PC Crashing- 14900k, RTX4090

[marianko]

Hello, 

I´ve built new computer for professional work in Archicad, Autocad, Twinmotion, Adobe CC... (and also some gaming )

The problem is my computer is crashing. Sometimes BSOD, other times just app would crash.

 

SPECS: 

MOBO: ASROCK Z790 PG-ITX / TB4

CPU: Intel i9 14900k

RAM: G.SKILL 96GB KIT DDR5 6 400 MHz CL32 Trident Z5 RGB Black

AIO: Be quiet! Silent Loop 2 280

GPU: MSI GeForce RTX 4090 GAMING X SLIM 24G, 24GB GDDR6X

PSU: ASUS ROG LOKI SFX-L 1000W

Fans: Noctua

Case: SSUPD Meshlicious 

build post with pics:

 

 

When I game or do work in Adobe or other demanding tasks everything is stable. Synthetic benchmarks also 100% stable over night. Temps within reason. 

Only when using twinmotion or archicad i get crash, which sucks, bcs I mostly use those for my work. 

 

when using Twinmotion i get

EXCEPTION_ACCESS_VIOLATION reading address 0x0000000000000238

usually when trying to apply some texture to material 

 

When using Archicad, sometimes program will quit, sometimes I get BSOD, other times my computer will just turn off instantly. 

When it will only quit I can open it again and it will crash in a few seconds. However when I restart my PC it works for a while, until first crash, and then it is the same thing over again.

There is no error code in Archicad. If it is possible to log it somehow, please help me. 

 

What I´ve tried so far:

run my projects on different machines: works without crashing

Reinstall (and update) BIOS, reset to default values

Remove XMP, now running base 5200mhz

increase or decrease voltage offset of CPU (by little)

Set power limit of CPU from 4096W (default) to other recommended values

Reinstall programs causing problems

Nvidia Studio driver and Gaming driver, different versions.

 

I am lost here, need help. I am under warranty, so I can replace components, but they are from different shops (bcs prices) and I dont want to send good pieces using postal services away for month if not needed. I need good route for troubleshooting, I want to help and provide any evidence or logs if necessary. 

 

Thanks for your help,

If we can solve it together I would highly appreciate it. 

 

[Qrdello]

I'd guess you're getting the worst case scenario with the 14th gen i9 crashfest, meaning the samples themselves are incredibly unstable which adds to mobo manufacturers developing their boards with ridiculous power-related settings. 

Right now I would either look up a for a new UEFI that includes intel baseline power profile, or significantly declock the i9 (some say even to less than 5ghz) so as to at least check if that makes it stable. You can also be really unfortunate and be in those 20-something% customers that just got really bad chips will have to RMA it  

[marianko]

OK now i am going to try following: 

 

clean install of OS (probably wont help, I´ve already tried safe mode and different drive with different install)

 

Iwill buy different Intel CPU, i5 14500 or something. I can return it in 14 days period. If it will help I will RMA my CPU. 

I´ll let you all know

 

Thanks

[Alex Atkin UK]

19 minutes ago, marianko said:

OK now i am going to try following: 

 

clean install of OS (probably wont help, I´ve already tried safe mode and different drive with different install)

 

Iwill buy different Intel CPU, i5 14500 or something. I can return it in 14 days period. If it will help I will RMA my CPU. 

I´ll let you all know

 

Thanks

You shouldn't do any of that until you have followed the advice already given.

[rippy4500]

Lock the core multiplier to 5.5GHz, the intel 14th gen crashing problem is from the crappy boosting algorithm pushing the cpu too hard.

[Agall]

48 minutes ago, marianko said:

Hello, 

I´ve built new computer for professional work in Archicad, Autocad, Twinmotion, Adobe CC... (and also some gaming )

The problem is my computer is crashing. Sometimes BSOD, other times just app would crash.

 

SPECS: 

MOBO: ASROCK Z790 PG-ITX / TB4

CPU: Intel i9 14900k

RAM: G.SKILL 96GB KIT DDR5 6 400 MHz CL32 Trident Z5 RGB Black

AIO: Be quiet! Silent Loop 2 280

GPU: MSI GeForce RTX 4090 GAMING X SLIM 24G, 24GB GDDR6X

PSU: ASUS ROG LOKI SFX-L 1000W

Fans: Noctua

Case: SSUPD Meshlicious 

build post with pics:

 

 

When I game or do work in Adobe or other demanding tasks everything is stable. Synthetic benchmarks also 100% stable over night. Temps within reason. 

Only when using twinmotion or archicad i get crash, which sucks, bcs I mostly use those for my work. 

 

when using Twinmotion i get

EXCEPTION_ACCESS_VIOLATION reading address 0x0000000000000238

usually when trying to apply some texture to material 

 

When using Archicad, sometimes program will quit, sometimes I get BSOD, other times my computer will just turn off instantly. 

When it will only quit I can open it again and it will crash in a few seconds. However when I restart my PC it works for a while, until first crash, and then it is the same thing over again.

There is no error code in Archicad. If it is possible to log it somehow, please help me. 

