“Resolve keeps crashing” is a symptom, not a diagnosis. A crash at launch, a freeze when one H.265 clip is imported, a Fusion failure on one node, and a whole-computer lockup during export need different tests. The fastest route is to protect the project, reproduce one trigger, and change one variable at a time.
Before touching drivers or reinstalling anything, write down the exact Resolve version and build, Free or Studio, operating system version, GPU, RAM, footage format, and the last action before the failure. That short record prevents the troubleshooting loop where five settings change and nobody knows which one mattered.
First decide whether it crashed, froze, or exhausted the system
| Symptom | What it means operationally | First useful test |
|---|---|---|
| Resolve disappears to the desktop | Application crash | Reproduce once, then create a diagnostics archive |
| Resolve says Not Responding but the OS remains usable | Application hang or a long operation | Watch whether CPU, disk, or GPU activity continues before force quitting |
| The display resets or the whole machine locks | Driver, hardware, power, thermal, or system-level failure is possible | Check OS event and system logs, temperatures, and stability outside Resolve |
| Only one render stops or fails | Delivery-path failure, not automatically an application crash | Use the render-failed workflow |
| Playback drops frames but Resolve stays open | Performance problem | Use the slow-playback diagnosis |
A busy application can look frozen while it builds waveforms, analyzes media, loads a large project, or processes a heavy effect. Give the process a short observed interval and watch CPU, disk and GPU activity rather than repeatedly clicking the window. A repeatable hang still matters, but describe it accurately. If the symptom doesn't fit this table, use the Resolve troubleshooting map before changing unrelated settings.
Protect the project before experimenting
If Resolve still opens, duplicate the affected timeline and choose File > Export Project to create a DRP. The Resolve 21 Reference Manual says standard Export Project omits stills and LUTs; a separate Export Project with Stills and LUTs command includes them. Neither DRP option bundles source media or generated cache, so preserve those separately. For important work, also back up the project library from Project Manager.
Open Preferences > User > Project Save and Load and verify both controls. In Resolve 21, Live Save is on by default and saves changes incrementally. Project Backups works independently, keeps timed project versions, and opens a recovered backup as an independent project rather than overwriting the original. Check the retention schedule and confirm that the backup location is available.
Don't reset preferences, delete a project library, or remove application-support folders until you have a restorable copy. A troubleshooting step isn't safe merely because it appears in a short video.
Build a reproducible crash note
Open a plain text file and capture one failure in this format:
- Resolve 21.x.x build number, Free or Studio.
- Operating system version and recent OS update.
- GPU model, driver version, VRAM, RAM, and free disk space.
- Project and timeline name.
- MediaInfo details for the source used at the crash point.
- Exact action: “Open Color page, select clip 43, enable temporal noise reduction.”
- Result: crash to desktop, hang, display reset, or full system restart.
- Whether the same action fails again after relaunch.
If it can't be repeated, note what was different before the next occurrence. Intermittent failures often become understandable after two or three observations reveal the shared effect, media source, display state, or background task.
Use the trigger to choose the branch
| Repeatable trigger | Best isolation test | Evidence it produces |
|---|---|---|
| Resolve crashes before the Project Manager appears | Disconnect optional hardware, test a clean OS user where practical, and inspect launch logs | Separates application startup from one project |
| One project crashes while others open | Open a backup or duplicate, then disable the last suspect effect or media group | Project-specific boundary |
| Importing one folder crashes | Add small batches, then individual files | Identifies one malformed or unsupported asset |
| One effect or Fusion comp crashes | Bypass the effect, node group, or third-party plugin | Effect/plugin/GPU branch |
| Moving or minimizing the window crashes | Test one display, remove overlays, and compare the current certified GPU driver | Display/driver/window-state branch |
| Export crashes | Render a short range around the repeatable frame | Separates one clip/frame/effect from general encoding |
Test a new project before blaming the installation
Create an empty local project, add one known-good file, make a simple timeline, and play it. If the clean project is stable, the installation can launch, decode that file, and run the basic edit path. Return to a duplicate of the problem project and isolate what it adds.
If the clean project also fails, the scope shifts toward the application build, GPU driver, system resources, optional hardware, preferences, or OS. Reinstalling the same build without collecting evidence may change nothing because media, drivers, plugins, and preferences remain outside the application bundle.
Isolate media without converting the whole job
A crash on import or at the same timeline frame can be caused by one incomplete image, damaged video, unusual audio stream, or unsupported codec profile. Copy the suspect item to a local drive, inspect it with MediaInfo, and test it in an empty project. Compare the actual video and audio streams with Blackmagic's codec matrix rather than trusting the filename extension.
When a folder import crashes, use a binary search: import half the files, then half of the failing half, until one asset remains. Blackmagic Support has recommended the same divide-and-conquer method for a repeatable folder-import crash. Don't delete the suspect source. Move a copy out of the test set and preserve the original for recovery or vendor analysis.
If a phone or screen recording imports but crashes during seeking, create one constant-frame-rate intermediate and repeat the same action. If the intermediate is stable, you have evidence about the source path. You don't yet have a reason to transcode every unrelated camera file.
