AMD's MI355X Outperforms NVIDIA's B300 in Open-Source AI Model
AMD's MI355X has outperformed NVIDIA's B300 in running the open-source AI model Kimi K3, achieving better performance per dollar. Kimi K3, with 2.8 trillion parameters, is a large model that requires significant VRAM, but AMD's MI355X offers a cost-efficient alternative to NVIDIA's hardware.
Key points
- AMD's MI355X has outperformed NVIDIA's B300 in running the open-source AI model Kimi K3.
- Kimi K3, developed by wafer.ai, has 2.8 trillion parameters and requires significant VRAM to run.
- AMD's MI355X offers a cost-efficient alternative to NVIDIA's B300, with a price per GPU that is 2.4 times cheaper on average.
- The MI355X has been used to achieve a performance of ~952 tok/s/node, making it a viable option for running large AI models.
- The use of MI355X for running Kimi K3 marks a significant milestone in the development of open-source AI models.
AMD's MI355X Outperforms NVIDIA's B300 in Open-Source AI Model
AMD's MI355X has made a significant breakthrough in the world of artificial intelligence (AI) by outperforming NVIDIA's B300 in running the open-source AI model Kimi K3. Developed by wafer.ai, Kimi K3 is a large model with 2.8 trillion parameters that requires significant VRAM to run.
The use of MI355X for running Kimi K3 marks a significant milestone in the development of open-source AI models. With its cost-efficient alternative to NVIDIA's hardware, AMD's MI355X has opened up new possibilities for organizations looking to develop and deploy large-scale AI models.
The performance of MI355X in running Kimi K3 is a testament to the capabilities of AMD's hardware. With a performance of ~952 tok/s/node, MI355X has proven itself to be a viable option for running large AI models.
The implications of this breakthrough are significant. With the increasing demand for AI models, the availability of cost-efficient hardware like MI355X will enable more organizations to develop and deploy large-scale AI models, leading to innovations in various industries.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.