Software 44590 Published by

MoltenVK 1.4.2-rc1 has launched as the first release candidate for Apple's Vulkan portability layer, bringing full Vulkan 1.4 compatibility to macOS, iOS, and tvOS. The update raises minimum system requirements to macOS 12.0 and iOS/tvOS 15.0 while introducing native sampler min/max filtering and gl_DrawID support for improved shader execution. Developers will see targeted fixes for older Mac 1 GPUs, AMD subgroup size corrections, and critical swapchain recreation patches, though some Apple Silicon channel corruption work was reverted pending further review. 



MoltenVK 1.4.2-rc1 Lands, Brings Vulkan 1.4 Support to Apple Platforms

MoltenVK 1.4.2-rc1 is officially out, marking the first release candidate for the Vulkan 1.4 update on macOS, iOS, and tvOS. It dropped yesterday, July 20, 2026, roughly eight months after the v1.4.1 stable release. The Khronos Group’s portability layer finally tackles a heavy backlog of GPU-specific regressions while raising the minimum system requirements.

Moltenvk

Raising the Floor and Adding Draw Indexing

Keep in mind that the minimum platform requirements just jumped. You now need macOS 12.0, iOS 15.0, or tvOS 15.0 to run this build. That effectively ends support for older Intel-based Macs and legacy iOS devices. Apple always phases out older Metal API support eventually, and MoltenVK is following that path.

The headline feature here is native sampler min/max filtering. It moves texture reduction conversion out of generic datatypes and straight into the pipeline. That should speed up certain drawing calls without forcing the driver to guess your workflow. You will also find the long-requested gl_DrawID and DrawIndex support for macOS. That brings Apple’s platform closer to parity with Windows and Linux for draw indexing.

The real work, though, is in the bug fixes. This release tackles a heavy backlog of GPU-specific regressions. The team disabled discarded-fragment store checks and combined sampler argument buffers on older Mac 1 hardware with NVIDIA or Intel graphics. AMD subgroup sizes got shifted back to 32 for correct simdgroup operations. Apple Silicon got a pass too, though the channel corruption fix for color render targets was actually pulled in this RC. The revert means that bug is still lurking.

MoltenVK is notoriously sensitive to shader translation, and this build shows exactly why. The KHR_external_semaphore_fd changes were reverted after MTLSharedEvent caused noticeable performance hits. SPIRV-Cross, the shader translation library, got updated and finally converted from a git submodule to a symlink for easier maintenance. A small change. But it saves contributors a lot of headache.

Several other fixes are buried in the commit history. Swapchain recreation for same-sized windows no longer drops down to 1x1 drawables. Cached Metal textures for swapchain images finally get invalidated correctly. Buffer and heap synchronization issues were patched, and argument buffer alignment padding was moved to the end of descriptor sets. Heavy lift behind the scenes.

Expect more release candidates before the final 1.4.2 drops. The team will need to figure out an alternative to the reverted semaphore changes and dig into why that Apple Silicon channel corruption fix failed. For now, you can grab the prebuilt tarballs for macOS, iOS, or the private API variant directly from the Khronos Group GitHub repository.

Head here to download the release.