 

What I´ve tried so far:

run my projects on different machines: works without crashing

Reinstall (and update) BIOS, reset to default values

Remove XMP, now running base 5200mhz

increase or decrease voltage offset of CPU (by little)

Set power limit of CPU from 4096W (default) to other recommended values

Reinstall programs causing problems

Nvidia Studio driver and Gaming driver, different versions.

 

I am lost here, need help. I am under warranty, so I can replace components, but they are from different shops (bcs prices) and I dont want to send good pieces using postal services away for month if not needed. I need good route for troubleshooting, I want to help and provide any evidence or logs if necessary. 

 

Thanks for your help,

If we can solve it together I would highly appreciate it. 

 

No matter what, I would have that 14900k at a max 105W TDP, same with the RTX 4090 at 90% TDP. Kind of ridiculous to have those components in SFX knowing they can draw in excess of 600W combined at stock settings.

[marianko]

7 minutes ago, Agall said:

No matter what, I would have that 14900k at a max 105W TDP, same with the RTX 4090 at 90% TDP. Kind of ridiculous to have those components in SFX knowing they can draw in excess of 600W combined at stock settings.

I dont understand why this would be crazy. I have 280mm AIO with aditional 120mm fan blowing cold air right to it, my CPU is not throttling under load. at least after I removed crazy power limit   and my GPU is provided with direct fresh air from outside and have 3 aditional fans blowing hot air away from it. When I game it is usually around 65 deg. c. 

 

I am going to try to lock multiplier to 5,5 GHz, or even lower. Will let you know. 

[Agall]

2 minutes ago, marianko said:

I dont understand why this would be crazy. I have 280mm AIO with aditional 120mm fan blowing cold air right to it, my CPU is not throttling under load. at least after I removed crazy power limit   and my GPU is provided with direct fresh air from outside and have 3 aditional fans blowing hot air away from it. When I game it is usually around 65 deg. c. 

 

I am going to try to lock multiplier to 5,5 GHz, or even lower. Will let you know. 

Its a wattage/volume concern with neighboring components. Just because the major components with active cooling are okay doesn't mean the rest of the system is.

[marianko]

17 minutes ago, Agall said:

Its a wattage/volume concern with neighboring components. Just because the major components with active cooling are okay doesn't mean the rest of the system is.

Oh I get it, all my temps seem to be ok, but I cant tell you you are wrong as I obviously have some kind of problem here  

[marianko]

Intel Base Power Profile settings- loaded

P core ratio- all cores 5 Ghz max now

E core ratio- all cores 4ghz max now

Twinmotion and Archicad both still crashing. 

This is awful to diagnose because both can run up to 1 hour without crashing. 

 

Any ideas? 

 

 

[Agall]

1 minute ago, marianko said:

Oh I get it, all my temps seem to be ok, but I cant tell you you are wrong as I obviously have some kind of problem here  

I spent 7 years doing mITX only builds. Without significant negative pressure, its common for hot spots to pop up on the motherboard or even backside of the graphics card. This was a problem when GPUs were maybe 200W and CPUs were maybe 100W, like with a Titan X and 4790k.

 

14900k doesn't lose significant performance, if any in gaming at 105W. If any, 105-150W being the suite spot for Intel right now, 105W being the absolute goat for AM5. RTX 4090 barely loses performance at 90% TDP, approximately 400W. Still a significant amount of heat in such a small package.

 

If it puts it into perspective, my Fractal Design North is effective SFF for an overclocked RTX 4090. You'd have to show some panel-less pictures for me to suggest anything outside of lowering max TDPs, but I wouldn't doubt if the stability issue is related to such.

[BiG StroOnZ]

6 hours ago, marianko said:

Intel Base Power Profile settings- loaded

P core ratio- all cores 5 Ghz max now

E core ratio- all cores 4ghz max now

Twinmotion and Archicad both still crashing. 

This is awful to diagnose because both can run up to 1 hour without crashing. 

 

Any ideas? 

 

Some interesting settings that might assist if you haven't tried already,

 

First off, make sure any type of MCE ("MultiCore Enhancement") setting is disabled if your motherboard has these settings. Also, reset CMOS to start fresh. I know you said you possibly tried some of this, but I don't know of the exact values; as you don't mention them, so here are some other recommendations:

• Set SVID behavior to Typical Scenario or Trained (if you have this option).
• Set short duration turbo power = 253.
• Set long duration turbo power = 253.
• Set max core/cache current or current limit = 307Amps.
• Set Package Power Limit 1 = 253.
• Set Package Power Limit 2 = 253.

You can also try:

• 288 on short duration and 288 PL1, with 253 on long duration and 253 on PL2 with the same 307Amps setting.

Also, make sure you use LLC3 or LLC4 (Load Line Calibration controls the load line impedance/slope, and "load line" Adaptive [or Active] voltage positioning). Alternative, or additional settings:

• Load Line Spec (LLC) 1.100 mOhm (Standard or Normal).
• DC LL 1.100 mOhm.
• AC LL 1.100 mOhm.

