13900K - Out of Video Memory or BSOD during shader compilation in two UE5 games

H

hu76

Limp Gawd
Joined
Oct 23, 2023
Messages
138
Hi. I am new. So welcome forum. So first my pc:

13900K stock
2x16 GB DDR5 GSKILL 6800
Rtx 4090 Gigabyte Gaming
Aorus Elite Z790 AX
SSD KINGSTON 2TB
Seasonic 1300W PX ATX 3.0 PCIE 5.
win 11

Each new install of nvidia drivers, and launching Remnant 2 ( the same UE5 engine like Lords OTF ), then on first launch ,during shader compilating it throws OUT OF VIDEO MEMORY and it can BSOD.

Next launches are fine.Just on first launch of the game, during first compiling shaders.

The same thing happened on game Lords of The Fallen, on first launch during shader compilating. And next launches are fine.



But like i said it will bsod again if i install new drivers and launch game for first time.



Also. I tested by Karhu Ram test no errors. Cinebench R23 passing. All games are stable beside that. Prime95 Small FT stable no errors.

Should i change ram,cpu or what is your opinion? Thx. Also found many reports with the same hardware, with the same issues in those games during first shader compilating.

This is from Remnant 2 official respons:

https://www.remnantgame.com/en/news/article/11551423

"--Out Of Memory when Loading (Intel 13th generation CPU’s)--​

We have identified an issue on some Intel 13th generation CPU’s where upon startup the game will display a message about being out of video memory or the crash reporter will pop up referencing an issue with decompressing a shader. If you experience this problem, you will likely also see it in other DX12 games.

If your CPU is overclocked, try setting it back to the defaults. If you’re not overclocked or that doesn’t work, try installing Intel Extreme Tuning Utility:

https://www.intel.com/content/www/us/en/download/17881/intel-extreme-tuning-utility-intel-xtu.html and lowering your “Performance Core Ratio” from 55x to 54x."
 
Last edited:
Are you on the latest motherboard BIOS available? Currently F9 for your board

Did you try the suggestion of lowering the performance core ratio as suggested by the Remnant developers?

If you turn off XMP and run the RAM at JEDEC 4800MHz or whatever, does it still happen?
 
Hi. My bios is updated ,nothing helped. Also dont try turn xmp to off. Overall all is stable, but that bsod can happen in both games only in shader compilating ,on first launch after game update or new installed drivers.

Not try lowering the performance core ratio as suggested by the Remnant developers.
 
Furious_Styles said:
If it works fine on 2nd launch just ignore it. Sounds like a UE5 engine problem. Not worth wasting time over.
Click to expand...
yes it works ok now. But thx. To be sure i tested all my games and Battlefield 2042 conquest 128 players. No issues. : P
 
Last edited:
I upgrade drivers to 545.92 for Alan Wake 2. And launched for first time Hogwart Legacy on newest drivers. It crashed three times to dekstop when i launched. Just before shader compilating crashed two times,and one time at shader compile. On fourth try i was able to run game,and now is launching fine everytime. Why i had that three crashes on boot? Is my cpu not stable?
 
hu76 said:
I upgrade drivers to 545.92 for Alan Wake 2. And launched for first time Hogwart Legacy on newest drivers. It crashed three times to dekstop when i launched. Just before shader compilating crashed two times,and one time at shader compile. On fourth try i was able to run game,and now is launching fine everytime. Why i had that three crashes on boot? Is my cpu not stable?
Click to expand...
You should try dropping the cpu speed s notch as suggested :).

hu76 said:
Not try lowering the performance core ratio as suggested by the Remnant developers.
Click to expand...
 
  • Like
Reactions: hu76
like this
Also want to mention that Hogwart Legacy is the same UNREAL ENGINE but ver.4.

