So I was lucky enough to get a 6800 xt on launch direct from AMD and received it yesterday. I am having a little bit of an issue with the performance in game regarding the boost clocks and GPU load randomly dropping. While in game the gpu load will drop to 40ish% and the gpu clock will drop to 550ish mhz. This will cause a framerate drop to 15-20 and stuttering for a few seconds before the boost ramps back up. I have noticed this most of the time while playing Destiny 2 and going from a tunnel area with no npcs to an area that has a bunch. This happens both while using the preset performance options in the current version of the AMD software and applying a manual OC (it happens less with the OC). Currently I have the OC set to: min freq. of 2400, max 2600, voltage 1100, power limit at +15%, Zero rpm fans disabled, radeon chill disabled, radeon boost disabled.
Full system setup is:
Asus x570 prime pro
3800x with PBO enabled
32GB gskill trident z 3200mhz xmp
Corsair RM750i psu
All storage is on samsung 970 evo plus NVME drives with no raid set up
CPU is water cooled and holds 50-60° and boosting to 4.6ghz and only 25-30% load. GPU temp is only 65-70° and power draw is 300w or less the whole time.
Has anyone else run in to this yet or figured out the reasoning?
Don't @ me that I should have used 140mm fans.... only had black 120s in stock when the build was done
Im having the same issue with mine. Randomly drops to 500mhz in game...
I am seeing it far far less when I can manage to keep the GPU load up and manually OC but it feels like as soon as it hits an easier part to run the system is going in to a power saving mode instead of keeping the boost up.
Its strange tho. Im only having this issue while playing black ops cold war and modern warfare. On games like CSGO, Apex Legends, and NFS heat the clocks are stable at 2400-2500mhz...
I’m going to try tonight to super sample my resolution to 6880x2880 and see if that will force the card to hold boost clocks and load. I’ll let you know what I find... maybe some of these titles are just too easy for it to run?
@Staarfox wrote:Its strange tho. Im only having this issue while playing black ops cold war and modern warfare. On games like CSGO, Apex Legends, and NFS heat the clocks are stable at 2400-2500mhz...
Otays, was just testing for the last 25-30 min and it turns out the answer is crank it up. I set up initially with 200% resolution scale but this was a bit of an overshoot at a whopping total of 19.9 million pixels (just under 8k resolution for the total render). It was definitely playable and held solid around 45 FPS and the boost clocks stayed pretty consistent at roughly 2580 MHz. I dialed back the resolution scale to 150% so it was rendering at 5160x2160 and that was the sweet spot with boosts holding solid at 2540-2560 and the framerate never dipping below 85 FPS. I guess the solution to our problems is crank everything to 11
I have those issues on witcher 3 and Il2-sturmovik:Great Battles but when i alt+TAB it runs back to normal boosting and keep running on full potential.
Other problem is when i exit games and back to desktop i have stutters on mouse cursor moving like my card is producing no clocks at all. When im moving super fast circles with mouse it goes smoothly but slow moving mouse cursor is stuttering. Only restart helps. I found that when i run a movie clip its goes back to normal but when i close app, mouse on idle stutter again.
My spec:
Aorus Z390 Pro, Intel I9 9900KF; G-skill trident Z 2x16 Gb 3600mhz, Sapphire 6800xt, Corsair RM750.
I managed to find a different solution. It seems that there is a issue with AMDs software. I tried running a slight OC on my card through ASUS GPU TWEAK II - which resulted in stable clocks during the whole day. The clocks stayed stable at more aggressiv overclocks aswell and without changing any ingame settings to put more load on the card.
Give it a try :)
resets frequencies due to bad components and bad software or even because of temperature. There are no other problems
i suspect a bug in the game - or either a driver bug
new Arch always have children illnesses
I have the 6800 and mine is doing the same thing.
The clocks dynamically change based on temps/power consumption etc. also depends on the instructions, i found out that on some games, or games that use RT the clocks drop a lot but if you try to force them higher (by setting the minimum to more) you get lower performance, my config rn is 500mhz-2560mhz. 1125mv. fast timing ram at 2150mhz, power limit to +15% and a custom fan speed so it ramps it up to a 100% when over 95ºC. this far I've got a boost in every game you can think of, the clock speeds stays at 2300-2400 and jumps to over 2550 if the scene is less demanding
I am very aware of how boost clocks function... The behavior was repeatable and not tied to temp or power. My OC is stable and I currently have the #6 spot on timespy with my cpu/gpu combo. What is happening is not the GPU dropping to base clock but the core will drop to 500MHz or less for a short time and then come right back up to the normal 2556 it holds while boosted. I am suspecting that there is a driver issue happening because if the load on the card is more demanding (like rendering at 150-200% resolution in some games) this does not happen. I have not had this happen at all with cyberpunk on 3440x1440 ultra with native rendering. The problem is isolated to less demanding titles. Temps when fully heat soaked are 60-66 on the core and 80-90 on the junction.
Ok so;
CPU: R9 5900X (Noctua NHD15)
MB: ASUS TUF x570 Plus WiFi
GPU: Gigabyte RX6800XT 16gb OC
Memory: 32gb Corsair Vengeance 3200mhz
Cooling: all fans, 7 Noctua 3000rpm fans in total, cooling is not an issue.
PSU: ASUS ROG 850w platinum
I’ll start off by saying this is my SECOND RX6800XT, having had a replacement due to a faulty card. I am now in the process of returning this card because the clock is like a yo yo. My FPS in games is way below what’s expected of the card and my rig (I game at 1440p and can barely hit average 90fps in some titles that I should be hitting 140+) it’s inconsistent though and sometimes it will work perfectly at my monitors refresh rate of 165. It always gets worse over time despite no thermal throttling taking place (temps rarely pass beyond 55c under load).
