NVIDIA, Intel and AMD graphics hardware all includes a dedicated video encoding block - NVENC on NVIDIA, Quick Sync Video on Intel and AMF (Advanced Media Framework, using the VCN engine) on AMD. These chips encode H.264 and HEVC, and on recent generations AV1, many times faster than a CPU encoder while leaving the processor mostly free. The trade-off is that, at the same file size, a good software encoder such as x264 or x265 at a slow preset usually still delivers slightly better quality.
This guide explains what each vendor's encoder supports, how to check what your own PC can do, which settings matter, and when GPU encoding is the right choice over the CPU.
How hardware encoding works
A hardware encoder is a fixed-function circuit on the graphics card or processor. It is separate from the 3D shader cores used for games, so it can run alongside other work with little performance cost. Because it is fixed-function, it can only do what it was designed for: the codecs, profiles, bit depths and resolutions built into that generation of silicon. A driver update cannot add AV1 encoding to a card that lacks the hardware.
Hardware encoders make faster, simpler decisions than a slow software preset. They generally use fewer analysis passes and less look-ahead, which is why software encoders tend to win on quality per bit. Modern generations have narrowed that gap considerably, and for many uses - screen recordings, game footage, quick previews, high-bitrate intermediates - the difference is hard to see. For a direct comparison of the two approaches, see NVENC vs x264.
What each vendor supports
Support depends on the generation of the chip, not the brand alone. The table below is a broad summary; individual models within a family can differ, so always confirm against your exact GPU.
| Vendor | H.264 | HEVC | AV1 encode | FFmpeg encoder names |
|---|---|---|---|---|
| NVIDIA (NVENC) | GeForce 600 series onwards | Later GTX 900 cards onwards; much improved from RTX 20 / GTX 16 (Turing) | RTX 40 and RTX 50 series | h264_nvenc, hevc_nvenc, av1_nvenc |
| Intel (Quick Sync) | Core processors with integrated graphics since 2nd generation | Roughly 6th/7th generation Core onwards (10-bit from 7th) | Arc discrete GPUs and Core Ultra processors with Arc graphics | h264_qsv, hevc_qsv, av1_qsv |
| AMD (AMF / VCN) | Most Radeon GPUs and Ryzen APUs of the last decade | RX 400 series onwards and recent APUs | RX 7000 (RDNA 3) and RX 9000 (RDNA 4), plus RDNA 3-based APUs | h264_amf, hevc_amf, av1_amf |
A few practical points sit behind that table:
- NVIDIA consumer cards have historically limited how many encode sessions can run at once. The limit has been raised over time and varies by driver, so check if you plan to run several encodes in parallel.
- Intel Quick Sync is part of the integrated graphics. If your desktop has a separate graphics card, the integrated GPU may be disabled in the BIOS, and Quick Sync will not be available until it is enabled. Processors with an "F" suffix have no integrated graphics at all.
- AMD encoders improved markedly in recent generations, particularly for H.264, which older AMD cards handled less well at low bitrates than their competitors.
If you are deciding between codecs for GPU output, H.264 vs H.265 and AV1 vs HEVC cover compatibility and efficiency in more depth.
Check what your PC can actually encode
The quickest way to find out what FFmpeg can see on your machine is to list its encoders and filter for the hardware ones. In a Windows command prompt:
ffmpeg -hide_banner -encoders | findstr "nvenc qsv amf"
An encoder appearing in that list only means your FFmpeg build includes it. It does not prove that your hardware and driver support it. The reliable test is a short real encode: if the card cannot do the job, FFmpeg fails with an error such as "No capable devices found" or a message about an unsupported codec or profile.
Before blaming the hardware, update your graphics driver. Hardware encoders rely on the driver, and an old driver is the most common reason a supported card fails to encode, especially for AV1 and 10-bit HEVC.