I launched other games ( not on unreal engine ): Quake 2 rtx, Battlefield 2042,Cyberpunk 2077, Far Cry 6, Control and not crashed at launch after new drivers. But no issues at all :)
 
Last edited:
Also last question. Please read:
https://www.reddit.com/r/intel/comments/13o29w5/13900k_will_no_longer_run_dx12_games_crashingctds/

Someone had few 13900K and all showing bad symptoms ,like crashing during shader compilation or throwing wheas 19.
Someone said that this can be a flow in microarchitecture. Also degradation on time.
What do you think?
Maybe should i change to 14900K and try a luck or it will be the same situation? in shader compilation no stability on first launch and wheas 19 on first launch during shader compilation
 
So last question. If in second launch is fine and no crashes and no wheas 19 loggers then keep it pc?:) Thank you and sorry for asking again. Just dont wanna long RMA process or maybe buying new cpu or ram or what else lol : )
 
Last edited:
I installed newest nvidia drivers and launched for first time: Lords of The Fallen,Battlefield 2042,Hogwart Legacy ( first deleted shader cache folder like developer suggested ) and no wheas and no crashes in first launch. I just turned off msi afterburner. So its ok. But thx all.
 
hu76 said:
Also last question. Please read:
https://www.reddit.com/r/intel/comments/13o29w5/13900k_will_no_longer_run_dx12_games_crashingctds/

Someone had few 13900K and all showing bad symptoms ,like crashing during shader compilation or throwing wheas 19.
Someone said that this can be a flow in microarchitecture. Also degradation on time.
What do you think?
Maybe should i change to 14900K and try a luck or it will be the same situation? in shader compilation no stability on first launch and wheas 19 on first launch during shader compilation
Click to expand...
so i was just about to post about this situation from something i was reading in the steam forums according to someone that this IS a degradation issue that's been happening with 12-14th gen intel cpu's. i'm going to post his first post and link to the thread if it interests you and you want to read more. (https://steamcommunity.com/app/315210/discussions/0/4204741842669464852/?tscn=1707987443) this would actually be something somebody like Gamers Nexus should look into but he's too much of an intel fanboy. but seems to be kind of widespread and could either be grounds for a recall or class action issue and if you ask me it's from either the bent chip thing or them running to high of voltage/wattage causing the degradation or a combination of both but looks like at this point they are aware of the problem and are trying to just keep sweeping it under the rug.
------------------------------------------

To everyone experiencing shader compilation "Out of video memory trying to allocate a rendering resource" errors
Your CPU is faulty and has Core errors. It's not your GPU, its your CPU

Crashing with this sort error message (and others, including straight up CTDs) when performing shader compilation is famously caused by broken CPUs, namely the Intel 13x and 14x series, and more specifically (but not solely) the 13900k and 14900k variants.

Anyone affected by shader comp crashing on the aforementioned CPUs just needs to open Event Viewer and do a parse for 'WHEA-Logger' warnings. Inside those entries, look for 'Translation Look Aside Buffer' and 'Internal Parity Error'. There will be lots. That's a bad CPU.

Furthermore, the CPU core errors, which are PCore-based can be surfaced by running OCCT whilst applying a very slight OC via BIOS, with SVID=Typical and LLC=4. OCCT will out Core errors almost immediately and then throughout the rest of the run. You may even get OCCT to report Core errors without these BIOS tweaks, depending on how bad the CPU has gotten. Attempt any this at your own risk. I'm simply reiterating how I was able to isolate down.

