AMD has released chips for AI tasks, mainly for the implementation of large language models or LLMs.
The Instinct M1300A (processor) and Instinct MI300X (accelerator) were presented for LLM management and training. The company announced that the memory characteristic of MI300X volume is 1.5 times bigger when compared with all the other versions. Also, both chips have significant improvements in energy efficiency and memory capacity.
MI300X was introduced on Azure according to the agreement between companies. Microsoft announced that the Azure ND MI300X VM is already available in preview. In addition, Meta also revealed information about the integration of MI300 processors.
For more productive data processing, APU chips combine GPU and CPU. The MI300A is a good choice for more effective computing, faster model training, and improved energy efficiency. AMD also noted that their new product has an increase in memory capacity. The availability of unified memory greatly simplifies the functioning of the system, as there is no need to move information from several devices. Pricing for the MI300A APU is not yet available.
In addition to Instinct M1300A and Instinct MI300X, the company also announced Ryzen 8040. This processor will have enough built-in AI functions. AMD emphasizes that the new series has a neural processor and higher performance when performing AI tasks. The processor has a higher performance when responding to video by almost 65 percent, and in games, this number is even higher when compared to competitors.
There is a high probability that in early 2024, Lenovo, Acer, Dell, Hewlett Packard, Asus, and Razer devices will be released with the Ryzen 8040 chip.
The Ryzen AI software platform has been announced by AMD, it is a great thing because developers using the Ryzen processor can create AI models. There will be access to the basic speech recognition models LLM, Whisper, and in particular Llama 2.