Yes.
In my experience, current drivers are very roughly 10% faster (in GPU limited scenarios) than 22.5.1 in DX11 games.
The RX 6000 series, especially with post 22.5.1 drivers is very hit or miss in this game, in VR.
AMD RX 6000 series driver issues & workarounds
Yes.
In my experience, current drivers are very roughly 10% faster (in GPU limited scenarios) than 22.5.1 in DX11 games.
The RX 6000 series, especially with post 22.5.1 drivers is very hit or miss in this game, in VR.
So basically its -avoid- AMD if you are playing ED in VR. Sad.
I dont know how many people are going to throw a considerable amount of money into something that is "hit and miss".
How widespread is this? Is there any correlation between CPU's etc?
Last edited:
The game works very well in flat screen mode. I have an all AMD system with an RX 6800XT and Radeon software v. 22.11.2, and I haven't noticed any graphics glitches with it. Performance is very good, too.
I play a lot of VR, and apparently you can get issues with game craches at load, warped planets etc. Not great.
VR issues seem to be widespread. However, I can't test them as the systems I have with AMD GPUs are running Windows Server and my WMR headset doesn't work these OSes.
Some people still report non-VR issues, but I cannot duplicate these and strongly suspect most come down to user error.
No, nothing vaguely reliable, nor would I expect such a correlation.
VR + some Radeon RX 6000 series part = common issues.
If your use case is VR + Elite: Dangerous Odyssey, go with an NVIDIA GPU.
I've just discovered that most remaining fps issues for me were resolved by upgrading to 5800 x3d CPU
You were most likely CPU limited then, and not a specific 6000 card issue.
I rock the 5800x3D as well, and it also kicked new life into my trusty old 2070. But VR performance still is not quite there... and most certainly not for anything newer than my Rift S.
Surprisingly not. My previous Ryzen 3700X was never taxed more than 5-10% - I think its just the added cache architecture of the new chip that solves almost everything for my setup.
The game will only fully load 2-3 threads at the best of times (partially loading several more, depending on how many worker threads have been set) and a less efficient memory subsystem (which is impossible to separate from the CPU when all the cache and the memory controllers are on the CPU) causing the CPU to waste huge numbers of cycles waiting to access system memory are exactly what made it so CPU limited on your 3700X.
A good rule of thumb is that if the GPU isn't pegged at essentially maximum utilization the entire time, it's a CPU and/or memory performance limitation.
Just switched from NVidia to Radeon 6900 XT ... fresh install of everything (OS and Drivers) and all seems fine except the orbit lines and gravity lines.
Searching for a workaround shows this has been a problem for a while now. Are there any options to improve these yet? Running on normal monitor (not VR).
Thanks!
Searching for a workaround shows this has been a problem for a while now. Are there any options to improve these yet? Running on normal monitor (not VR).
Thanks!
This is typical. VR is where issues persist.
Outside of huge amounts of super sampling, no.
Since this thread is still active, I figured I'll try here to see if there is any hope.
I'm running the Valve Index VR with the latest Steam VR version.
Radeon RX 6900XT driver version 22.11.2
Latest version of Odyssey.
Game runs just fine in space and in stations. Even with all settings turned up. Beautiful.
Some planets are borked. I managed to land at a station on a borked planet, and upon entering the docking bay the whole PC died like there was a power outage. Upon resetting and logging in again, it repeats the power shutdown. I know I can load pancake mode and get out of the station, non-vr runs 100% fine. But logging in on a planet in vr causes this power shutdown.
Is there a workaround for this other than driver 22.5.1?
I really don't want to downgrade because the other games I play really benefit from the newer driver and even Elite looks much better under the new driver. Downgrading to 22.5.1 causes a very noticeable drop in the lighting/shadow quality.
I have tried the floating doubles fix and it did nothing. I have not attempted the shader fix as it did not seem to apply. Help/thoughts?
I'm running the Valve Index VR with the latest Steam VR version.
Radeon RX 6900XT driver version 22.11.2
Latest version of Odyssey.