Need more proof? This is a thread I stumbled over when investigating my own exact same problem.

https://www.reddit.com/r/buildapc/comments/11uftum/rtx_4090_i9_13900k_pc_build_crashing_with/

And this quickly became drenched all the same issues...

https://www.reddit.com/r/intel/comments/13o29w5/13900k_will_no_longer_run_dx12_games_crashingctds/

and the list goes on and on and on and on, and this is only a small fraction of whats out there...

https://www.reddit.com/r/remnantgame/comments/17lpz8g/13900k_not_stable_on_stock_during_first_shader/

https://www.reddit.com/r/intel/comments/12bybl5/something_wrong_with_13900k/

https://www.reddit.com/r/intel/comments/17ku82f/13900k_degradation_i_have_faulty_cpu_and_try_my/

https://www.reddit.com/r/intel/comments/17aob62/13900k_wheas_19_or_bsod_during_shader_compilation/

https://www.reddit.com/r/pcmasterrace/comments/17gb3p0/13900k_out_of_video_memory_or_bsod_during_shader/

The only solution is to either drop the PCore multiplier down, I got stable at 52x but it can vary (and still degrades, so needs more ongoing effort) or RMA. You can drop the PCore using Intel XTU but again, use at your own risk (although its fairly straightforward for anyone with basic PC tech understanding).

*I'm on my 4th 13900k in just over a year*, which has so far been working for a few months.

This is a CPU hardware defect. Nothing in software or OS or BIOS (other than a BIOS-level downclock), will provide a stable system. You can trust me, given I spent close to 6 weeks troubleshooting my own system before isolating it to the CPU.

----

Edit: Just want to make one point really clear; these errors don't generally manifest out of the box. That is, it takes time before the CPU becomes affected. In my case, every CPU took between 1-3 months before the exact same behaviour started to show. DX12 and shader comp, either at initial shader comp or during gameplay, would crash the game client out.

One that continually caused these errors for me was Fortnite DX12 with all the settings ramped to top or one step down from the top

The issues started small. Very occasional errors as per the title, when trying to load the game, but clicking away and trying again would mostly work. Then sometimes work. Then never work, indicating the CPU was degrading. Then even in times where it would load to the menu, trying to get to the staging area would crash out, again with the same error. Checking the game log, it pointed to the shader comp causing a fatal error.

Replacement CPUs fixed all that. For a while. But in those cases, the game would load up and go into a match but during gameplay, would CTD. Checking the game log again, it showed that shader decompression, due to traversal, was again causing a fatal error.

Every time these errors happened, using OCCT with the settings I stated, would show Core errors.

The list of games affected is huge. Every DX12 shader based game will crash at least during initial shader comp or if not, as above, some time after. If shader comp can be gotten past, it might seem OK but as soon as you update the GPU driver, or the game is patched, it will perform shader comp again and that's why some people can see it as hit and miss.
 
Last edited:
d3athf1sh said:
so i was just about to post about this situation from something i was reading in the steam forums according to someone that this IS a degradation issue that's been happening with 12-14th gen intel cpu's. i'm going to post his first post and link to the thread if it interests you and you want to read more. (https://steamcommunity.com/app/315210/discussions/0/4204741842669464852/?tscn=1707987443) this would actually be something somebody like Gamers Nexus should look into but he's too much of an intel fanboy. but seems to be kind of widespread and could either be grounds for a recall or class action issue and if you ask me it's from either the bent chip thing or them running to high of voltage/wattage causing the degradation or a combination of both but looks like at this point they are aware of the problem and are trying to just keep sweeping it under the rug.
------------------------------------------

To everyone experiencing shader compilation "Out of video memory trying to allocate a rendering resource" errors
Your CPU is faulty and has Core errors. It's not your GPU, its your CPU

Crashing with this sort error message (and others, including straight up CTDs) when performing shader compilation is famously caused by broken CPUs, namely the Intel 13x and 14x series, and more specifically (but not solely) the 13900k and 14900k variants.

Anyone affected by shader comp crashing on the aforementioned CPUs just needs to open Event Viewer and do a parse for 'WHEA-Logger' warnings. Inside those entries, look for 'Translation Look Aside Buffer' and 'Internal Parity Error'. There will be lots. That's a bad CPU.

Furthermore, the CPU core errors, which are PCore-based can be surfaced by running OCCT whilst applying a very slight OC via BIOS, with SVID=Typical and LLC=4. OCCT will out Core errors almost immediately and then throughout the rest of the run. You may even get OCCT to report Core errors without these BIOS tweaks, depending on how bad the CPU has gotten. Attempt any this at your own risk. I'm simply reiterating how I was able to isolate down.

