Skip to content

Intel i9-13900K / 14900K instability: BIOS workarounds for Metashape crashes

  • Status: unverified
  • Applies to: Metashape Pro 2.x and Metashape Standard 2.x — same hardware-level issue, independent of Metashape version
  • Edition: Pro / Standard
  • Diátaxis: how-to
  • Confidence: high
  • Last reviewed: 2026-05-29

Confidence: high. The CPU instability is widely attested in the Agisoft forum and across other CPU-heavy workloads (Unreal Engine shader compilation, scientific Python, BMW renders); BIOS workarounds are forum-attested with three independent users reporting success. This is a hardware-platform issue, not a Metashape bug.

Problem

Your workstation runs an Intel i9-13900K, i9-14900K, or similar 13th/14th-gen "K"-series desktop CPU. Metashape crashes inconsistently during Align Photos (specifically the "estimating camera locations" sub-stage) or Optimize Cameras. The crashes appear random — the same project may align cleanly one run and crash another with identical parameters. Other CPU-heavy applications may show similar instability.

This is not a Metashape bug. Intel acknowledged in 2024 that 13th and 14th-gen K-series CPUs have a microcode-level instability related to elevated voltage / clock requests under sustained load, particularly affecting heavily-binned high-end SKUs. The issue manifests across many applications; Metashape's Estimate Camera Locations stage is among the most affected because it sustains 100% CPU on all cores for minutes-to-hours.

Diagnostic ladder

Before applying BIOS workarounds, rule out conventional hardware faults. Agisoft support's published diagnostic order:

  1. Memtest86 — multithreaded RAM check (run from bootable USB). Test each module individually to isolate failures. Tip: Metashape's own access pattern can stress RAM more than Memtest, so a clean Memtest run does NOT rule out RAM instability.
  2. Intel Processor Diagnostic Tool — official Intel CPU stress test, free download.
  3. Prime95 with AVX2 — community-standard CPU stress test that exercises the AVX2 instructions Metashape uses heavily.
  4. Intel Extreme Tuning Utility (XTU) → Stress Test → AVX2 — Intel's tuning + diagnostic suite.

"If such random crashes occur during CPU/RAM-intensive steps, it is recommended to: - perform multithreaded Memtest86 checks, - confirm if crashes happen regardless the number of RAM modules used, - try reducing RAM frequency in BIOS, - try CPU stress test using Prime95 (with AVX2), - try Intel Processor Diagnostic Tool, - try Intel XTU CPU stress test." — Agisoft support, 2023-10-31, Metashape 2.1 (permalink)

If all four diagnostic tools pass cleanly and Metashape still crashes, the BIOS workarounds below are the most reliable remediation reported in the forum.

BIOS workarounds (forum-attested)

Each workaround restricts the CPU's maximum sustained clock or voltage slightly, which prevents the instability that triggers crashes. The cost is a 5-15% performance reduction on heavily- threaded workloads.

Workaround 1 — Cap Turbo Power Limits (PL1 / PL2)

In BIOS, set Turbo Power Limits (PL1 and PL2 / "long-duration" and "short-duration" power caps) to fixed values rather than the motherboard's default AUTO setting. Recommended caps:

  • PL1 (long-duration): 253 W (Intel's official 13900K/14900K base TDP) — this is what the chip is rated for sustained.
  • PL2 (short-duration): 253 W — same as PL1, eliminating the brief Turbo overshoot.

This is reported to fix many crashes:

"I think it could be a setting in the Gigabyte BIOS that is causing the issues. In the BIOS, under CPU settings: 'Turbo Power Limits' is default on 'AUTO' [...] Changing from 'AUTO' to 'Intel POR' currently seems to be rectifying the issue. This has worked for me on a couple of small (700-800 image) datasets, now trying on a 63k image dataset." — 3dMB Ltd, 2024-01-19, Metashape 2.1 (permalink)

The setting name varies by motherboard:

Vendor Setting path
ASUS Ai Tweaker → Internal CPU Power Management → Long Duration Power Limit / Short Duration Power Limit
Gigabyte Tweaker → Advanced CPU Settings → Turbo Power Limits
MSI OC → CPU Specifications → Long Duration Power Limit / Short Duration Power Limit
ASRock OC Tweaker → CPU Configuration → Long Duration Power Limit