Now onto the memory, this could be the crux of your issue IMO (trying to get that 96GB kit stable):

• At least 1.365V for DRAM VDD (Voltage for the DRAM Data Lines)
• 1.365V DRAM VDDQ (Voltage for the DRAM Output Buffers)
• 1.8V for DRAM VPP (Programming Voltage for EEPROM).
• CPU VDDQ - make sure it's around 1.4V, but under.
• VCCSA or SA Voltage (System Agent)- 1.110-1.120V is a good value to shoot for (up to 1.4V is possible, but definitely under 1.35v is more common).
• VCCIO - If available, would increase this too, but slightly in increments (definitely under 1.4v).

If these values don't work at XMP or lower, 5200MT/s or 4800MT/s (I would even try lowering Memory Frequency with these settings to see if you can get it stable all the way down to about 4400MT/s). After that point, if that still doesn't work; I would take one stick out and see if it was a memory compatibility issue as 96GB is IME heavy stressing on the IMC. I know there are people who get lucky and have no stability problems with these new kits, but I would still be suspicious of getting them stable despite being 2x2 and not 4x4. 

 

Finally, as recommended by @rippy4500 too:

• Set a Fixed Core Clock by locking Performance Core Ratio to a fixed value or Sync ALL-Core Ratio between 5.5-5.7GHz.
• 1.2-1.35v VCORE is acceptable starting values to play around with at these clocks (keep it under 1.45v though).
• Play with Ring and Cache settings 4.4-4.7 Ring/Cache

On a side note: Try playing with any combination of these settings during troubleshooting, something has got to give in at some point. 

[marianko]

Thanks, my computer seems to be stable. I will try to play with it more.

 

I am now running at 4Ghz across all cores with package power limit at 253.

 

I hate it, I should´ve bought i5 or something, save my money and have around same performance lol. 

 

Sad story. I can RMA it, but the same or worse piece can return, or can show marks after a few months. I want intel to take proper responsibility. Is there some kind of lawsuit?  

[Agall]

54 minutes ago, marianko said:

Thanks, my computer seems to be stable. I will try to play with it more.

 

I am now running at 4Ghz across all cores with package power limit at 253.

 

I hate it, I should´ve bought i5 or something, save my money and have around same performance lol. 

 

Sad story. I can RMA it, but the same or worse piece can return, or can show marks after a few months. I want intel to take proper responsibility. Is there some kind of lawsuit?  

You could always disable the E cores to reduce overall multicore and prioritize the P cores. Also setting a lower TDP limit.

[Qrdello]

1 hour ago, marianko said:

Thanks, my computer seems to be stable. I will try to play with it more.

 

I am now running at 4Ghz across all cores with package power limit at 253.

 

I hate it, I should´ve bought i5 or something, save my money and have around same performance lol. 

 

Sad story. I can RMA it, but the same or worse piece can return, or can show marks after a few months. I want intel to take proper responsibility. Is there some kind of lawsuit?  

That's exactly the thing - i7 and i9 users had paid significant penny for their CPUs and what they receive is a pile of trash that needs adjusting to just be usable whilst it should be idiot proof right out of the box and be working on any mobo with sufficient PD. To make things worse Intel won't say any specifics or admit they had made serious miscalculations because they're at war with AMD and will not prove otherwise. That's just plain stupid, not because of the rivalrly per se, but the fact that they have been releasing knowingly a faulty product advertised as a godsend. And that would be acceptable still, had it's issues been limited to low or mid-tier boards, but high-end product is affected the same if not even more than aformentioned.

[BiG StroOnZ]

4 hours ago, marianko said:

Thanks, my computer seems to be stable. I will try to play with it more.

 

I am now running at 4Ghz across all cores with package power limit at 253.

 

I hate it, I should´ve bought i5 or something, save my money and have around same performance lol. 

 

Sad story. I can RMA it, but the same or worse piece can return, or can show marks after a few months. I want intel to take proper responsibility. Is there some kind of lawsuit?  

 

I would definitely play with it more and take a look at the list of settings I presented: as any combination of those should help with stability immensely. To the point where I'm certain you can do much better than 4GHz All-Core.

 

I wouldn't regret it too much, while the out-of-box experience is lackluster with your CPU, I'm sure you will get the quirks down and get it stable at something more acceptable; like 5.4-5.6GHz All-Core with the 96GB memory kit at least at 5200-5600MHz. I would even look at adjusting the E-Cores as well. People seem to be having good luck locking those to around 4.2-4.4GHz, while figuring out an acceptably stable Ring/Cache Ratio.

 

Here is a nice guide that might make understanding those settings I posted easier:

 

Asus Maximus Z790 and Intel i9-13900k/14900k - An overclocking and tuning guide. | Overclock.net 

 

To be forthright, both 12th/13th/14th Gen and Zen4/AM5 have had their fair share of platform headaches. Mostly related to DDR5 initially. There were also some controversies with the 7900x3D and 7950x3D regarding its performance later. Same goes with the 14900K/14900KS and its stability issues.

 

This is always commonplace when it comes to fairly new architectures. So, take it as you will.