Need more proof? This is a thread I stumbled over when investigating my own exact same problem.

https://www.reddit.com/r/buildapc/comments/11uftum/rtx_4090_i9_13900k_pc_build_crashing_with/

And this quickly became drenched all the same issues...

https://www.reddit.com/r/intel/comments/13o29w5/13900k_will_no_longer_run_dx12_games_crashingctds/

and the list goes on and on and on and on, and this is only a small fraction of whats out there...

https://www.reddit.com/r/remnantgame/comments/17lpz8g/13900k_not_stable_on_stock_during_first_shader/

https://www.reddit.com/r/intel/comments/12bybl5/something_wrong_with_13900k/

https://www.reddit.com/r/intel/comments/17ku82f/13900k_degradation_i_have_faulty_cpu_and_try_my/

https://www.reddit.com/r/intel/comments/17aob62/13900k_wheas_19_or_bsod_during_shader_compilation/

https://www.reddit.com/r/pcmasterrace/comments/17gb3p0/13900k_out_of_video_memory_or_bsod_during_shader/

The only solution is to either drop the PCore multiplier down, I got stable at 52x but it can vary (and still degrades, so needs more ongoing effort) or RMA. You can drop the PCore using Intel XTU but again, use at your own risk (although its fairly straightforward for anyone with basic PC tech understanding).

*I'm on my 4th 13900k in just over a year*, which has so far been working for a few months.

This is a CPU hardware defect. Nothing in software or OS or BIOS (other than a BIOS-level downclock), will provide a stable system. You can trust me, given I spent close to 6 weeks troubleshooting my own system before isolating it to the CPU.

----

Edit: Just want to make one point really clear; these errors don't generally manifest out of the box. That is, it takes time before the CPU becomes affected. In my case, every CPU took between 1-3 months before the exact same behaviour started to show. DX12 and shader comp, either at initial shader comp or during gameplay, would crash the game client out.

One that continually caused these errors for me was Fortnite DX12 with all the settings ramped to top or one step down from the top

The issues started small. Very occasional errors as per the title, when trying to load the game, but clicking away and trying again would mostly work. Then sometimes work. Then never work, indicating the CPU was degrading. Then even in times where it would load to the menu, trying to get to the staging area would crash out, again with the same error. Checking the game log, it pointed to the shader comp causing a fatal error.

Replacement CPUs fixed all that. For a while. But in those cases, the game would load up and go into a match but during gameplay, would CTD. Checking the game log again, it showed that shader decompression, due to traversal, was again causing a fatal error.

Every time these errors happened, using OCCT with the settings I stated, would show Core errors.

The list of games affected is huge. Every DX12 shader based game will crash at least during initial shader comp or if not, as above, some time after. If shader comp can be gotten past, it might seem OK but as soon as you update the GPU driver, or the game is patched, it will perform shader comp again and that's why some people can see it as hit and miss.
Click to expand...
thx for such large post :) Actually all is fine. After installed new drivers i launching every dx12 game with shader compilation for the first time and its not crashing. So i guess i m fine,i think ;)

Tried: Battlefield 2042, Remnant 2,Lords of The Fallen, Hogwart Legacy and first launch after new installed driver , no crash. Uff ;)
 
to anyone still interested in this. the original post from the steam forums i mentioned the other day is now making headlines at Tom's Hardware and they have reached out to intel for a response. i posted it in hardware news if you want to keep up with what happens.
 
