I have learned, after much hunting and very little info found, that there’s a known but little advertised issue with DX12 and Intel 13th and 14th generation CPUs.
I had been getting weird, out of video memory crashes on a variety of games defaulting to using DX12, some with no ability to even get to the Start screen to go into settings and change to DX11.
It turns out that on systems like mine (I’m running a 14900KF not overclocked yet), you have to go through the following steps to resolve the problem:
A) Install the Intel Extreme Tuning Utility from their website. Latest version for your OS is easily discoverable (I’m running Windows 11 Pro).
B) You will have to go into your BIOS to turn Undervoltage Protection on (Enabled) before you can install or use the utility. You will get a warning about this on first attempt to install. Do so. Carefully.
C) Once running the Utility, simply Manually set the P-Core speed to 52-54. It will likely be higher by default. You’re literally slowing down your CPU, but at this level of performance, that doesn’t matter. At all. Running my CPU at 6GHz is easy on my system with my cooling solution…. but it won’t run DX12 games. This is the fault of DX12.
D) Keep the Intel XTU on your system. You can easily change this setting back to default for everything else, in case that tiny nub of extra performance is useful to you outside DX12 games.
Editorial: this is a ridiculous state of affairs, and must be remedied by the DX12 Team at Microsoft. I hope that will be done before the launch of MS2024, if the plan is to force everyone to DX12. Otherwise, it could easily be seen as an intentional effort by Microsoft to lock Intel CPUs out of their OS and Games markets, which I imagine could lead to some pretty serious business conflict.
Hope this helps some other poor soul out there. I have literally spent months diagnosing this problem, with no solid info on the web available to help.
Interesting. I am sure this will prove useful for anyone else experiencing this symptom. In my case, I have 14900KF in my Alienware system with a 4090. It runs perfectly with no such issues. In fact it is an absolute beast and has endless horsepower.
I don´t see the relation between OOM and CPU, nor even between DX12 and CPU as DX12 indeed is expected to reduce CPU usage… But anyway, if you are facing an unstability problem for whatever reason, why don´t you reduce the clock of E-cores instead? That way you won´t sacrify performance in games and still reduce the overall CPU usage a bit.
I would try to change the CUDA memory usage policy at Nvidia´s panel first. That setting has an option to allow usage of RAM to help in case of high VRAM usage scenarios, which is indeed intended to prevent the OOM situations on the video card like the one you describe. DX12 uses more VRAM that DX11 so that makes more sense to me that playing around with CPU. You may want to explore that option before killing your CPU performance.
Interesting, but the whole point is that the 14900 series and 13900 series will not run DX12, period, without downclocking their P-Cores, no matter how much VRAM or system RAM you have installed.
It’s not a question of fiddling with Nvidia tweaks, believe me, I’ve tried them all, including using Nvidia Profile Inspector. I have literally over 10 years of experience with Nvidia GPUs. It’s not a “wrong setting” there.
I think at this point there is an actual coverup going on around the real problem. I have MSFS2020 running fine on my system, with very high, knowledgeable settings, using DX11. DX12? Not even going to experiment.
People can hide their head in the sand around this issue, but it’s going to bite MS hard when it comes to insisting on DX12 for launch of 2024. Watch for what happens with the upcoming MSI Claw when it launches.
Unless the DX12 team comes up with a fix by then. Fingers crossed.
II dunno for sure, but I think there are lots of folks running 13th and 14th gen processors in the sim in DX12 without encountering the issues you are having. Perhaps that’s why you found so little information on the “issue”.
And I’m not negating your experience in saying that. Rather, if you were to take all of the permutations of CPU, RAM, storage, BIOS settings, motherboard chipset, drivers, GPU, FS settings, I think it’s inevitable that multiple people will have your experience. Are there other variables you can change without breaking the bank too much? An inexpensive set of RAM sticks, maybe at a lower speed?
I took a more extreme approach and did an OS rebuild, twice, because I was having wretched instability. I’ve taken the approach of starting from nothing and working up to the simplest configuration that will work.
For example, I decided my machine wasn’t going to also be a Docker development machine, that I wouldn’t run WSL on it, and so I could disable certain virtualization-related settings. I install my NVidia drivers as clean installs every time, and skip installing GeForce Experience. Not saying you need to do all these things, but that’s just the path I followed - to not have the machine be everything I could possibly want it to do and just make it work with FS. And then from there, you can start layering things on slowly, until you encounter trouble.
That problem exists and it is not correct only to blame Microsoft and your statement about Raptor Lake and RL-R will not run dx12 is not correct at all and not helping anyone. Better to be helpful to people that has the problem.
You also need to give more info:
what is your ram? 6000 or 5600 etc. have you tried 2 sticks 5200?
what is your cooling? Aio? Water? What size?
did you do all bios updates? Or your mobo is using out of the box bios?
when your dx12 game crashed, you started it and cpu was 100%?
I don´t mean there´s no issue, only that I don´t see the relation between the described factors. I found some posts from other users having the same situation like you, however all seem to be lost in the sea trying potential fixes that basically mean reducing CPU load in one way or the other. That´s why I think the issue may not be directly linked to DX12 itself but linked to system stability in general.
14th gen are brand new on the market. Are you using a BIOS dated after 14900k release to market (was around Oct 2023 if I´m not wrong), and most important that already includes any support/stability improvements for that new CPU series in the release notes? Unfortunatelly such BIOS versions releases are not always as fast as the HW release to market itself.
