I Tested the Top Web Browsers' RAM Usage, Here's What I Found

I tested four browsers on the same MacBook Pro M3, same nine tabs, and one of them was eating 23 percent of total system memory while sitting still. The numbers are not what you would guess.

Chrome, Edge, Brave, and Firefox, two states each, four cross-checked memory tools, ninety-second stabilization windows, two runs averaged. This video walks through the full methodology, the cold-launch numbers, the nine-tab honest workday, and the part that surprised me: the gap between Chrome and Brave, two browsers running the same engine, is almost entirely the weight of the advertising economy. You are paying for ads in RAM.

Then the structural answer to why Chromium is so heavy: site isolation overhead post-Spectre, V8's greedy allocator, persistent service workers, and the cultural shift that happened when RAM got cheap. Firefox proves none of it is inevitable.

This is a benchmark video, but the conclusion is not "switch to Firefox." It is more uncomfortable than that.

━━━━━━━━━━━━━━━━━━
RELATED VIDEO ON THE CHANNEL
━━━━━━━━━━━━━━━━━━
Stop Shipping a Browser (The End of Electron) — the same Chromium tax measured here, but applied to every desktop app on your dock. Slack, Discord, VS Code, all shipping their own browser. If this video made you ask why the browser is the heaviest software you run, that one answers what happens when you stop accepting the answer.

━━━━━━━━━━━━━━━━━━
CHAPTERS
━━━━━━━━━━━━━━━━━━
0:00 850 MB doing nothing
0:50 Methodology and tools
1:52 Empty browser results
2:14 Chrome misreports its own RAM
2:40 Nine tabs, the honest workday
3:39 Brave's borrowed win
4:14 Why Chromium is so heavy
5:18 Firefox proves it is not inevitable
5:47 The verdict no one wants
7:00 The hidden Electron tax

━━━━━━━━━━━━━━━━━━
CONCEPTS AND TOOLS MENTIONED
━━━━━━━━━━━━━━━━━━
- Activity Monitor (macOS), htop, top, ps aux for cross-checked memory accounting
- chrome://memory-internals, chrome://discards, chrome://serviceworker-internals
- edge://discards, brave://memory-internals, about:memory, about:processes
- Site isolation (eTLD+1 process model) and Spectre/Meltdown mitigations (2018)
- V8 JIT cache, Blink, SpiderMonkey, Fission (Firefox process model)
- Service workers and persistent background registration
- The Chromium tax and Electron-shipped browsers in desktop apps

━━━━━━━━━━━━━━━━━━
FAQ
━━━━━━━━━━━━━━━━━━
**Which browser uses the least RAM in 2026?**
Firefox. In this benchmark it used 480 MB on an empty new tab and 2.9 GB on a nine-tab honest workday — roughly 30 to 43 percent less than Chrome on the exact same machine and content.

**Is Chrome still the worst browser for RAM?**
Yes. Chrome topped the chart at 850 MB cold-launched and 4.2 GB on nine tabs, crossing 23 percent of total system memory on an 18 GB MacBook Pro. Edge and Brave use the same Chromium engine and both came in lighter.

**Does Brave really save memory, or is it just the ad blocker?**
Mostly the ad blocker. Brave is built on Chromium, so by every architectural measure it should match Chrome. The 1.1 GB gap between them is almost entirely ad inventory and tracker scripts that Brave refuses to download by default. Install uBlock Origin in Chrome and most of the difference disappears.

**Should I switch to Firefox for performance?**
For RAM, yes — especially on 8 GB laptops or tab-heavy workflows. The trade-off is page paint feel: in this testing, JavaScript-heavy apps like Notion ran perceptibly slower in Firefox. Faster on memory, slower on micro-interactions.

**Why does Chrome use so much RAM in the first place?**
Three structural reasons: site isolation spawns one process per site after Spectre/Meltdown, each carrying 40 to 70 MB of fixed overhead; V8 allocates aggressively to cache compiled code; and service workers persist in the background after tabs close. The fourth reason is cultural — RAM got cheap, and the engineering pressure to be lean disappeared.

**Are these results trustworthy if Chrome misreports its own memory?**
That is exactly why this video uses four tools cross-checked. chrome://memory-internals undercounted by about 18 percent versus Activity Monitor on renderer and GPU processes. The OS-level numbers (top, htop, ps aux) were treated as ground truth.

#BrowserBenchmark #ChromiumTax #WebPerformance #FirefoxVsChrome #ProgrammingDeepDive

Видео I Tested the Top Web Browsers' RAM Usage, Here's What I Found канала Macro Lens