So.....what is this new AMD GPU?

It looks like AMD's Instinct MI300X GPUs are starting to ship.

With its launch, AMD is addressing the growing demand for more powerful computational tools in these sectors. The GPU's designated name, 'LamniAI,' suggests its primary application in running large language models (LLMs) for enterprise-scale tasks.

The MI300X is related to the MI300A and is engineered specifically for data center applications. While the MI300A was the industry's first data center-grade accelerated processing unit (APU) integrating general-purpose x86 CPU cores and CDNA 3-based parallel compute processors, the MI300X diverges in its design philosophy.

The MI300X omits the x86 CPU cores found in the MI300A, instead emphasizing its CDNA 3 chiplets. This shift results in the MI300X having 304 compute units, significantly increasing from the MI300A's 228 compute units. In terms of memory, the MI300X has 192 GB of HBM3 memory, supporting a peak bandwidth of 5.3 TB/s.

The development of the Instinct MI300X is part of a broader trend in GPU evolution, where companies like AMD and Nvidia continuously innovate to meet the escalating demands of modern computing.

A comparison with Nvidia's chipset evolution provides a perspective on the journey that AMD is now on:

• "Kepler" Series: K10, K20, K40, K80

• "Maxwell" M40

• "Pascal" GP100

• "Volta" GV100

• "Ampere" GA100

• "Hopper" GH100

• "Hopper-Next" H200

• "Blackwell" GB100

• "Xavier" GX100

I don't know about you, but I am excited about what is next!