock reduction, clock blocker has helped me a lot in that regard.

Those green squares however goes beyond the usual problems with the R9 3xx series, could be a driver specific issue with the game or a bigger problem on your GPU.

Its true AM3+ CPUs don't compare with Intel offerings at the same price, but if you OC it you will get pretty much enough performance for any game you throw at it (60+FPS), big exception being poorly optimized engines like Source (CSGO, L4D2 and such) or Assassin's Creed ported versions.

 

iturnip

Commendable
Nov 7, 2016
15
0
1,510


Sorry I accidently clicked found solution instead of reply.

I have an 8350.
 

iturnip

Commendable
Nov 7, 2016
15
0
1,510


Yeah this is a rig I bought a few years ago, besides the gpu which I bought at the start of the year. I just want it to last me until the next range of CPUs come out next year and replace motherboard, ram and cpu all in one go.
 


Well it seems like you have your next move planned out. I do have some words of warning though.

Remember the problems with the RX 480 reference when it drew too much power from the PCIe slot? Although that problem has been solved via software, several hardware issues are tangled up in the custom RX 480 market. Simply put, some companies have opted for lower tolerance MOSFETs than what were seen on the reference card. Cards with power delivery systems that are not even up to spec with the reference version include the PowerColor Red Devil, the Gigabyte G1 GAMING and possibly others that I haven't seen be put through extensive testing yet.

However, the red team is not the only one experiencing problems with custom cards... The green team has their own set of issues as well. One of the most recent of these issues is the MOSFETs actually failing on any cards with the EVGA ACX 3.0 cooler on them. Since this cooler is seen on EVGA's GTX 1070 and 1080 cards, nearly all of these cards are affected. But there is a new BIOS available for all affected cards and you can contact EVGA to get free thermal pads to connect the different components of the power delivery systems to the cooler and backplate. Why this wasn't done at factory is beyond me.
 
First off the gpu needs to be in the closest slot to the cpu which should be pcie slot 1. Secondly BF1 is not as much cpu intensive as it is GPU intensive. I use 2 gtx 1080's and play at 3440x1440 and get high usage. My wife runs 2560x1080 using a single 1070 and its pretty much maxed at 100% usage while playing with fps locked at 144. Make sure your gpu is not getting hot and downclocking itself. Use a custom fan profile inside the settings of msi afterburner and set the fan to be like 100% at say 78c or whatever temp you want. This will ensure your gpu stays cool. Report back after having done these things. Also make sure your motherboard drivers and BIOS are up to date.
 

amtseung

Distinguished


The R9 family of cards and subsequent drivers were notorious for downclocking cards based on load, even when heat was not an issue. AMD called it Powertune or something. It was their attempt at competing with Nvidia at being power efficient. To me, it's the single worst thing in the development of AMD drivers. It needs to be removed and discarded ASAP.

OP, you have a couple options here. You can run Clockblocker as was suggested above, or run RadeonPro, which can similarly lock your graphics card at max clock speeds and voltages all the time, or you can take the easy way out by selling the 380 and buying a 1050ti. With the first two options, noise and heat will become an issue, and your power bill will go up. With the third option, it'll cost more money up front for better performance and far less headache in the future.

Oh, and DX12 in BF1 isn't truly DX12. It's DX11 wrapped in DX12. You can google it for more information.

If Afterburner says "0 mV" and you haven't overclocked the card, that's normal.

Always put the primary graphics card in the top most PCIe x16 slot.

There's no way an R9 380 can saturate all the PCIe lanes running at 2.0 speeds, let alone 3.0, if my 380x isn't able to. No cause for concern here.

TL;DR AMD Drivers are at fault here, and you can't fix it. Workarounds listed above.
 