d3athf1sh said:
so i was just about to post about this situation from something i was reading in the steam forums according to someone that this IS a degradation issue that's been happening with 12-14th gen intel cpu's. i'm going to post his first post and link to the thread if it interests you and you want to read more. (https://steamcommunity.com/app/315210/discussions/0/4204741842669464852/?tscn=1707987443) this would actually be something somebody like Gamers Nexus should look into but he's too much of an intel fanboy. but seems to be kind of widespread and could either be grounds for a recall or class action issue and if you ask me it's from either the bent chip thing or them running to high of voltage/wattage causing the degradation or a combination of both but looks like at this point they are aware of the problem and are trying to just keep sweeping it under the rug.
------------------------------------------

To everyone experiencing shader compilation "Out of video memory trying to allocate a rendering resource" errors
Your CPU is faulty and has Core errors. It's not your GPU, its your CPU

Crashing with this sort error message (and others, including straight up CTDs) when performing shader compilation is famously caused by broken CPUs, namely the Intel 13x and 14x series, and more specifically (but not solely) the 13900k and 14900k variants.

Anyone affected by shader comp crashing on the aforementioned CPUs just needs to open Event Viewer and do a parse for 'WHEA-Logger' warnings. Inside those entries, look for 'Translation Look Aside Buffer' and 'Internal Parity Error'. There will be lots. That's a bad CPU.

Furthermore, the CPU core errors, which are PCore-based can be surfaced by running OCCT whilst applying a very slight OC via BIOS, with SVID=Typical and LLC=4. OCCT will out Core errors almost immediately and then throughout the rest of the run. You may even get OCCT to report Core errors without these BIOS tweaks, depending on how bad the CPU has gotten. Attempt any this at your own risk. I'm simply reiterating how I was able to isolate down.

Need more proof? This is a thread I stumbled over when investigating my own exact same problem.

https://www.reddit.com/r/buildapc/comments/11uftum/rtx_4090_i9_13900k_pc_build_crashing_with/

And this quickly became drenched all the same issues...

https://www.reddit.com/r/intel/comments/13o29w5/13900k_will_no_longer_run_dx12_games_crashingctds/

and the list goes on and on and on and on, and this is only a small fraction of whats out there...

https://www.reddit.com/r/remnantgame/comments/17lpz8g/13900k_not_stable_on_stock_during_first_shader/

https://www.reddit.com/r/intel/comments/12bybl5/something_wrong_with_13900k/

https://www.reddit.com/r/intel/comments/17ku82f/13900k_degradation_i_have_faulty_cpu_and_try_my/

https://www.reddit.com/r/intel/comments/17aob62/13900k_wheas_19_or_bsod_during_shader_compilation/

https://www.reddit.com/r/pcmasterrace/comments/17gb3p0/13900k_out_of_video_memory_or_bsod_during_shader/

The only solution is to either drop the PCore multiplier down, I got stable at 52x but it can vary (and still degrades, so needs more ongoing effort) or RMA. You can drop the PCore using Intel XTU but again, use at your own risk (although its fairly straightforward for anyone with basic PC tech understanding).

*I'm on my 4th 13900k in just over a year*, which has so far been working for a few months.

This is a CPU hardware defect. Nothing in software or OS or BIOS (other than a BIOS-level downclock), will provide a stable system. You can trust me, given I spent close to 6 weeks troubleshooting my own system before isolating it to the CPU.

----

Edit: Just want to make one point really clear; these errors don't generally manifest out of the box. That is, it takes time before the CPU becomes affected. In my case, every CPU took between 1-3 months before the exact same behaviour started to show. DX12 and shader comp, either at initial shader comp or during gameplay, would crash the game client out.

One that continually caused these errors for me was Fortnite DX12 with all the settings ramped to top or one step down from the top

The issues started small. Very occasional errors as per the title, when trying to load the game, but clicking away and trying again would mostly work. Then sometimes work. Then never work, indicating the CPU was degrading. Then even in times where it would load to the menu, trying to get to the staging area would crash out, again with the same error. Checking the game log, it pointed to the shader comp causing a fatal error.

Replacement CPUs fixed all that. For a while. But in those cases, the game would load up and go into a match but during gameplay, would CTD. Checking the game log again, it showed that shader decompression, due to traversal, was again causing a fatal error.

