I do have an unusual set up in that I’ve put together multiple SSDs (from previous system) an HD (WD4004EX) and a few NVMes on the mobo and M.2 storage expansion card that came with the mobo. Could be there’s some sort of PCIe bus contention going on there that adds to the issue, though on the face of it that doesn’t seem likely. No identifiable conflicts in Device Manager, either.
I have seen your Asus BIOS info (I own another Asus Mobo too) and latest version clearly supports 14th Gen CPUs. Support is there since 1402, so BIOS is fine.
I still think you may be facing a system stability issue here, as reducing CPU clocks clearly mitigates the problem. You may want to try XMP II instead of XMP I. I have also a 6400 MHz Corsair RAM with an i9-12900ks and XMP I is not stable at all and has never been stable in any of my setups during the past years, resulting even in BSODs. If XMP II is not available you may have Asus OptiMem II available instead and you can also use it.
Make a copy of your current settings just in case, then load BIOS default settings (pressing F5), change to XMP II and select the correct RAM profile. Asus settings are always well optimized and most settings can be left on Auto or to the values proposed by Asus defaults (which is most cases will be also “Auto”). I never had to manually tune my BIOS in years by using this method and I always had stable systems.
About current market practices and known technical issues, well… There´s still a chance of that, sure. Windows itself is not handling P and E cores properly and HAGS is not yet stable at all, resulting in many problems when running DX12 or games that use frame generation like MSFS. The arrival of 14th Gen just put more stress on systems on top of that. Problem may not be the CPU itself but the whole current ecosystem. HW seems to evolve faster than the rest, while it still needs the rest to perform optimally.
Cheers
1 Like
Yes this all sounds spot on to me. I kind of feel an obligation given the newness of the hardware to keep tweaking to discover and publish issues I find; will continue to do so. Trying XMP II will be my next test at some point.
I’ve tried the VRM load line fiddling, legacy game mode enabled…. makes no difference. Only thing that helps decisively is using XTU as primitively as possible, just to downclock the P-Cores. It is all quite odd to be sure! 
EDIT: So after a little more experimentation… nah, I’m just going to stick with what’s working now. Even got ASUS “AI Optimized” Overclocking working fine, also GPU III Tweaked my 4080, all tested fine, just to be safe then downclocked the P-Cores with XTU to 54, Flew around in my new R66 over Sydney (great add-on by Cowan, btw! All their chopper simulations are insta-buy for me at this point!), then exited to main menu and… lockup. End-tasked MS Flight Simulator, restarted and… HANG. Shut down power, went back into BIOS, Fixed All the Things (back to normal, not overclocked, but XMP II this time to be a wild man… and it worked!), went back to GPU Tweak III, Fixed All the Things (back to Default Mode after doing its full OC Scan), and everything’s fine again and I’ll just XTU to ironically DOWNclock my CPU when I need to play a DX12 game.
It’s all so silly! I feel like I’m Alice in Wonderland here, all this engineering overkill for what, a few MHz extra? With mysterious instabilities introduced and LOUDER fans as a result, and no discernible change in performance? None of it is really worth the hassle, as I should have guessed at the outset. And what’s the point of overclocking things if you then have to downspeed them to get them to run?! O_o → it’s just nuts.
So my OP stands: use Intel’s XTU in this upside down world to slow down your P-Cores if you’re having any trouble with running games under DX-12, one and done. Period. Then Revert to Normal after a playsession. Play with all this AI overtuning and overclocking and overdoing stuff if you like numbers. Like all the pretty fan colors, it is, er, engineering m--------. I’ll leave it at that.
Stock i9-14900k here, with 64GB 6800 MHz RAM, and RTX4090 OC. No such issue with DX12 here. it was running normally for me. I have since switched to DX11 again on the advice of YT setup for VR, but I had DX12 running fine on my PC for a few weeks. I don’t think the issue is as basic as a problem with 13900 or 14900 cpus, there is something else happening.
2 Likes
I will refer you to this:
Not smoking anything. Real issue. Feel free to deny. “Works on my machine” is something no QA tester would even say as a response, but of course we’re users, not QA here. So I understand your and others’ refusal to acknowledge the reality of this problem.
Easily solvable, anyways; just downclock your 14900K/13900K 100MHz from its default with XTU, problems with DX12 solved. The fact that such a slight change stops all lockups and crashes and failures to launch is indicative, to me, that there is a definite timing issue present and I think it’s probably in the Intel chips, as it’s been reported both in games using the UE5 engine and those not, and with benchmarks like Cinebench r/23 and r/24, and on different Z790 motherboards, both by ASUS and MSI. So that would tend to narrow it down either to the LGA1700 Chipset (most likely) or the CPU itself. OTOH it doesn’t happen (for me) with DX11 at all, even at overclocking, which points to DX12 as the culprit, my point here in the first place. Interaction of DX12 with LGA1700/14th-gen Intel CPUs. I’ve done my volunteer work isoiating this; it’s up to MS/Intel to pick up the ball and run with it, or not, now.