Game runs just fine in space and in stations. Even with all settings turned up. Beautiful.
Some planets are borked. I managed to land at a station on a borked planet, and upon entering the docking bay the whole PC died like there was a power outage. Upon resetting and logging in again, it repeats the power shutdown. I know I can load pancake mode and get out of the station, non-vr runs 100% fine. But logging in on a planet in vr causes this power shutdown.
Is there a workaround for this other than driver 22.5.1?
I really don't want to downgrade because the other games I play really benefit from the newer driver and even Elite looks much better under the new driver. Downgrading to 22.5.1 causes a very noticeable drop in the lighting/shadow quality.
I have tried the floating doubles fix and it did nothing. I have not attempted the shader fix as it did not seem to apply. Help/thoughts?
Your computer is experiencing a hardware fault. Most likely your power supply can't handle the strain, but it could also be overvolting, undervolting, too high temperature or any other hardware issue. This can be caused by incorrect or badly configured driver settings, particularly when it comes to overclocking.
Upon reading this, your first reaction will most likely be defensive, and you'll want to reply with something like "it never happened before", or "it doesn't happen in other games!", or as you already stated "it doesn't happen in pancake mode!".
But this is squarely in the realm of hardware issue and you should investigate and consider removing any overclocking on the CPU, Memory or GPU.
This is not typical, and not something software should even be able to cause. 'Normal' failure behavior is a crash to menu or desktop, with or without some sort of error reported by the game, not a system shutdown.
Maybe the capacity of the PSU was exceeded, or the PSU is faulty? Or the GPU has too high overclock?
Thanks all for taking the time to reply.
So at this point I've spent a LOT of time (days) testing and here is what I've gotten and why I feel it's some sort of driver+Elite issue.
First, this is not a new system or build. I just decided I needed to make the driver update move for other applications/games.
I'm running an AMD 3900X CPU, 12 cores. And a 1200W Thermaltake power supply.
I've tested the PP as much as I am capable and it seems that both rails, all of the voltage limiters, and the thermal limiters are all responding correctly.
Tested the CPU and GPU both with OC settings and non-OC. Manual and in-built OC settings. Under multiple game and driver configurations, and with 3rd party software.
(The CPU does throw a overvoltage trip on core 7 when running Folding@Home using more than 12 threads after about 30 minutes of work. This might be a processor issue but unverified.)
During testing, none of the monitoring software I have or any of the loggers shows or records anything out of the usual. But the error does seem to be a overvolt threshold trip triggered by the CPU. But whatever it is, nothing is logged out of place on any of the loggers across any core of the GPU or CPU. I've also live monitored everything and nothing even comes close to hitting a limit.
I have tested both 22.5.1 and 22.11.2 drivers. All crash issues are only on the 22.11.2 driver. For testing purposes I'm on a borked planet, docked in an underground hanger at a planetary port under thargoid attack.
Using the 22.5.1 driver I can run the game under maximum settings with my preferred OC settings, in VR with Steam set to 100% resolution at 144hz and the ingame HMD at 1.0 and the upscaler set to CAS at 1.25. No issues other than running 45-60 FPS while at port.
Using the 22.11.2 driver, the system crashes under those settings upon game load/login of the ship/port 17 seconds after login. It seems to be loading the last planetary details and shaders. All crashes using the 22.11.2 are at the same point.
Same for OC and non-OC system settings of any configuration.
Same regardless of Steam resolution setting (I've done 20%-150%).
Same with in-game settings at VR low or VR Ultra.
(During all of this testing I did finally get a proper game crash and error reporter that I submitted to FDev. 1 out of over 30 cycles.)
(I've reverted back to the 22.5.1 multiple times throughout all of this testing and it's always fine and stable.)
But here is the kicker... Late yesterday after I made my original post I found a stable setup...