I absoloutely agree. This is the most nonsense post I think I have ever witnessed. How did this person come to this absurd conclusion?
13th & 14th Gen are not compatible with DX12?? It boggles the mind. So Raptor Lake and DX12 don’t like each other as they’re fighting over a few 100MHZ P-Core. WTF. Sounds like a him problem to me.
“if the plan is to force everyone to DX12. Otherwise, it could easily be seen as an intentional effort by Microsoft to lock Intel CPUs out of their OS and Games markets, which I imagine could lead to some pretty serious business conflict”.
Yep, latest BIOS release for my ASUS Z790 Hero mobo (1801). Could be there are more improvements on the way, but that motherboard has been out awhile. CPU and RAM confirmed on QVL. RAM: Corsair 6000MT, run at XMP 1 on the mobo. GPU: ASUS Tuf Gaming 4080 16GB OC Edition, not overclocked yet (none of this really needs overclocking, IMHO). Lots of NVMe memory on the motherboard. BIOS set to Optimized Default with just XMP 1 setting on to bring the RAM up from stock 4800 to 6000MHz. CPU will do 6 GHz without any heat issues in my set up, plenty of cooling including Coolermaster MP240 PL Liquid Flux (I could bump that up but my temps as indicated with HWINFO64, AIDAExtreme and ASUS’ Armoury Crate remain quite cool, in the 40s (oC), with CPU Package running at 75-77oC, which is standard. Changing fan speeds makes no difference in whether things run better or not, the fan curves have all been optimized (6 chassis fans in all, 2 fans which never run on the 4080 ). Oh – latest NVidia driver for this card, all GEForce interference turned OFF.
The various tweaks I’ve gathered from people around the 'net are:
Try enabling Legacy Gaming Mode (what? for a DX12 game using NVidia DLSS?)
Try setting the AC Line Load on the CPU to 4, IA voltage to 1.02 (not trying that… yet, very basic overclock).
Best tip and what I’m using now: install Intel XTU (latest version for my Windows 11 Pro Preview (Beta) Build OS, then use it to downclock the P-Cores from 57, what it defaults at, to 54. A hack, but it works, for all DX12 games. Immortals of Aveum is an excellent test case. Without downclocking, it won’t even launch, and causes hard lockups; with downclocking to 54 (5400MHz, basically, from 5700MHz, default), it runs very smoothly at ultra settings and 2560x1440.
That’s where I am now.
The connection I’ve seen is simply that NOTHING I’ve used that doesn’t require DX12 (sorry for double-negative) has caused any crashes, hiccups or lockups on my system. And MSFS2020 using DX11 runs on 2560x1440 with Ultra settings and NVidia DLSS as smooth as glass. Turn on DX12 and BOOM!
So at this point this is less an issue of some particular confirmed microcode interaction or timing issue (I’ve had a friend at Epic Games who works on the Unreal 5 engine comment to me privately on this, and he suggests the possibility of a race condition induced by the 14900KF at non-overclocked speed <it’s worse if I try overclocking it>, but that’s just an extremely educated guess on his part), and more deduction by process of elimination, Sherlock Holmes style.
The reality is: this is an issue. A known issue, and one that needs to get resolved.
I have informed people elsewhere; it’s up to them to figure it out. It’s also up to MS and Intel to figure it out. Good luck.
I’ve built 5 computers since, the 12900K was released. A 12900K, A 2nd 12900K, 13900K, a 2nd 13900K and now a 13900KS. All with 3090 TI’s and none of them have an issue like you’re having. I would try running it in a different Windows User just to see if it’s the Windows user profile that’s corrupted, or running a Windows In-Place upgrade/repair using the Windows ISO, mounting it to wherever on the pc, starting it and go through the process of an upgrade/reinstall that won’t affect anything on the system besides the OS. Or try all the steps in between those two options. But most of the time issues like that are with the system itself, worth a shot anyway. DX12 isn’t going to get “fixed” anytime soon. Especially if you can’t get Microsoft to replicate the issue themselves.
DX12 out of memory message can occur with CPU instability, usually during transient loads such as loading screens. If you’re getting this regularly, you could have an unstable CPU due to too low voltage, even at stock. I got these all the time loading BF2042 when overclocking the 13900K too far, usually accompanied by a nice WHEA parity error on a CPU thread. Increasing the voltage or lowering the frequency elimated this issue.
This is why OP “fixed” it by downclocking the CPU frequency. CPU parity errors, even when corrected, can throw this error.
Could well be a low voltage thing, I’ll look into that. All my testing, benchmarking hasn’t demonstrated any issues, I’m not overclocked, and I’ve gone thru full sfc/scannow and DISM /Online /Cleanup-Image /RestoreHealth (pardon syntax errors) cycles with no issues detected (along with chkdsk on any related drives, including swap).
I’m loathe to go back to repair/restore, as I’m on Insider Preview (Beta) build now, and I expect issues trying to restore from my W11 Home DVD if I try it.
Will live with where I’m at for now, thanks for the suggestions though. I’m more concerned about how this might affect others than what it does in my system at this point. I’ve got it under control. I won’t worry about anyone else from here on out; I’ve provided my info, it’s there for them to use if needed. Over and Out.