FahCore_a8 v0.0.12 segfault on Ubuntu 24.04 LTS / Kernel 6.14 - CPU work units crash immediately

[Mark_folding]

Subject: FahCore_a8 v0.0.12 segfault on Ubuntu 24.04 LTS / Kernel 6.14 - CPU work units crash immediately
System Information:

OS: Ubuntu 24.04 LTS
Kernel: 6.14
CPU: AMD Ryzen 9 5900X 12-Core Processor (24 threads)
GPU: NVIDIA GeForce RTX 5070 Ti
RAM: 62.72 GiB
F@H Client: v8.4.9
Username: Hemlok
Team: 250299

Problem Description:
All CPU work units fail immediately after starting with "Core was killed" / "Core crashed with exit code 0" errors. GPU folding works perfectly without issues.
Symptoms:

CPU work units (using FahCore_a8 v0.0.12, Gromacs core) consistently crash within seconds of starting
Work units complete only "1 out of 2500000 steps (0%)" before crashing
GPU work units (using Core22, OpenMM) complete successfully
Pattern repeats with every CPU work unit assigned

Example from F@H logs:
20:55:43:I1:WU185:Project: 16959 (Run 19, Clone 116, Gen 991)
20:55:43:I1:WU185:Calling: mdrun -c frame991.gro -s frame991.tpr -x frame991.xtc -cpt 5 -nt 12 -ntmpi 1
20:55:43:I1:WU185:Steps: first=-1817467296 total=-1814967296
20:55:44:I1:WU185:Completed 1 out of 2500000 steps (0%)
20:56:51:E :WU185:Core was killed
20:56:51:E :WU185:Core crashed with exit code 0
20:56:51:E :WU185:Core returned FAILED_1 (0)
20:56:51:E :WU185:Run did not produce any results. Dumping WU
```

**Note:** The negative step numbers (`first=-1817467296 total=-1814967296`) appear unusual and may indicate a corrupted work unit or core issue.

**Kernel logs (dmesg) show segmentation faults:**
```
[15374.949026] FahCore_a8[13727]: segfault at fffffffe2e338230 ip 000000000071b329 sp 000076475dffad70 error 5 in FahCore_a8[31b329,406000+e6d000] likely on CPU 15 (core 3, socket 0)
[15374.949036] Code: 6a 59 d1 c4 41 7a 2c cd c5 7a 2c c1 c5 fa 2c fa c4 c1 7a 2a c1 4d 63 d1 4c 8b 4c 24 30 c4 41 0a 2a f0 c5 12 5c e0 c5 f8 57 c0 <c4> 01 1a 58 2c 91 4d 63 c8 c5 fa 2a c7 4c 8b 44 24 38 c4 41 72 5c

[15493.242565] FahCore_a8[14043]: segfault at fffffffe17b866b0 ip 000000000071b329 sp 000072e7f6ffcd70 error 5 in FahCore_a8[31b329,406000+e6d000] likely on CPU 20 (core 10, socket 0)
[15493.242579] Code: 6a 59 d1 c4 41 7a 2c cd c5 7a 2c c1 c5 fa 2c fa c4 c1 7a 2a c1 4d 63 d1 4c 8b 4c 24 30 c4 41 0a 2a f0 c5 12 5c e0 c5 f8 57 c0 <c4> 01 1a 58 2c 91 4d 63 c8 c5 fa 2a c7 4c 8b 44 24 38 c4 41 72 5c
```

**FahCore_a8 details:**
```
Core: Gromacs
Type: 0xa8
Version: 0.0.12
Date: Jan 16 2021
Time: 19:24:44
Compiler: GNU 8.3.0
Platform: linux2 4.15.0-128-generic
Bits: 64
Mode: Release
SIMD: avx2_256
OpenMP: ON
CUDA: OFF
Root Cause:
The FahCore_a8 binary (compiled for kernel 4.15.0-128) is experiencing memory access violations (segfaults) when running on the modern kernel 6.14. The core is attempting to access invalid memory addresses, causing immediate crashes.
Troubleshooting steps attempted:

Fixed v7 to v8 config file migration issues (removed deprecated options: gpu, fold-anon, pci-bus, pci-slot)
Deleted and re-downloaded FahCore_a8 (rm -rf /var/lib/fah-client/cores/fahcore-a8*)
Full service restart
Tried multiple different CPU work units - all crash with identical symptoms
Verified no DNS/network issues (other computers on network fold successfully, GPU folding works)
Checked dmesg logs - confirmed segmentation faults in FahCore_a8

Workaround:
Currently running with CPU folding disabled (<cpus value="0"/> in config.xml). GPU folding continues to work flawlessly.
Question:
Is there a newer version of FahCore_a8 available that's compatible with Ubuntu 24.04 / kernel 6.14? Or is this a known issue with a planned fix?
The core appears to be nearly 4 years old (Jan 2021) and may need to be recompiled for modern Linux kernels.

[muziqaz]

There is nothing wrong with the core. Your CPU/system is unstable

Run full system stability test, like y-cruncher for extensive amount of time to determine if the system is stable or not. The way yours craps out straight away, it might be temperature of the cpu

[Mark_folding]

How can I test this suggestion further? I stress tested each section of this recently assembled computer one at a time. This is an open case system with air cooling, and it settles in at 70deg C at 100% load. Isn't that well under what this CPU should be capable of? I've been folding on many machines over many years but this is the first time that I've seen this. If I have a flakey CPU, I'm nor sure how to confirm. I used AI to walk me through one test at time for each major component, but I don't have another MB to swap to and have not had time to swap the PSU. That said, this is a top brand 1200w so I'm not expeting that to be a factor when the folding fails on the first unit every time.

mark@titan:~$ stress-ng --cpu 0 --timeout 15m --metrics-brief
stress-ng: info: [14383] setting to a 15 mins, 0 secs run per stressor
stress-ng: info: [14383] dispatching hogs: 24 cpu
stress-ng: info: [14383] note: 24 cpus have scaling governors set to powersave and this can impact on performance; setting /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor to 'performance' may improve performance
stress-ng: metrc: [14383] stressor bogo ops real time usr time sys time bogo ops/s bogo ops/s
stress-ng: metrc: [14383] (secs) (secs) (secs) (real time) (usr+sys time)
stress-ng: metrc: [14383] cpu 32984624 900.00 21320.83 13.10 36649.54 1546.11
stress-ng: info: [14383] skipped: 0
stress-ng: info: [14383] passed: 24: cpu (24)
stress-ng: info: [14383] failed: 0
stress-ng: info: [14383] metrics untrustworthy: 0
stress-ng: info: [14383] successful run completed in 15 mins, 0.01 secs

## Diagnostic Summary & Final Steps
We have now tested every major component that can be reliably checked with software, and they have all passed with flying colors.
-Memory (RAM): Passed
-Graphics Card (GPU): Passed
-Processor (CPU) & Cooling: Passed
-Storage (NVMe SSD): Passed

When the CPU, RAM, GPU, and storage all test clean, suspicion falls on the two components that are very difficult to test with software: the Power Supply Unit (PSU) and the Motherboard.

[muziqaz]

run y-cruncher, which loads your system fully

[toTOW]

Does it happen with all WUs or only with this one ?