System can be OC or not, no effect. But I'm currently using my preferred OC settings. Steam rez at 100% 144hz. Game settings at maximum with HMD at 1.25 and upscaler on FSR (I'm using balanced for visual preference right now, but have tried all four settings). Game runs fine. Getting about 75FPS around the planetary port. The planet is borked and I can do the relog near the planet surface to correct it. I can fly around and do everything in the AX CZ around the port, dock, launch, land on the surface with no issues.
But the FSR is less attractive and readable than the CAS at 1.25.
The moment I switch back to CAS or to the 'normal' upscaler setting under game settings, system crash. CAS uses a lot more power than the FSR and drops my frame rates to the 45-60 range in the testing area. The system is definitely under more stress. But everything is well within tested levels. I just can't figure out how/why the CAS triggers it on a planet. It's fine in space and at orbital ports of any variety. Whatever it is with planet tech/lighting/shaders in combination with 22.11.2 causes it on CAS and not on FSR. 22.5.1 is fine in the planetary scenario and 22.11.2 is fine off planet but not on with CAS. 22.11.2 on planet is fine with FSR.
Last edited:
Did you create a bug report for that?
This somewhat resonates with my issue - borked planets and a CTD when trying to start the game while sitting on a borked planet, see https://issues.frontierstore.net/issue-detail/56183 - but with me, it's only a reasonably (we're taking ED, after all) clean CTD.
I have mailed with support (wanted a module blueprint fragment at that time) and tried a few things they suggested, up to a clean re-install of 22.11.2. Nothing really helps with the borked planets - but doing a game file validation seems to force the game into recreating the planetary graphics data, after which the planets look fine for this game session (but will be borked again the next time). I rolled back to 22.5.1 yesterday, and all seems to be fine so far - but I won't be sure until it continues to work today.
This somewhat resonates with my issue - borked planets and a CTD when trying to start the game while sitting on a borked planet, see https://issues.frontierstore.net/issue-detail/56183 - but with me, it's only a reasonably (we're taking ED, after all) clean CTD.
I have mailed with support (wanted a module blueprint fragment at that time) and tried a few things they suggested, up to a clean re-install of 22.11.2. Nothing really helps with the borked planets - but doing a game file validation seems to force the game into recreating the planetary graphics data, after which the planets look fine for this game session (but will be borked again the next time). I rolled back to 22.5.1 yesterday, and all seems to be fine so far - but I won't be sure until it continues to work today.
Over voltage or over temperature? Where are you seeing this?
The board's VRM can only deliver vcore on a per-CCX granularity (and the per-core LDOs are generally bypassed on the desktop platforms), so if one core is being overvolted, every core on the same CCX should be registering this. The board should also not be delivering more voltage than requested by the CPU, so unless you're running a manual offset or running an excessive load-line calibration value, an actual overvoltage scenario implies a serious firmware bug or a hardware defect with the board or CPU.
Post 22.5.1 drivers overhauled the D3D11 driver and increased performance. Performance isn't free and either comes from reducing the actual amount of work being done, or leveraging the hardware available more completely; in the latter case, power demands go up. Elite: Dangerous Odyssey is also an outlier in more ways than one and it's entirely possible that the combination is resulting in transient power draw that's revealing issues with your hardware that other tests miss.
FSR renders at a lower internal resolution, the other super-sampling options do not (unless a value below 1.0x is chosen). It's entirely expected for the FSR options to be less demanding, especially in a scenario where the game will be bottlenecked by something other than the GPU.
Regardless, crashing is not atypical with post-22.5.1 drivers in VR, but a crash that causes the system to power down absolutely is.
What happens if you underclock/undervolt the GPU? You can do this right in Radeon Settings. Performance -> tuning -> custom/manual -> enable GPU tuning -> enable advanced control. Reduce the max frequency to 2000MHz or so and see if it crashes in the same manner.
Also, it bears repeating:
If you are experiencing ANY hardware or software issue that you believe is due to an AMD driver, don't just report it to Frontier, also report it to AMD
If you are experiencing ANY hardware or software issue that you believe is due to an AMD driver, don't just report it to Frontier, also report it to AMD