AMD is making available the new version of its open software, ROCm 6.1.3 – ensuring the strategy of expanding its compatibility with its GPU catalog, including Radeon cards. The intention is to allow everyone to run artificial intelligence workloads on ROCm.
Among the features presented in this latest version, the focus is to further increase its accessibility, scalability and make a wider range of hardware compatible with the software. Check out the main changes below:
- Multi-GPU support, enabling the construction of scalable AI desktops for multi-server and multi-user solutions;
- Beta-level support for the Windows subsystem on Linux, allowing these ROCm solutions to work on a Windows OS-based system;
- Support for the TensorFlow Framework, offering more options for developing artificial intelligence
AMD ROCm 6.1.3 can now also support four Radeon RX or Radeon PRO GPUs simultaneously, allowing you to scale up configurations with data parallelism – where each GPU acts independently on computational inference and generates a response.
New AMD ROCm extends functional parity from data centers to desktops, enabling artificial intelligence research and development on readily available and affordable platforms
Andrej Zdravkovic, senior vice-president of AMD
With the arrival of the AMD ROCm 6.1.3 version, the user will be able to access the Windows subsystem on Linux, also known as WSL 2, at Beta level. In simpler terms, this will allow you to run Linux-based AI tools on the Windows system without the need for a dedicated Linux system or dual-boot setup.
In addition, qualification for the TensorFlow framework was obtained, bringing a new AI development option in addition to the well-known PyTorch and ONNX – extending support to the open ecosystem of frameworks and libraries.
See more:
A chegada da GPU AMD Radeon PRO W7900 Dual Slot
Among the compatible GPUs with AMD ROCm 6.1.3 software support is the all-new Radeon PRO W7900 Dual Slot. In terms of specifications, it has the Navi 31 XTX chip with 6,144 cores and 96 computational units. It has a 384-bit memory bus and comes with 48 GB GDDR6 memory – which reaches up to 864 GB/s of bandwidth.
The Dual Slot design choice is ideal for use on servers and also features 96MB of Infinity Cache and 192 AI accelerators – featuring one of the manufacturer’s most capable GPUs. Its suggested retail price is $3,499 and is made to work alongside large LLMs with speedy performance.
AMD’s intention is that users can equip the GPU in their workstation and have high inference accuracy in LLMs. An example are the LLaMA-2 and LLaMA-3 models, with 70B parameters quantified in INT4 and which require at least 35GB of GPU memory – making the Radeon PRO W7900 a suitable option.
Fonte: TechPowerUp
Join the Adrenaline offer group
Check out the main offers on hardware, components and other electronics that we found online. Video card, motherboard, RAM memory and everything you need to build your PC. By joining our group, you receive daily promotions and have early access to discount coupons.
Join the group and take advantage of promotions
Source: https://www.adrenaline.com.br/amd/amd-rocm-6-1-3-permite-colocar-4-placas-radeon-juntas/
