Qualcomm Patents a Graphics Chip That Converts Video Data Mid-Process Without Pausing
Every GPU juggles two incompatible ways of storing image data, and switching between them usually means stopping to reformat everything first. Qualcomm's new patent describes hardware that handles that conversion in the background, while the chip is already working.
What Qualcomm's on-the-fly format conversion actually does
Every time your phone processes a photo, a video frame, or a camera feed, the image data is packed into memory in one of two ways: either all the color channels mixed together (called interleaved), or each color channel stored separately (called planar). Different apps, cameras, and AI models often expect different formats, and that mismatch creates extra work.
Right now, graphics chips often have to pause and reformat the data before they can start the real processing. That's a small but real cost, and it adds up fast when your phone is handling a live video stream, running a background AI model, or editing photos in real time.
Qualcomm's patent describes a GPU that performs this format conversion while it's already processing the data, rather than treating it as a separate step. The chip detects it's working with interleaved data and flips it to the planar layout on the way through, then signals that the conversion is done.
… perform, during a processing of the interleaved data, a conversion of a layout for the interleaved data to a layout for planar data …
Translation: It changes how video data is organized while the chip is actively working on it.
How the GPU swaps data layouts during processing
At a technical level, the patent describes a processor (specifically a GPU or similar graphics/compute chip) that can handle two common memory layouts for image or graphics data:
- Interleaved format: color values for each pixel are stored together, one after another (think R, G, B, R, G, B across a row). This is common from cameras and encoders.
- Planar format: all the red values are stored together, then all the green, then all the blue. This is often what AI inference engines and video processing pipelines prefer.
Normally, converting between these two layouts requires a dedicated preprocessing pass, meaning the GPU has to read the data, rewrite it somewhere else in memory, then start the actual work. That's extra memory bandwidth and extra latency (delay).
This patent's approach embeds the conversion inside the processing pipeline itself. The chip receives a signal that the incoming data is interleaved, performs the layout conversion during the processing step rather than before it, and then outputs a signal confirming the conversion happened. No separate reformat pass, no second round-trip to memory.
The patent covers the hardware logic to do this, the control signals that trigger and confirm the conversion, and the architecture for feeding the result downstream in the correct planar layout.
Aspects presented herein relate to methods and devices for data or graphics processing including an apparatus, e.g., a GPU.
Translation: The patent describes a graphics processor designed to handle video formats more efficiently.
What this means for phones handling video and AI workloads
For phones and tablets, this kind of efficiency matters most in scenarios where data is moving fast: live camera processing, real-time video filters, or AI models running on the GPU. Each unnecessary memory read/write burns power and adds delay. Qualcomm's run of mobile GPU efficiency filings points to steady pressure to squeeze more out of the same battery budget.
For you, the benefit would be subtler than a spec-sheet number: slightly faster photo processing, smoother video, or an AI feature that drains less battery because the chip isn't doing redundant work behind the scenes. It's an infrastructure improvement rather than a visible feature, but that's exactly where phone performance gains are made today.
That makes this Qualcomm's 12th filing we've tracked since July in our GPU rendering race watchlist, building on their work around faster screen rendering and sharp rotating graphics.
Format mismatches between camera sensors, graphics processors, and AI chips are a persistent, invisible tax on every image your phone touches. Each time a video frame passes from one component to another in the wrong data arrangement, the chip has to stop, rearrange the bytes, and start again, burning time and battery at rates that add up quickly across a continuous camera feed.
Absorbing that rearrangement directly into the graphics hardware, rather than treating it as a separate chore, is a well-proportioned answer to a well-scoped problem. It does not reinvent anything fundamental; it closes a specific gap that affects nearly every image-processing task a modern phone runs.
How much this matters in practice depends on how often your device hits that gap, and on continuous camera or video workloads, that is very often.
There are more where this came from
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings
15 drawing sheets from US 2026/0260310 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →