A "roofline analysis" reveals that LLM performance is limited by the slower of two factors: the time it takes to fetch model parameters from memory (memory-bound) or the time it takes to perform matrix multiplications (compute-bound). Optimizing performance requires identifying and addressing the correct bottleneck.

A technique from cryptography, the Feistel network, makes any function invertible. When applied to neural network layers ("RevNets"), it allows activations from the forward pass to be re-calculated during the backward pass instead of stored. This trades extra compute for a massive reduction in memory footprint during training.

Separating inference into "prefill" (memory-bound) and "decode" (bandwidth-bound) tasks is a game-changer for hardware longevity. It allows older GPUs to be used for prefill tasks indefinitely, extending their useful economic life from 3-4 years to 10-15 years, a boon for data centers and their financiers.

The necessity of batching stems from a fundamental hardware reality: moving data is far more energy-intensive than computing with it. A single parameter's journey from on-chip SRAM to the multiplier can cost 1000x more energy than the multiplication itself. Batching amortizes this high data movement cost over many computations.

Spreading a model's layers across multiple GPU racks (pipeline parallelism) is a strategy to overcome memory capacity limits on a single rack. However, for inference, it offers no latency improvement; the total time remains the same. Its sole benefit is in memory capacity management for enormous models.

OpenAI achieved a major reduction in the cost of running its models through purely software and algorithmic improvements, such as quantization and smarter caching. This demonstrates that efficiency innovation can be as impactful as acquiring more hardware, suggesting a path to overcoming compute bottlenecks without relying solely on expensive chips.

Instead of running an entire inference task on a single GPU, the next efficiency leap will come from breaking it down. Tasks like prefill, attention, and feed-forward networks will be routed to specialized chips, such as SRAM-based accelerators, that are best suited for each job, dramatically improving performance and ROI.

The "Attention is All You Need" paper's key breakthrough was an architecture designed for massive scalability across GPUs. This focus on efficiency, anticipating the industry's shift to larger models, was more crucial to its dominance than the attention mechanism itself.

Model architecture decisions directly impact inference performance. AI company Zyphra pre-selects target hardware and then chooses model parameters—such as a hidden dimension with many powers of two—to align with how GPUs split up workloads, maximizing efficiency from day one.