Every time these errors happened, using OCCT with the settings I stated, would show Core errors.

The list of games affected is huge. Every DX12 shader based game will crash at least during initial shader comp or if not, as above, some time after. If shader comp can be gotten past, it might seem OK but as soon as you update the GPU driver, or the game is patched, it will perform shader comp again and that's why some people can see it as hit and miss.
Click to expand...
Did you ever take care to manually set your bios so that it runs your CPU at Intel's actual stock settings?

I have a feeling this is some sort of microcode change from Intel at some point, which has ended up not playing nice with Motherboard Brand's "default" settings, which are actually pushing power limits, voltage curves, etc, beyond Intel's actual stock settings.
 
Last edited:
chameleoneel said:
Did you ever take care to manually set your bios so that it runs your CPU at Intel's actual stock settings?

I have a feeling this is some sort of microcode change from Intel at some point, which has ended up not playing nice with Motherboard Brand's "default" settings, which are actually pushing power limits, voltage curves, etc, beyond Intel's actual stock settings.
Click to expand...
I think it's something Intel has done, for sure. Or, quality is just out the window.

Returned the 14900KS for instability. Chip was garbage, even at reduced clocks, voltages, etc. No settings were stable on mine. Even chrome crashed when clocks were down to 4000. RAM dropped to 4800. Nothing got it stable.

Dropped a 12600k into the mb, since it was sitting in a box, and zero issues. Even running the same RAM XMP @7600.

Considering keeping it or trying the 12900KS instead of putting the 13900KS back in. The 13 I had to reduce RAM clocks for stability.

So, something is going on with 13/14 gen at the high end. If I hadn't gotten an MC deal on the 12th way back I wouldn't have upgraded. Buying new today? I'd look at AMD on the high end. Either seem fine in the middle or lower.
 
wareyore said:
I think it's something Intel has done, for sure. Or, quality is just out the window.

Returned the 14900KS for instability. Chip was garbage, even at reduced clocks, voltages, etc. No settings were stable on mine. Even chrome crashed when clocks were down to 4000. RAM dropped to 4800. Nothing got it stable.

Dropped a 12600k into the mb, since it was sitting in a box, and zero issues. Even running the same RAM XMP @7600.

Considering keeping it or trying the 12900KS instead of putting the 13900KS back in. The 13 I had to reduce RAM clocks for stability.

So, something is going on with 13/14 gen at the high end. If I hadn't gotten an MC deal on the 12th way back I wouldn't have upgraded. Buying new today? I'd look at AMD on the high end. Either seem fine in the middle or lower.
Click to expand...
Ok but you didn't actually answer my question: Did you ever take care to manually set your bios so that it runs your CPU at Intel's actual stock settings?
 
chameleoneel said:
Ok but you didn't actually answer my question: Did you ever take care to manually set your bios so that it runs your CPU at Intel's actual stock settings?
Click to expand...
No, I never touched anything in the BIOS when I tried for two weeks to tweak the "reduced clocks, voltages, etc."

Yes, and FAR BELOW THEIR STOCK SETTINGS.

Garbage chip is garbage.

Temps were good down there, tho. Barely in the low 60's running nearly 4 minutes of C23 at 4500/4000. But, it still crapped itself.
 
VIPs: Intel, you may have a problem?

Intel: Huh? What?

VIPs: There might be problem that causes your 13th and 14th gen CPUs to fail. Could be in the architecture itself.

Intel: What do you suggest we do?

VIPs: Maybe contact your legal team. You could be liable.

Intel: No. We have a better idea.

VIPs: What are you going to do?

Intel: We're calling Rashee Rice!