Workaround 2 — Cap the Performance Active-Core Tuning Ratio

In Intel XTU (or BIOS, where exposed), set the Performance Active-Core Tuning Ratio (per-core multiplier) to a value below the chip's stock turbo. For 13900K stock turbo is 58×; capping to 53× has been reported as stable:

"I just wanted to provide an update that I was able to fix the crashing issue by going into the Intel(R) extreme tuning utility and setting the Performance Active-Core Tuning Ratio of all of the cores to 53x." — colbyrand, 2024-04-02, Metashape 2.1 (permalink)

This caps each core's turbo to 5.3 GHz (53 × 100 MHz BCLK) instead of the stock 5.8 GHz. The 8.6% clock reduction substantially improves stability under sustained load.

Workaround 3 — Disable Thermal Velocity Boost

If both workarounds above don't suffice, disable Thermal Velocity Boost (TVB) entirely in BIOS. TVB is the most aggressive opportunistic-clock feature on K-series CPUs and is the first to trigger the instability under thermal/power stress.

Workaround 4 — Replace the motherboard

In one user's case, replacing the Gigabyte Z790 motherboard with an ASUS Prime Z790-P fully resolved crashes that no BIOS adjustment fixed:

"I tried to replace every part of the system, frames, graphics card, disk, but with no results. Only replacing the motherboard with an ASUS Prime Z790-P allowed it to work without errors." — geodamkru, 2023-11-23, Metashape 2.1 (permalink)

This points to motherboard-level VRM (voltage regulator module) quality / firmware as a contributing factor on certain Gigabyte Z790 boards. Before attempting motherboard replacement, ensure the board's BIOS is up-to-date — Intel released microcode patches in mid-2024 (Intel Reference Code 0x125+) that address some of the instability on the BIOS side.

Verifying the fix

After applying a BIOS workaround:

  1. Reboot. BIOS power-limit changes don't take effect until reboot.
  2. Re-run the failing project. Use the same dataset that reproduced the crash; expected behaviour is a complete alignment with no crash.
  3. Monitor with HWInfo64 (Windows) or s-tui (Linux) during alignment. Watch CPU package power draw — confirm it stays below the new PL1 limit. Watch CPU package temperature — it should stay below 90°C under sustained load.

Crash-report locations for Agisoft support

If you submit a crash report to Agisoft support, the dump files are at:

OS Path
Windows C:\Users\<username>\AppData\Roaming\Agisoft\Metashape Pro\Crash Reports\pending\
Linux ~/.agisoft/metashape pro/Crash Reports/pending/
macOS ~/Library/Application Support/Agisoft/Metashape Pro/Crash Reports/pending/

Attach both the .dmp and .extra files for each crash — support uses both to determine the failure mode (CPU exception, GPU driver crash, RAM corruption, etc.).

Caveats

  • The workarounds reduce maximum performance ~5-10% in exchange for stability. On batch-processing workstations this is usually an acceptable trade. For interactive work with smaller projects, the stock clocks may still work fine.
  • Intel microcode updates (mid-2024) addressed some causes of the instability at the firmware level. Users on the latest BIOS may not need manual workarounds. Check your motherboard vendor's BIOS release notes for "Intel microcode 0x125" or newer.
  • The thread is still active. Behaviour on newer Intel generations (Core Ultra Series 1 / Series 2 desktop) is uncharacterised in this thread. If you have a Core Ultra workstation that exhibits similar symptoms, search the forum for "Core Ultra crash" before assuming the same fix applies.
  • DDR5 RAM stability is a separate concern. Some users in the source thread reported that RAM frequency / module count affected stability:

"The more RAM sticks you use - the bigger chance of RAM instability and the lower maximum frequency at which memory is stable." — PolarNick, 2023-10-16, Metashape 2.1 (permalink)

4×32GB DDR5 sticks at the kit's rated XMP speed often run unstable; reducing to 2×32GB or dropping the XMP profile to default JEDEC speeds can help. This is a general DDR5 motherboard / CPU memory-controller limitation, not a Metashape issue. - AMD Ryzen 7000 / 9000 series does NOT exhibit this particular instability. If you are buying a new workstation primarily for Metashape stability, the AMD platform may be a lower-risk choice in 2024-2025.

See also

References