A real memory-bandwidth measurement, in your browser: the button below streams through a buffer far larger than your CPU's cache and reports copy bandwidth in GB/s. The full test adds separate read and write passes on a bigger buffer.
| Configuration | Theoretical peak |
|---|---|
| DDR4-2666 dual channel | 42.7 GB/s |
| DDR4-3200 dual channel | 51.2 GB/s |
| DDR5-5600 dual channel | 89.6 GB/s |
| DDR5-6000 dual channel | 96.0 GB/s |
| DDR5-8000 dual channel | 128 GB/s |
Two things move your real number more than the sticker speed. Channel count: a single stick runs single-channel and halves bandwidth — the most common self-inflicted downgrade in prebuilt PCs; two matched sticks fix it for free. XMP/EXPO: RAM ships running at standard speed until you enable its profile in the BIOS; DDR5-6000 kits routinely run at 4800 in machines where nobody flipped the switch. A single-threaded browser test reads well below the theoretical peak by design — but it reads proportionally, so single-channel or disabled-XMP systems still show up clearly low. The RAM latency calculator covers the other half: why CL30 at 6000 beats CL36 at 5600.
The browser test is single-threaded JavaScript through one core; theoretical peak needs every channel saturated from multiple cores. Expect a fraction of peak — the value is comparative: single-channel systems and disabled XMP show up clearly low versus similar machines.
Task Manager → Performance → Memory shows 'Slots used'; CPU-Z's Memory tab shows channel count directly. Two matched sticks in the right slots (usually 2 and 4) enable dual channel.
More than most upgrades ago: modern CPUs with high frame rates are memory-sensitive, and 1% lows improve measurably from DDR5-5600 to 6000+ on current platforms. Channel count matters most of all — never game on a single stick.