Need swagger when walking away from a crash you caused? Call Rashee Rice today. 1-800-THUG-LIFE
 
Last edited:
wareyore said:
Dropped a 12600k into the mb, since it was sitting in a box, and zero issues. Even running the same RAM XMP @7600.
Click to expand...
I was running a 13900K for months with zero issues with any games. The only reason I switched to an i5 was due to heat and combined power draw with the 4090. I realized then that you really don't need to have an i7 or i9 to have an excellent gaming PC, and from what I'm seeing now I made the right choice.
 
Great, 13900K user here, thought I was unnaffected as I've not really noticed any major crashing, but then, I don't tend to play a alot of very recent games.

But I looked in my event viewer for WHEA errors anyway, and sure enough, the last 3 new games I've installed have apparently thrown WHEA errors:

(*edit* Just to note, that's the full total of just 9 WHEA errors in my logs since Jan 2023)

WHEA errors.png

I've not really noticed any crashing though, but something that HAS happened in the last few months or so is that games will freeze for about 5 seconds, then carry on perfectly fine. I'd put it down to nvidia driver error, but damn, guess now I know otherwise? :(

Are there any free demos/games that I can quickly run to test for the error?

Running OCCT, no errors so far... *Edit* >> No errors from anything.

*Edit* Just to be clear, I've NOT ONCE had the specific 'VRAM out of memory' error that's been mentioned, in any game, at all, so I'm not sure how useful my use case is here, but sharing nonetheless.
 
Last edited:
Are people having these problems with AMD cards?

I got a 14900k on the way to pair with a Nitro 7900xtx
 
I believe it's an intel CPU hardware issue, easily rectified in software with almost no real-world performance loss, they likely already have the fix.

The trouble is when you roll it out, what spin do you give?
 
l88bastard said:
Are people having these problems with AMD cards?

I got a 14900k on the way to pair with a Nitro 7900xtx
Click to expand...

From what I've seen in reddit reports, most aren't stating their gpus. Those that do list it, it is a 4090 and a few 3090 cards. Most of the reports seem to be issues on the i9 chips than the i7 chips and only 13/14th gen stuff. I would still test the heck out of a 14900k to check for instability no matter the gpu.
 
l88bastard said:
Are people having these problems with AMD cards?

I got a 14900k on the way to pair with a Nitro 7900xtx
Click to expand...
IMO, enforce Intel stock limits and update your bios as soon as updates come out, for the next 2 or 3 updates (and re-do the actual intel limits, after each update).

You might lose a couple thousand in Cinibench all core. But, you will lose zero gaming performance, use a lot less power, and it could avoid most of the issues happening, until a real fix is issued in bios updates.
 
It is funny people don't mention PSU
Power supply is fundamental system component - especially for these CPUs

Imho it is Intel's design flaw to release 300+W CPU especially given motherboard manufacturers on one side cut costs on good components and on the other side use "throw more voltage - shouldn't hurt" tactics to avoid stability issues. It is all nice and dandy when you have say 95W CPU that maybe can run at 160W - then if you up this by say 40W its still within some reason. Then also PSU quality doesn't matter as much.

The issue here is that settings are not like they should be - hefty OC overhead so that even worst silicon can handle clocks with ease. Instead Intel is selling pre-overclocked chips barely managing to run at stock settings. 100MHz more and they wouldn't be stable. That is bad design.

Personally I used 13600KF with manual OC and it ran fine. Replaced it to 13900KF and cooling to direct die water cooling and I couldn't quite get stability I expected despite everything seemingly being okay. Replaced PSU to something actually good and so far it has been rock solid.

That said this issue might be something else - would be good if someone made proper tool to induce activity patterns which reveal the issue. I suspect that if its compilation then its lots of random spikes. I compile programs a lot myself and compilation in general is an amazing stability tester. In fact I did at one time validate system with prime95, memtest, cinebench in a loop, etc. but would still get random issues with compilation. I had to loosen memory timings to fix that issue.
 
For affected pool of users, games that used to run with perfect stability for more than a year on the same exact chip suddently began crashing during shader pre-caching within 2-3 seconds of the process without CPU even getting hot. Good chips turning bad for many people at once during the same time frame? What are the chances of that? These chips didn't just become less stable, but completely unstable in specific situations.

