Procesamiento mediante GPU¶

El procesamiento mediante GPU hará posible usar la tarjeta de video para el procesamiento, en vez de la CPU. Esto podría acelerar sus tiempos, dado que las GPU modernas se encuentran diseñadas para realizar una gran cantidad de cálculos numéricos. Por otra parte, también presentan algunas limitaciones al procesar escenas complejas, debido a una memoria generalmente más limitada, así como inconvenientes de pérdida de interactividad al usar la misma tarjeta gráfica para la presentación en pantalla y el procesamiento de imágenes.

Para activar el procesamiento mediante GPU, ir a Preferencias ‣ Sistema ‣ Dispositivos para procesamiento con Cycles y seleccionar CUDA, OptiX, HIP, oneAPI o Metal. Luego, cada escena deberá ser configurada para usar procesamiento por GPU en Propiedades ‣ Procesamiento ‣ Dispositivo.

Tecnologías de procesamiento¶

Blender soporta distintas tecnologías para el procesamiento mediante GPU, dependiendo del fabricante de la tarjeta y del sistema operativo usado.

CUDA – NVIDIA¶

CUDA is supported on Windows and Linux and requires a NVIDIA graphics cards with compute capability 3.0 and higher. To make sure your GPU is supported, see the list of NVIDIA graphics cards with the compute capabilities and supported graphics cards.

OptiX – NVIDIA¶

OptiX is supported on Windows and Linux and requires a NVIDIA graphics cards with compute capability 5.0 and higher and a driver version of at least 495.89. To make sure your GPU is supported, see the list of NVIDIA graphics cards.

OptiX takes advantage of hardware ray-tracing acceleration in RTX graphics cards, for improved performance.

GPU acceleration for OpenImageDenoise is available for compute capability 7.0 and higher, which includes all NVIDIA RTX cards.

HIP – AMD¶

HIP is supported on Windows and Linux and requires a AMD graphics card with the RDNA1 architecture or newer. Both discrete GPUs and APUs are supported.

Las GPU incluyen:

Radeon serie RX 5000
Radeon serie RX 6000
Radeon serie RX 7000
Radeon serie RX 9000
Radeon Pro serie W6000
Radeon Pro serie W7000

Versiones mínimas de los controladores:

Windows: Radeon software 24.6.1 o Radeon PRO software 24.Q2
Linux: Radeon software 23.40 o ROCm 6.0

Please refer to AMD’s website for more information about AMD graphics cards and their architectures.

Hardware ray-tracing support is available with the most recent drivers. This can be enabled in the preferences, and is supported on Radeon RX 6000 and newer.

GPU accelerated denoising is available on discrete Radeon RX 6000 and Radeon RX 7000 GPUs.

Shadow caustics are not supported on Radeon RX 5000.

oneAPI – Intel¶

oneAPI is a computation library that is supported on Windows and Linux and requires a Intel® Arc™ graphics card with the Xe HPG architecture. Hardware acceleration for ray-tracing and denoising is supported.

Las GPU incluyen:

Intel® Arc™ serie A
Intel® Arc™ serie B

Versiones mínimas de los controladores:

Windows: Controlador gráfico XX.X.101.5518 de Intel
Linux: intel-level-zero-gpu paquete 1.3.27642, normalmente disponible a través del paquete intel-compute-runtime XX.XX.27642.38

Please refer to Intel’s website for more information about Intel graphics cards and their architectures.

GPU accelerated denoising is available on all supported GPUs.

Metal – Apple (macOS)¶

Metal is supported on Apple computers with Apple Silicon. macOS 13.0 or newer is required to support all features.

GPU accelerated denoising is available on Apple Silicon.

Limitaciones¶

Path Guiding is not supported on any GPU.
Open Shading Language is only supported for OptiX, with some limitations listed in the documentation.

Preguntas frecuentes¶

¿Por qué Blender responde con lentitud durante el procesamiento?¶

On older GPU generations, graphics cards can only either render or draw the user interface. This can make Blender unresponsive while it is rendering. Heavy scenes can also make Blender unresponsive on newer GPUs, when using a lot of memory or executing expensive shaders, however this is generally less of a problem.

The only complete solution for this is to use a dedicated GPU for rendering, and another for display.

