YuvalGGYuvalGG

Tools

Memory test

Test your RAM in the browser: read, write and copy speed, memory latency, and a chart that shows where the CPU caches end and main memory starts. No download, nothing uploaded.

Memory test

Measures read, write and copy speed and memory latency from JavaScript in one Web Worker. Everything runs in this tab; nothing is uploaded. Browser numbers are much lower than native benchmarks, and the score is relative and uncalibrated.

RAM reported by the browser

—

Rounded; Chrome tops out at 32 GB (8 GB on Android)

Test size (memory the tab uses)

Bigger sizes reach further past the CPU caches. If the browser refuses, the test halves the size and tells you.

Takes about 10 seconds:

  1. Read speed
  2. Write speed
  3. Copy speed
  4. Latency sweep (16 KB up to the buffer size)

Close games and heavy apps, keep this tab in front, then press Start.

One browser thread cannot use all the bandwidth your memory has, so native tools such as AIDA64 show much higher GB/s. This test does not check your RAM for errors. Compare runs on the same browser and machine, before and after a change.

How to use the memory test

  1. Close games, video exports and extra browser tabs. Plug a laptop into power, because many laptops run slower on battery.
  2. Pick a test size. 256 MB suits most PCs; 512 MB reaches further past the caches; 64 MB is gentler on phones.
  3. Press Start memory test and keep the tab in front for about 10 seconds. If you switch tabs, the test pauses, because browsers slow down background tabs. Press Resume to redo the step that was interrupted.
  4. Read the read, write and copy speeds, the latency, and the chart. Press Copy results to paste them into a forum post or a note.
  5. Change one thing, such as enabling XMP or EXPO or switching the power mode, and run the test again in the same browser.

What the test measures

The page reserves one large block of memory inside a Web Worker, a piece of JavaScript that runs on its own thread, and times four jobs on it:

  • Read adds up every number in the block, so the processor has to pull all of it from RAM.
  • Write fills the whole block with a new value.
  • Copy copies the first half of the block onto the second half. The speed counts each byte copied once.
  • Latency follows a chain of random jumps, one per 64-byte line, so the processor cannot guess the next address. The time per jump is the wait for one piece of data. The test repeats this for block sizes from 16 KB up to 256 MB (or the size you picked, if smaller).

Speeds are in GB/s, where 1 GB is a billion bytes. Each speed is the middle (median) result of all the passes in 1.5 seconds. The 0–100 score puts read, write and copy speed and the latency of the biggest block on a log curve. The curve is relative and uncalibrated: it is not based on any named RAM kit or processor.

Reading the latency chart

Processors keep recently used data in small, fast caches close to the cores. While the test block fits in a cache, each jump takes a few nanoseconds. When the block grows past a cache level, the time per jump steps up, and once it is far bigger than every cache, most jumps go to main memory and the line flattens out at a much higher value.

The lines on the chart mark the bigger jumps. They are a rough hint of where your cache levels end, not exact cache sizes: the browser, other apps and how the operating system maps memory all blur the steps. Processors with a very large last-level cache keep the line low for longer before it climbs.

Why browser numbers are lower than native tools

  • One thread. A single core usually cannot use all the bandwidth of dual-channel memory. Native benchmarks such as AIDA64 load many cores at once.
  • JavaScript, not native code. The browser checks every array access and decides how the code is compiled, so the loops are slower than hand-tuned native code.
  • Address translation counts too. Random jumps over hundreds of megabytes also pay for the processor looking up where each page of memory lives, so the latency here reads higher than the figures native tools report.
  • Browser limits. Browsers cap how much memory one tab can take and make their timers less precise for security reasons. The test runs each job long enough that timer precision does not matter, but very large sizes may be refused.

This page also does not look for memory errors. For that, run Windows Memory Diagnostic (typemdsched.exe in the Start menu; it restarts the PC to test) or MemTest86, which boots from a USB stick and tests the RAM outside Windows.

How to check your RAM speed, XMP and EXPO in Windows

  1. Press Ctrl + Shift + Esc to open Task Manager, open Performance and clickMemory. The Speed line shows the speed your RAM runs at now, and Slots used shows how many sticks are installed.
  2. Compare that speed with the rating on the kit, such as DDR5-6000. If Windows shows a lower number, such as 4800 on a DDR5 kit, the memory is running at its default (JEDEC) speed.
  3. To run at the rated speed, enter the BIOS (UEFI) during start-up and enable the memory profile: XMP (Extreme Memory Profile) on Intel boards, EXPO on AMD AM5 boards. Some boards can also load an XMP profile on AMD. The menu name differs by board maker.
  4. If the PC will not boot or becomes unstable after enabling the profile, clear the CMOS (see the motherboard manual) and leave it off, or update the BIOS and try again.