Solution
you can lock clock speeds through regular gpu software just setting it to run max performance at all times. Not sure amd software has that but nvidia does in the control panel. No need for some random 3rd party software. Also Noise is not an issue ESPECIALLY if you use headphones and most 3rd party coolers even at 100% fan speed aren't audible over regular pc speakers either. Secondly power bill going up? From what? He already runs a 380 turning his fan speeds up on it will not cause ANY noticeable difference in electricity use wtf are you talking about and neither will running the card at full speed all the time. The card can run at max speed all the time but unless a load is on it it's not really drawing anything extra power wise. Honestly what i would do is use DDU software and reboot into safemode then use the highly recommeded setting of removing drivers then boot into windows regularly and download and install the newest drivers from amd. As stated make sure your motherboard drivers and bios are up to date too and report back.
 

amtseung

Distinguished
If there was an option to lock the clock speeds within AMD's own drivers, no one would be complaining about these issues. The problem is that it was removed, when there used to be one in CCC. None of the other third party overclocking utilities, including GPUTweak, Afterburner, Trixxx, and even AMD's own overdrive have the functionality to do so. The fact that it was quietly removed when Crimson first came out says something.

Fan speed isn't what's going to cause heat and a power bill increase. It's how Clockblocker and RadeonPro's clock locking works. It doesn't just lock your GPU at max speeds, it locks it at max voltage. They pretty much pin your GPU at 100% everything, regardless of load, to prevent Powertune from doing its thing by axing voltage and dropping C-states harder than Skrillex drops the bass. And by noise, it could be fan speed, it could also be coil whine. AMD GPU's tend to have more coil whine than their Nvidia counterparts. Wearing headphones or blasting your speakers is a potential workaround, but not everyone lives alone in an acoustically sealed box, not everyone has tolerant neighbors, and not everyone has the head shape (or ear canal shape in the case of IEM's) for comfortable long term use of headphones or earphones.

As I said earlier, reinstalling drivers won't fix the issue. It may, in fact, make it worse, as I found out the hard way. Powertune is working as intended by downclocking your card based on load and by downscaling the power target, so that maxing the power target slider at 120% is more like... 60%. Custom bios time boiz.
 

Adrenalinemech

Commendable
Nov 18, 2016
2
0
1,510


Download latest Crimson Driver ver 16.11.4. Click Gaming tab, then Global Settings tab and make sure Power Efficiency is turned OFF! No need for 3rd party software like ClockBlocker. Go play and have fun! I have an XFX R9380 4gb card and have not had these issues with the Crimson drivers since March earlier this year. Am also running an FX8150 processor OC'd @ 4.2ghz and I play BF1 @60 fps on medium and high settings with little difference between the two.
 

amtseung

Distinguished


Actually, what's funny is that 16.11.4 broke my custom bios that had been so good at mitigating any voltage and core clock fluctuations that had plagued me since the day I bought that card. I had a nice stable 1.2ghz core and un-overclocked memory at +66mV dialed in with the amazing Sapphire Nitro cooler keeping it in check. After the update, BAM, card was unable to go above 900mhz under any load before dropping core voltage harder than Skrillex. I suddenly found my power limit scaling and temp limits set back to their old values, and being unable to scale the true power limit anywhere above 75-80% anymore, the voltage and core clock slapping the top and bottom of the charts back and forth drove me nuts once again.

I know that this is a fringe case, and definitely an example of a worst case scenario, but if it's happened to me, it very well could happen to someone else.

Instead of attempting to do more custom bios stuff or reinstalling clockblocker, I solved the problem by spending $300 (woohoo for wonky sales pricing) on a 6GB MSI GTX1060 gaming X. I've had no problems since day 1, boosts as high as is promised on the box in OC Mode, and the microstutter is gone. Praise the lord, I mean papa vanu...
 

iturnip

Commendable
Nov 7, 2016
15
0
1,510


Hey mate

I went into my crimson settings and I already have power efficiency turned off so I guess I am going to have to try clockblocker. This GPU has been the worst buy for me, only bought it at the start of the year and payed $350 Australian for it. I am just saving up some money to upgrade to a 1070 next year with the release of the next gen of CPU offerings.