Settings that matter
Rate control
Hardware encoders offer their own constant-quality modes rather than x264-style CRF. On NVENC this is variable bitrate with a constant quality target (-rc vbr -cq), on Quick Sync it is -global_quality, and on AMF it is typically constant QP or a quality-based VBR mode. The numbers are not interchangeable with CRF values, so run short tests rather than copying a CRF you use with x264. For streaming or strict size limits, CBR or VBR with a maximum bitrate are also available on all three.
Presets
NVENC uses presets from p1 (fastest) to p7 (best quality). Quick Sync and AMF have similar speed-versus-quality options. Slower presets cost relatively little time on a GPU, so for file encoding it is usually worth choosing one of the higher-quality presets. A typical NVENC HEVC command looks like this:
ffmpeg -i input.mp4 -c:v hevc_nvenc -preset p6 -rc vbr -cq 24 -b:v 0 -c:a copy output.mp4
Other options worth knowing
- Look-ahead and adaptive quantisation - NVENC's
-rc-lookahead,-spatial-aqand-temporal-aqcan noticeably improve quality on difficult content. - B-frames - supported on most modern hardware for H.264; HEVC B-frames need newer generations (Turing onwards on NVIDIA).
- 10-bit - HEVC and AV1 10-bit encoding is widely supported on recent hardware and helps reduce banding in gradients. Hardware H.264 is generally 8-bit only; 10-bit H.264 encoding has only appeared on the newest generations, such as NVIDIA RTX 50.
- Hardware decoding - using the GPU to decode as well (for example
-hwaccel cuda) can speed things up further, but it complicates filtering, so keep it simple until you need the extra speed.
When to use the GPU and when to use the CPU
| Situation | Better choice | Why |
|---|---|---|
| Live streaming or recording while gaming | GPU | Minimal impact on the game or CPU |
| Large batches where turnaround matters | GPU | Much higher throughput |
| Smallest file at the best quality | CPU (x265, SVT-AV1, x264 slow) | Better compression efficiency |
| Archival masters | CPU | Lossless and high-quality options are software-based |
| Previews, proxies, quick shares | GPU | Speed matters more than the last few percent of efficiency |
A sensible approach is to use the GPU by default and switch to a software encoder when file size or quality per bit is the priority, such as a final upload or a file you will keep for years.
Common problems
- Encoder not found - your FFmpeg build may not include it, or the driver is missing. Update the driver and check the encoder list.
- Unsupported profile or pixel format - for example asking an older card for 10-bit HEVC or 4:4:4 chroma. Fall back to 8-bit 4:2:0 or use the CPU.
- Blocky output at low bitrates - hardware encoders struggle more than software at very low bitrates. Raise the quality target or use a slower preset.
- Laptops with two GPUs - Windows may assign the app to the integrated GPU. Set the preferred GPU in Windows graphics settings if you need the discrete card.
Doing it with Movie Encoder
Movie Encoder supports NVIDIA NVENC, Intel Quick Sync and AMD hardware encoders when the PC supports them. It detects what your card and drivers can really encode, lets you choose CPU or GPU automatically or manually, includes an encoder benchmark and a configurable fallback, and only offers hardware AV1 when your hardware supports it.
Hardware constant quality, CBR, VBR and max bitrate settings sit alongside the usual software options, and the command preview shows the exact FFmpeg arguments, so you can see precisely which encoder and settings each job will use.
Frequently asked questions
Is GPU encoding lower quality than CPU encoding?
At the same file size, a slow software encoder such as x264 or x265 usually edges ahead, but modern hardware encoders are close. Raising the bitrate or quality target slightly on the GPU typically closes the gap.
Can a driver update add AV1 encoding to my graphics card?
No. AV1 encoding needs dedicated hardware, which arrived with NVIDIA RTX 40, Intel Arc and AMD RX 7000 series. Driver updates fix bugs and add features but cannot add a codec the chip lacks.
Why does FFmpeg list an NVENC encoder my card can't use?
The encoder list shows what your FFmpeg build was compiled with, not what your hardware supports. A short test encode is the reliable way to confirm.