Tools such as CPU-Z show a memory frequency that is half the MT/s rating, for example about 3000 MHz for DDR5-6000. That is normal: DDR memory moves data twice per clock. With two sticks, use the slots the motherboard manual recommends so they run in dual channel.

For games, enough RAM matters first: if Task Manager shows memory nearly full while you play, more RAM helps with stutter. My guide on how to increase FPS covers XMP and EXPO with the other free wins, and best CPU for gaming lists the RAM speeds AMD prints for its processors. Minecraft players can also check how to allocate RAM to Minecraft.

Common questions

Why are the GB/s numbers so much lower than AIDA64 or my RAM rating?

The page runs JavaScript on one thread inside the browser. One thread cannot use all the bandwidth of dual-channel memory, the browser adds safety checks to every array access, and the browser decides how the work runs. Native tools run optimised code on many threads at once. Use this page to compare your own runs in the same browser, not to compare with a native benchmark.

Does this test check my RAM for errors?

No. It measures speed, not errors. To check for faulty memory, use Windows Memory Diagnostic (type mdsched.exe in the Start menu) or MemTest86, which boots from a USB stick and tests the RAM outside Windows.

Why is the RAM amount the page shows rounded or lower than what I have?

Websites only get a rough RAM value through navigator.deviceMemory. Chromium browsers such as Chrome and Edge round it to a power of two and cap it. Since Chrome 153 the desktop value goes up to 32 GB and Android stops at 8 GB; older versions never showed more than 8 GB. Firefox and Safari do not share it at all. Task Manager or Settings > System > About shows the real amount.

What test size should I pick?

256 MB is a good default on PCs and laptops. 512 MB reaches further past the CPU caches, and 64 MB is gentler on phones and devices with little memory. If the browser refuses the size you picked, the test halves it until it fits and tells you.

Will turning on XMP or EXPO raise my score?

Often a little, because the RAM then runs at its rated speed instead of a slower default. A browser test only shows part of the change, so check the Speed line in Task Manager to confirm the profile is active.

Why do my results change between runs?

Background apps, the power plan or battery mode, other open tabs and the browser version all change the result. Run the test two or three times and look at the typical result, not the best one.

Is it safe to run?

Yes. The test reserves the memory size you pick for about 10 seconds and releases it when it finishes or when you press Stop. Nothing is written to your disk and nothing is uploaded.

  • Sensitivity converter — Convert sensitivity between Overwatch, Valorant, CS2, and other titles using documented yaw values.
  • cm/360 calculator — Turn mouse DPI and in-game sensitivity into centimeters per 360° turn.
  • eDPI calculator — Multiply mouse DPI by in-game sensitivity for a same-game comparison number.
  • FOV converter — Convert horizontal and vertical FOV across 4:3, 16:9, 16:10, and 21:9. Includes Overwatch, CS2, and Valorant.
  • Frame-time and Hz calculator — Convert FPS to milliseconds, compare that wait to the monitor refresh, and read a 1% low as a frame time.
  • Mouse polling-rate tester — Measure how often the browser sees pointer events. Average and max Hz.
  • Mouse raw input test — See every mouse report as a dot, spot sensor skips, and check jitter, swipe speed, and acceleration.
  • Mouse test — Check mouse buttons, the scroll wheel, double-click chatter, and a pointermove polling-rate reading in the browser.
  • Keyboard test — Press keys on a US QWERTY board, then time your typing on real sentences or random letter strings.
  • Gamepad test — See every controller button, analog trigger and stick live, check stick drift with a suggested deadzone, and test vibration.
  • Processor test — Single-thread and all-thread CPU speed, scaling and a throttling check.
  • System info — OS, browser, GPU, CPU threads, RAM and screen your browser reports.
  • Graphics test — Detect your GPU and run a relative WebGL2 benchmark.
  • Bottleneck / FPS estimator — Check if the CPU or the GPU holds your FPS back, from usage you measured.
  • Custom gaming PC — What to check before you order parts, RAM included.

Comments

No comments yet. Be the first to share what you think.