Out of memory crashes with AE 2026.2 and Nvidia 5070

Hi everyone,

Quick update for those affected on this thread. ​@Roberto Tafuro has shared a set of workarounds that have made AE workable again for them on similar hardware. While we investigate the underlying issue, these are worth trying in the meantime:

Let me know how it goes. Thanks again for the patience and the detailed reporting.

Best,

Nishu

Hi everyone,

First, apologies for the delay in following up here.

The information on this thread we have collectively is substantial. I’ll escalate this issue and try to get someone from the product team to engage here.

A couple of asks that would help engineering move faster:

1. ​@Greg@Antifreeze Animation, Can you share a stripped-down version of a project that triggers the VRAM accumulation? Feel free to DM me the files if you prefer not to share them publicly.
2. For everyone affected: your exact AE version and NVIDIA driver version would help the team correlate.

Will follow up here as soon as I have something concrete from the team. Thanks for your patience and for the detail you've all contributed.

Best,

Nishu

I’m experiencing the same issue with a GTX4090 and 128GB of ram on a 32 cores in Windows 11 25h2. I’ve sent bug reports, i’ve opened a thread here, no solution and AE with no third party plugins is unusable. I’ve gone back to 2025. With a beast like the one i have, DaVinci is flying, while Adobe programs are basically useless. It’s a rant from a 20 years Adobe user.

It happens to me everyday, 4099 with 128GB of RAM and 32 cores. 

Bug Report: UI Panels Fail to Redraw (Black Screen) Post-VRAM Purge in v26.2.1

• OS: Windows 11

• Hardware: RTX 5070 (Dedicated VRAM: 12GB) | 128GB System RAM

• Driver: NVIDIA Studio Driver 591.44 (Rolled back to address previous stability issues)

• Mercury Transmit: Permanently DISABLED.

Steps to Reproduce & Isolation Results:

1. Baseline Test Setup: Confined the entire After Effects UI strictly to a single physical monitor to eliminate multi-monitor coordinate handoff issues. Secondary monitor was kept active for non-AE tasks only.

2. VRAM Accumulation: Performed standard timeline RAM previews until Dedicated GPU memory reached 10.5 GB / 12.0 GB total hardware utilization (with AfterFX.exe consuming ~6.3 GB of that allocation).

3. The Trigger: Executed an application purge command to flush memory.

4. The Failure State: Task Manager confirmed a successful VRAM release, dropping total dedicated GPU allocation back down to a stable 5.8 GB. However, the After Effects interface panels instantly blacked out and failed to redraw.

5. Workspace Isolation: Forcing a workspace layout reset to a single-monitor configuration ("Small Screen") altered the panel bounding boxes successfully, but the canvas elements within those panels remained entirely black and un-rendered.

Diagnostic Conclusion for Engineering:

The asset caching and VRAM garbage collection mechanics are functioning correctly (evidenced by the clean drop to 5.8 GB). The failure lies entirely within the v26.2.1 UI presentation engine. The application is dropping its underlying OpenGL/DirectX window drawing contexts during a memory flush and failing to re-initialize or resubmit those canvas texture handles to the Windows Desktop Window Manager (dwm.exe).

(Tested by Greg, using Gemini to walk through troubleshooting and write this up)

Noteworthy: Switching projects (without quitting) successfully refreshed screens.
Noteworthy: When AE project is isolated on one monitor, Purge>All Memory does indeed drop the Dedicated GPU memory. 

same thing happening to me. version 2024 worked absolutely fine, since 2025 and 2026 can’t work more than 20/30 min straight without purging memory or restarting. Even with the most simple scenes without any external plugins involved.

Interesting for the AE tech support:  
I opened AME to verify that the problem exists there as well. (AE is still open). 
AME crashed the render, but didn’t crash the app.
Closed AME. GPU ram released!
But… back in AE I now get black screens despite closing/opening the project, clearing all cache, etc. 
And GPU memory is back down to 2.6 (supposedly)!   Not healthy!

Noteworthy?  GPU maxes out, but Shared GPU Memory stays at zero… 
 

Tested, with no release of dedicated GPU
-- Delete suspect effects, including CC Force Motion Blur
-- Delete suspect .glb object
-- Open new project (without quitting After Effects)
-- Disable multi-core rendering
-- Render using Mercury Software-only (no hardware acceleration)

Disabling the Nvidia driver (via Windows Device Manager) does empty the memory, but also make AE hang.

My testing shows this issue does not exist for me in Blender. It is releasing GPU memory as expected.