Modern CPU RAM is typically of the DDR4 variety, offering $$20$$ to $$25$$ GB/s bandwidth per module with a $$64$$-bit-wide bus. To achieve higher performance, memory modules are usually paired to utilize multiple channels. CPUs generally feature between $$2$$ and $$4$$ memory channels, resulting in a peak memory bandwidth ranging from $$40$$ GB/s to $$100$$ GB/s. Additionally, there are often two memory banks per channel; for instance, high-end processors like AMD’s Zen 3 Threadripper provide $$8$$ memory slots to accommodate these configurations.

CPU RAM Architecture and Bandwidth

Reading from RAM requires first sending an address to the memory module, a setup process that takes roughly $$100$$ ns. Once established, subsequent data transfers are exceedingly fast, taking only $$0.2$$ ns per $$64$$-bit record. When reading a long contiguous sequence of records—known as a burst read—the initial high cost is amortized. Conversely, isolated random reads require the $$100$$ ns setup every time, making the first read $$500$$ times more expensive than subsequent burst transfers. Consequently, while a system might perform up to $$10,000,000$$ random reads per second, computational efficiency dictates avoiding random memory access whenever possible in favor of burst reads and writes.

Burst Reads versus Random Access in Memory

Accessing main memory (RAM) is significantly slower than hitting the CPU cache hierarchy. For example, a $$64$$ MB memory reference on a local CPU takes approximately $$46$$ ns, while a larger $$256$$ MB reference takes about $$75$$ ns. If the memory must be accessed across a multi-socket system (a remote CPU reference), the latency increases substantially due to the interconnect hop, taking roughly $$70$$ ns for a $$64$$ MB reference and $$120$$ ns for a $$256$$ MB reference.

Main Memory Latencies

Random Access Memory (RAM) serves as the primary memory of a computer, providing fast access to data required during active computation. For deep learning tasks, RAM is essential for temporarily storing and retrieving training data, intermediate weight vectors, and neural network activations before they are processed by the CPU or GPU.

Claude

A standard computer utilized by deep learning researchers and practitioners typically includes several fundamental hardware components: a processor (CPU) for executing programs, memory (RAM) for storing intermediate results, a network connection (like Ethernet) for data transfer, a high-speed expansion bus (PCIe) to attach accelerators, and durable storage (HDD or SSD) for housing training datasets.

Computer Components in Deep Learning

Dive into Deep Learning

The Central Processing Unit (CPU) is the centerpiece of a computer system, responsible for executing machine code. A modern CPU is composed of several sophisticated components: multiple processor cores for execution, a communication bus connecting the cores, and a hierarchy of local caches that provide high-bandwidth, low-latency memory access. Furthermore, modern CPUs incorporate specialized vector processing units designed to accelerate the linear algebra and convolution operations that are foundational to machine learning workloads.

Central Processing Unit (CPU) in Deep Learning

Random Access Memory (RAM) in Deep Learning

When a single computational device is insufficient for deep learning optimization, systems rely on networks and buses to synchronize processing by transferring data. The selection of these interconnects depends on design parameters such as bandwidth, cost, distance, and flexibility. For example, while WiFi is inexpensive and flexible, its mediocre bandwidth and latency make it wholly unsuitable for machine learning clusters. Instead, distributed deep learning environments utilize high-speed connections—ranging from standard $$1$$ GBit/s Ethernet to $$100$$ GBit/s interconnects in cloud instances—to efficiently synchronize model parameters across multiple systems.

Network Connectivity in Deep Learning Systems

The Peripheral Component Interconnect Express (PCIe) is a dedicated high-speed expansion bus that forms point-to-point connections between the CPU and critical components like GPUs, network interfaces, and storage. It offers massive bandwidth, such as up to $$32$$ GB/s on PCIe 4.0 in a $$16$$-lane slot, with single-digit microsecond latency (e.g., $$5$$ μs). However, PCIe links are a precious resource because processors only support a limited number of lanes. For instance, AMD's EPYC 3 has $$128$$ lanes, Intel's Xeon has up to $$48$$ lanes per chip, while desktop-grade CPUs like the Ryzen 9 and Core i9 have $$20$$ and $$16$$ lanes, respectively. Because GPUs typically require $$16$$ lanes each and must share the bus with other bandwidth-intensive peripherals, the number of GPUs that can connect to the CPU at full speed is inherently limited. To minimize packet overhead during transmission across these limited links, deep learning systems prefer large bulk data transfers.

PCI Express (PCIe) in Deep Learning

Durable storage, such as magnetic hard disk drives (HDDs) or solid-state drives (SSDs), provides long-term retention of information. In deep learning workflows, these components are typically connected via the PCIe bus and are responsible for the efficient transfer of massive training datasets to the system, as well as for securely storing intermediate model checkpoints.

Learn Before

Related

Learn After