Chrome crashes at 4,500 tabs despite 1.5 terabytes of RAM

Google Chrome has a well-earned reputation for swallowing system memory. In an ambitious stress test designed to find the absolute breaking point of modern web architecture, Linus Sebastian pushed a high-end enterprise system to its absolute limits. The mission was straightforward: determine how many active browser tabs a machine packing a massive 1.5 TB of DDR5 RAM and dual 96-core AMD EPYC 9684X processors could run before collapsing.

Rather than relying on manual clicking, the test used a custom Python-based automation script to open tabs in controlled batches. This approach systematically isolated hardware performance from human error. It also highlighted a sobering reality: modern web optimization is struggling to keep pace with raw hardware development.

Upgrading to Server-Grade Horsepower

To eliminate the hardware bottlenecks that plagued previous tests, the team made several critical architectural upgrades. First, they deployed dual AMD EPYC processors, tripling their previous core count to 192 physical and logical threads. This massive processing power tackled the multi-threaded load of running thousands of isolated browser processes.

Memory capacity was another key focus. Instead of the complex Load-Reduced DIMMs (LRDIMMs) used in earlier experiments, the team used twelve 128 GB DDR5 registered ECC modules. These newer modules handle extreme capacities much more efficiently by splitting the traditional 72-bit ECC channel into two independent 40-bit channels. This architectural change directly filters out signal noise without sacrificing latency.

Windows vs. Linux Performance Breakdown

The choice of operating system proved to be the most decisive factor in the experiment. The performance profiles of Windows 11 and Ubuntu diverged almost immediately under heavy loads:

• Windows 11: The OS struggled early with memory management. After passing 1,500 tabs, memory allocation flattened out. The system stopped loading page content properly and eventually hard-crashed at roughly 3,000 tabs.
• Ubuntu Linux: Linux handled the load much more gracefully. Its memory usage climbed linearly, indicating that it was actually rendering page content rather than quietly suspending processes. The team also tweaked the OS configuration, increasing the system's open file limit to 65,535 to prevent early file-descriptor bottlenecks.

The VRAM Bottleneck and Software Crashes

Despite the massive 1.5 TB system memory pool, the experiment was ultimately cut short by an unexpected bottleneck: video memory (VRAM). The server was running a basic graphics card designed primarily for video output rather than heavy rendering. As thousands of media-rich tabs requested GPU acceleration, the card quickly ran out of physical VRAM.

This resource exhaustion triggered severe UI lag and unstable rendering. Eventually, it caused a hard application crash on both operating systems. Even on Linux, which handled the resource load much better than Windows, Chrome collapsed at roughly 4,500 tabs. This was well short of the team's historical record, proving that modern web pages have become significantly heavier and more resource-intensive.

Web Bloat and the RAM Paradigm

This experiment offers a clear takeaway for tech enthusiasts: hardware is no longer the main limiting factor for extreme multitasking. Instead, the bottleneck has shifted to browser architecture and the sheer weight of modern web pages. The heavy script payloads, tracker networks, and security layers like Cloudflare make today's websites vastly more demanding than those of five years ago.

For users looking to optimize more modest systems, technologies like Linux's ZRAM offer an elegant solution. ZRAM dynamically compresses data directly in memory, helping systems avoid sluggish pagefile swaps to physical SSDs. However, as long as web bloat continues to outpace software optimization, even the most powerful enterprise systems will eventually find their limits.