¿Por qué una escena que es posible procesar con la CPU no se puede procesar con la GPU?¶

There may be multiple causes, but the most common one is that there is not enough memory on your graphics card. Typically, the GPU can only use the amount of memory that is on the GPU (see Would multiple GPUs increase available memory? for more information). This is usually much smaller than the amount of system memory the CPU can access. With CUDA, OptiX, HIP and Metal devices, if the GPU memory is full Blender will automatically try to use system memory. This has a performance impact, but will usually still result in a faster render than using CPU rendering.

¿Es posible usar varias GPU para el procesamiento?¶

Yes, go to Preferences ‣ System ‣ Compute Device Panel, and configure it as you desire.

¿El uso de varias GPU aumentará la cantidad de memoria disponible?¶

Typically, no, each GPU can only access its own memory.

The exception is NVIDIA GPUs connected with NVLink, where multiple GPUs can share memory at a small performance cost. This is can be enabled with Distributed Memory Across Devices in the preferences.

¿Qué se procesa más rápido?¶

This varies depending on the hardware used. Different technologies also have different compute times depending on the scene tested. For the most up to date information on the performance of different devices, browse the Blender Open Data resource.

Mensajes de error¶

En caso de problemas, asegurarse de tener instalados los controladores oficiales obtenidos en el sitio web del fabricante de la GPU o, en Linux, a través del administrador de paquetes de la distribución. Los controladores gráficos proporcionados por el fabricante de la computadora podrían estar incompletos o ser antiguos.

Error: Memoria agotada¶

Normalmente significará que no existe suficiente memoria en la GPU para almacenar la escena durante el procesamiento.

Nota

Una forma de reducir el uso de memoria es reduciendo la resolución de las texturas. Por ejemplo, las texturas de imagen de 8k, 4k, 2k y 1k ocupan respectivamente 256MB, 64MB, 16MB y 4MB de memoria.

El controlador OpenGL de NVIDIA perdió conexión con el controlador gráfico¶

Cuando una misma GPU es usada tanto para el despliegue en pantalla, como para el procesamiento, Windows tiene un límite de tiempo para que la GPU haga cálculos de procesamiento. Si se tuviera una escena particularmente pesada, Cycles podría requerir demasiado tiempo de la GPU. Reducir la opción Tamaño de celdas en el panel Rendimiento > Memoria, podría aliviar este inconveniente, pero la única solución real es usar tarjetas gráficas distintas para la visualización en pantalla y el procesamiento.

Otra solución podría ser aumentar el tiempo de espera, si bien esto disminuirá la respuesta de la interfaz durante el procesamiento de escenas pesadas. Aprender más aquí.

Versión de GNU no soportada¶

On Linux, depending on your GCC version you might get this error. See the NVIDIA CUDA Installation Guide for Linux for a list of supported GCC versions. There are two possible solutions to this error:

Usar un compilador alternativo

If you have an older GCC installed that is compatible with the installed CUDA toolkit version, then you can use it instead of the default compiler. This is done by setting the CYCLES_CUDA_EXTRA_CFLAGS environment variable when starting Blender.

Launch Blender from the command line as follows:

CYCLES_CUDA_EXTRA_CFLAGS="-ccbin gcc-x.x" blender

(Substitute the name or path of the compatible GCC compiler).

Eliminar comprobaciones de compatibilidad

If the above is unsuccessful, delete the following line in /usr/local/cuda/include/host_config.h:

#error -- unsupported GNU version! gcc x.x and up are not supported!

This will allow Cycles to successfully compile the CUDA rendering kernel the first time it attempts to use your GPU for rendering. Once the kernel is built successfully, you can launch Blender as you normally would and the CUDA kernel will still be used for rendering.

Error de CUDA: Falló la compilación del núcleo¶

This error may happen if you have a new NVIDIA graphics card that is not yet supported by the Blender version and CUDA toolkit you have installed. In this case Blender may try to dynamically build a kernel for your graphics card and fail.

In this case you can:

Check if the latest Blender version (official or experimental builds) supports your graphics card.
If you build Blender yourself, try to download and install a newer CUDA developer toolkit.

Normally users do not need to install the CUDA toolkit as Blender comes with precompiled kernels.