Separate GPU memory pressure from a driver failure
Noise reduction, Magic Mask, Super Scale, optical flow, large Fusion frames, many OpenFX plugins, and high-resolution timelines can increase GPU memory and processing pressure. If Resolve reports GPU memory full, follow the dedicated GPU memory workflow.
A display reset, driver timeout, or whole-system lock needs a broader check. Record the driver version, compare it with the current support notes and Resolve system requirements, and inspect the operating system's event or crash report. On Windows, test the appropriate supported driver branch for the installed Resolve build and GPU when one is specified. On macOS, GPU drivers arrive with the OS, so a Windows clean-driver tutorial doesn't apply.
Monitor temperatures and system stability during the repeatable test. Resolve can expose marginal cooling, power, RAM, GPU, or overclock stability because it loads several subsystems at once. That doesn't prove Resolve caused the hardware fault. Remove overclocks and test the machine at known-stable defaults before drawing conclusions.
Disable third-party plugins as a controlled test
If the crash begins when loading a title, transition, audio effect, LUT tool, or OpenFX plugin, duplicate the timeline and bypass that item. Update the plugin from its vendor only after confirming compatibility with your Resolve and OS versions.
A plugin can also fail while Resolve scans it at launch. Temporarily moving one suspect plugin out of its documented scan path can be a useful test, but record the original location and change only one product at a time. A current Maxon support example uses Preferences > Video Plugins and the product's own plugin cache for Universe. Don't apply product-specific file paths to unrelated plugins.
Clear only the cache tied to the failure
Render cache and optimized media are generated files. A damaged cache entry can reproduce a problem at one frame, so deleting the cache for the selected clip or timeline is a reasonable controlled test. Deleting every cache on every disk isn't a universal crash fix and can turn a short diagnosis into hours of regeneration.
Check free space and write access on the cache drive. A full or disconnected cache location can create failures that look like a bad effect. Keep source media, cache, proxies, and project-library backups conceptually separate so removing generated files never removes the only copy of a source.
Update or roll back with a rollback path
Check Blackmagic's Resolve support page for the current build and release notes. If the notes name your failure or your build is far behind, an update may be justified. If crashing began immediately after an application, OS, driver, or plugin update, the previous known-good combination is useful evidence.
Before changing Resolve, back up the project library and export important projects. Newer project libraries may not be usable in older Resolve versions. The safe update procedure covers the backup and verification sequence.
Create diagnostics while the failure is fresh
After reproducing the crash, reopen Resolve, open any project if needed, and choose Help > Create Diagnostics Log on Desktop. The command is documented in Resolve 21, and Blackmagic Support asks for that archive together with system information and, for project-specific failures, a DRP plus exact reproduction steps.
On Windows, save a System Information NFO rather than a text export when support requests it. On macOS, save the System Report. Don't post diagnostic archives publicly without reviewing the privacy implications. They can contain machine details, paths, plugin information, and project context. Share them through the support route you intentionally choose.
Know when to stop changing settings
Escalate with evidence when a clean project and known-good media still reproduce the failure, the same action crashes after a supported driver and application update, or the issue is tied to a specific project that survives a backup restore. Send the exact build, system report, diagnostics archive, minimal project or sample media where licensing permits, and concise steps.
For a whole-system restart, thermal shutdown, storage errors, memory-test failure, or repeated crashes in other GPU-heavy applications, troubleshoot the system before continuing production. An NLE preference reset cannot fix unstable RAM or a failing drive.
Frequently Asked Questions
Why does DaVinci Resolve keep crashing?
Common branches include one damaged or unsupported media file, GPU memory pressure, a driver or display failure, a third-party plugin, one project or effect, a full cache drive, and an application bug. Reproduce the exact trigger before choosing a fix.
Will clearing the cache stop Resolve from crashing?
Only when generated cache media is part of the repeatable failure. Delete the cache for the selected clip or timeline as a controlled test. A launch crash, plugin scan failure, unsupported source, or unstable GPU will not be repaired by clearing unrelated cache files.
Should I reinstall DaVinci Resolve?
Not as the first step. Test a clean project and known-good media, collect diagnostics, and isolate plugins, drivers, and project-specific factors. Reinstalling the same build may leave the real cause unchanged.
How do I get DaVinci Resolve crash logs?
After the failure, reopen Resolve and choose Help > Create Diagnostics Log on Desktop. If Resolve can't reach Project Manager, use Blackmagic's platform-specific CaptureLogs utility for your installed build. Keep the archive private unless you intentionally share it with support.
Can low RAM or VRAM make Resolve crash?
Resource pressure can contribute, especially with high-resolution media, Fusion, noise reduction, AI tools, and large frame buffers. Record RAM and VRAM use around a repeatable trigger, but don't treat capacity as proof when a driver, plugin, damaged file, or hardware fault can produce a similar failure.
Related guides
Use the Resolve troubleshooting hub when the trigger points away from a general crash. An explicit VRAM warning belongs in the GPU memory workflow, while a repeatable delivery failure belongs in the render-failed diagnosis.