Could be an intrinsic, unsolvable problem (except with the XTU hack) with this generation; no matter, Meteor Lake, Arrow Lake and Panther Lake are right around the corner. Meteor Lake will be spectacular on laptops and handhelds in particular (the 155H being used, for instance, in the upcoming MSI Claw). The NPU component of Intel’s next-gen CPUs is going to be a game-changer, not necessarily reflect in primitive benchmarks, of this I am convinced.
OK! Got a fix here, just in from Intel. On their Community Forum:
“To help you further, kindly install/run and perform the following items below:
- Intel Processor Diagnostic Tool - to check the functionality of all CPU cores.
- Intel Extreme Tuning Utility (Intel XTU) - run AVX2 test, if AVX2 test failure is seen.
- Go to BIOS settings:
-
Advanced (F7)
-
AI Tweaker > SVID Behavior: Change to “Intel’s Fail Safe”
-
Reboot the system and run the XTU test again to check if the AVX2 test can pass.
-
Run games to see if the issue happens again.”
Changed SVID Behavior in AMI BIOS 1801 on my ASUS mobo (see above) to Intel’s Fail Safe, and all DX12-induced crashing has ceased. Turns out using the XTU CPU stress test with AVX2 instructions, the tempt will fail unless SVID Behavior is changed from Auto to Intel Fail Safe.
Hope this is use full if anyone else runs into this issue.
Sorry, but this is not technically correct. SVID Behavior infuences the AC LL, which goes in to Vcore calculation. A higher AC LL = higher Vcore. Changing SVID Behavior to “Intel fail safe” forces AC LL to the highest possible that Intel defines, and on most chips would be a massive overvolt. AVX2 stable typically requires a higher Vcore compared to SSE instructions.
Therefore, my previous post still stands: this error is a sign of an unstable chip and caused by too low Vcore. However, I would not set SVID Behavior to Intel fail-safe. Manually increasing AC LL until stable will ensure you’re minimizing Vcore while maintaining stability.
3 Likes
Use the Intel provided CPU-Check-Tool. If everything is ok, I would not recommend to play around with the suggestions here.
1 Like
I’ve used the Intel Processor Diagnostic Utility. Everything scores as Passed.
@ncbartschi I will agree with you my 14900K, like all others, is unstable. I’ve reverted back to just not overclocking it, not using XMP, and downclocking it to 5400MHz using Intel Extreme Utility.
The Intel certification and advertising is apparently false, or not relevant for use of this chip with ASUS ROG Maximus Z790 Hero motherboards. My loss. Sorry to mislead. I’ll just live with it; it’s plenty fast enough “stock.” Just overpriced, clearly.
1 Like
Pretty sure that with minimal tuning of the correct parameters in your BIOS you would be able to get your 14900K rock stable. Just needs some reading… As ncbartschi implied, you issue is most probably related to a too aggressive Load Line Calibration Setup and/or wrong Vcore settings. If I am not entirely mistaken, there were some discussion about this issue especially on ASUS mobos when the 14000 series hit the market, but maybe I am mixing things here.
1 Like
Don’t use the Intel XTU app but set XMP in the BIOS.
Set whatever CPU speed you want to run in the BIOS.
Don’t use XTU.
3 Likes
Agreed - the combo of app and BIOS makes it more difficult to sort out which is doing what. I’d use XTU to measure, and perhaps make fleeting changes for that purpose alone, and then bake them into the BIOS.
2 Likes
It is probably more a motherboard issue than the CPU issue.
I have i9-13900k, 64 gb DDR5, 4090 system and don’t have those settings or software installed as the OP wrote about. The issue w/ DX12, in MSFS, is with Frame Generation - it’s bugged since SU14 where the screen shakes and shutters.
1 Like
I hear that - motherboard drivers, motherboard firmware, motherboard manufacturing defects… I also wouldn’t rule out a bad lot of CPU dies.
1 Like
I’m running a 14900K, stock clocks, XMP, DX12, a 4090 and TAA and frame gen and not had one single error, hang or crash. I’m not negating your situation, which sucks, but your first sentence is demonstrably false.
4 Likes
Phil, I think it’s the stock clocks that’s the thing. To get my system to “stock” yeah I can load optimized defaults, set XMP I and call it a day, I guess. But then you’re basically not using the chip you paid for. Why get an unlockable CPU when you are just running it stock?
1 Like
The K CPUs are better than the non-K CPUs.
The hope is that they last longer or have less troubles.
the K cpu’s have higher stock clocks. That’s what you’re paying for really. intel already boosted the clocks to about as high as they’ll go (to keep up with AMD’s performance). These days you’ll generally see very little gains from manually overclocking, unlike a decade ago when there was plenty of ‘free’ performance on the table.
1 Like
Alienware/Dell are notorius for intentionally limiting the performance of the CPU and GPU to extreme levels. So if you buy a 14th gen i9 on their systems would be like running an 11th gen i9 on other systems
1 Like