Crashes etc have happened, I know I it’s nothing to do with my setup, my PSU is fine and I use two 8pin cables. I have minor overclocks, less than you OP and my custom settings help improve stability to an extent.
(2300-2500mhz, max voltage, max out memory and change to fast timings, +15% power, custom fan curve that tops out at 50% (more on this later))
so I can be playing games and I’ll also run GPU-Z. I’ll notice clock speeds start going haywire after an hour or so but temps aren’t even at 50c. I have all fans set to static 900rpm which really keeps temps down for me.
for some reason the fans on my GPU will start increasing despite temps only reading 60 ish on hot spot and despite me having set a custom fan curve.
I’ll see my FPS just slowly die over time. Running benchmarks such as fur mark is completely fine for some reason, any game I play ends up with poor performance. (All settings on windows are correct, all drivers up to date, windows up to date, chipset and bios all up to date).
this is the same problem my first GPU suffered and I got a replacement. I’ll be getting a replacement for this one too. Pop in an old GPU (1080)and what do you know?? It works better than the 6800xt.
convinced a lot of these cards are busted from the get go.
The main offender where it’s most obvious is Baldurs Gate 3, this is run in;
fullscreen (1440@120, set my monitor to 120)
max settings
all Effects etc
it also occurs on;
destiny 2 (all max settings, fullscreen)
remnant from the ashes (via gamepass)
state of decay 2 (gamepass)
final fantasy XIV (full screen max settings, GPU barely has to work)
WoW (frames tend to settle around 70 in the end but GPU barely has to work)
Valheim (maxed out, starts at a nice 165 and slowly degrades to 50-60fps)
COD modern warfare (stutters but stays 100+ unless heavy action on screen)
Baldur's Gate seems to run fine in this video i found with a 6800 XT. (1) Baldur's Gate III High Settings 1440p RX 6800 XT Benchmark Test - YouTube
Perhaps it is a local system issue. You may want to try a selective start-up so that you can rule out any non windows background apps causing your performance issues and run the game only with nothing else running to see if that helps.
You may also want to consider a clean Windows installation perhaps.
To reboot in a selective startup:
Press WINDOWS-R on the keyboard to bring up the Run window, enter MSCONFIG and click OK.
Under the General tab, click "Normal startup" if it is not selected already.
Under the Services tab, put a check on "Hide All Microsoft Services" and then click "Disable All."
Under the Startup tab, these will still be disabled in Windows 10 in the Task Manager.
Click OK on the MSCONFIG window, and then click Restart when prompted.
Has your 3rd GPU acted any different? Or have you resolved your issue?
I am currently in somewhat of the same position where I RMA'ed my first GPU and seem to have the same random down clocking problems.
First of all I had to down clock my GPU below the "gaming frequency " advertised by AMD (2015Mhz), which helped a little.
But the issue still persited.
I also tested Matts version of testing for a local system issue and found that the downclocks occur when men ues of loading screens get rendered.
But according to my tests ( in Monsterhunter:World) with and without non system services no downclocking occurs while stutters are present ( as viewed in HwInfo64's min core clock row). In these tests I only followed the gameplay and din not tab out or open any option's menu.
For me this would also sadly sound like another RMA because I don't have access to system-wide debugging tools ( to maybe see which buffers are overrun or which device is responsible for not transferring the rendering data)
Hey Matt,
I did already try these 2 cases I fact my "overclocking profil" is attached as a picture.
I might have joined the wrong discussion because my testing yesterday and reveals that downclocking does not occur (with this overclocking profil and whilst sole gaming[e.g. opening no menues, maps,...) .
Still for me stutter occur which seem to be more frequent at the beginning of "matches" or at the end which leads me to believe it has something to do with loading assets for the first time or caching them.
I am going to try to load from different harddrives , more page file size and with XMP on or off.
EDIT: none of the variations I was able to do did affect the appearance of stutters (see attached video at timestamp : 0:07s-0:09s)
Pic 1: wattman profile ( downclocking below "gaming frequency for less stutters)
Pic 2: HW info for a 9:30 gamesession in Monsterhunter :World -> no downclocking below minimum coreclock
Video 1 ; displaying type of stutter randomly occurring
In the comments of my thread about this I posted a solution that completely fixed the problem for me. Im not the one who made it, I can’t find the original site. But it fully fixed the problem. Just remember to set a minimum clock in Radeon settings, I use 2.1ghz.
https://community.amd.com/t5/graphics/6800xt-clock-speed-problems-downclocking/m-p/496305#M80665
Thanks for your insight, sadly even disabling the Sleep modes I found out donwnclocking is still occurring at my end ( in contrast to my previous belief) . Even though it is only for one frame the GPU clock drops to 32 MHz ( from 1950 Mhz avg.).
I didn't find this with HWINFO, but looking at my recording ( see https://imgur.com/a/Qdeo7Xe 0:05 for one frame you can see the first reading in "GPU2" -category drop for around 1 or 2 frames) .
I also tried with another monitor and this time via hdmi and the same thing occurs when capped to 60Fps ( https://imgur.com/a/Yho3VS6)
This still seems weird, since I more or less feels like a something power related, but when I changed between my 3 PSUs I couldn't find a difference in performance / stutters occurring during gaming.
@PtotheG
for me this looks like a "windows system interrupt" - so you should do a clean install of windows and only install programs needed and test again.
also this can happen with "unstable" ram-oc (XMP is also oc) - so also test with DDR4 standards (DDR4-2400 etc)
also it can rarely happen that a GPU OC (even Factory OC) causes this - so try to use AMD standard clocks for 6800XT