Raptor Lake has been around for more than a year and yet these issues began gaining tracking starting just a few months ago. I know at least one game that crashes during shader compilation for affected users regains full stability when running in Windows 8 Compatibility Mode. That mode makes load times very long, but without any crashes and without framerate reduction. Here's a brief explanation of what Windows Compatibility modes do and may shed a bit of light on why such modes fix crashes discussed in this thread, at least for some games and for some people - https://superuser.com/questions/133746/how-does-the-compatibility-mode-in-windows-work .

Microsoft has been releasing a ton of security updates for the past few months and some of these updates were from Intel in regards to Indirect Branch Prediction exploits. I think some of these updates negatively affected stability the same way Spectre/Meltdown mitigations and microcodes negatively affected performance on older CPU's. Intel, NVidia, AMD, and Microsoft already investigate, but finding the root of the problem is going to take time.
 
Last edited:
Just tested shader compilation for a craahing game. It caused all CPU cores to reach TJ Max (100 C) in 5 seconds (without heavy load) when using default Win11 mode and it crashed. In Win8 compatibility mode, only a few cores reached 100 C and it did not crash. Everything was at stock factory settings.

My cooling and overall temps are great. I can't reach TJ Max even when running Prme95 for hours. Shader Compilation is the only process that causes such high temperatures. Load is at about 20% during that time. Shader Compilation itself must be using some special instructions.
 
Last edited:
OpenSource Ghost said:
Just tested shader compilation for a craahing game. It caused all CPU cores to reach TJ Max (100 C) in 5 seconds (without heavy load) when using default Win11 mode and it crashed. In Win8 compatibility mode, only a few cores reached 100 C and it did not crash. Everything was at stock factory settings.

My cooling and overall temps are great. I can't reach TJ Max even when running Prme95 for hours. Shader Compilation is the only process that causes such high temperatures. Load is at about 20% during that time. Shader Compilation itself must be using some special instructions.
Click to expand...
Interesting. It's like you stumbled onto Furmark for CPUs. Sounds like Intel may have a thermal throttling bug combined with that shader compiler causing a perfect storm of heat. I wonder if running Intel stock is really necessary or if you just need a test routine that causes an exceptional load on the CPU so you can find out how much power you can feed it without crashing or melting the CPU. What game and how are you compiling the shaders? You might actually have a useful test for power limit tuning.
 
Yes! Power current amp value adjustment fixed this issue for me. It was the one variable that made all the difference. When set to 512A, CPU doesn't throttle at all and just crashes during game shader compilation, which forces CPU temperature to 100C Tj Max, which normal, but CPU core clocks don't throttle at all. End result = crash or BSOD during game shader compilation. When I reduce amps to safe limit, all crashes stop and CPU begins to throttle a bit during Tj Max periods. If I reduce amps too far, then CPU throttles too much. Game benchmark results were not affected.

Given that the same game shader compilation processes were not crashing with power current set to 512A earlier, the actual root cause was CPU degradation due to Intel and Intel motherboards' makers using very aggressive power settings. I am still glad I can stabilize my CPU after watching that video!

Many thanks for that video!
 
zandor said:
Interesting. It's like you stumbled onto Furmark for CPUs. Sounds like Intel may have a thermal throttling bug combined with that shader compiler causing a perfect storm of heat. I wonder if running Intel stock is really necessary or if you just need a test routine that causes an exceptional load on the CPU so you can find out how much power you can feed it without crashing or melting the CPU. What game and how are you compiling the shaders? You might actually have a useful test for power limit tuning.
Click to expand...

Just run DX12 games that compile shaders before loading the game (or loading saved game) and not compiling them during gameplay. Good examples are The Last of Us and Horizon - Forbidden West.
 
You must log in or register to reply here.
Back
Top