System information
These values are browser-provided hints, not a hardware inventory. The memory hint is neither installed RAM nor free memory: a device with 32 GiB installed can report a smaller value. Vigilum cannot determine your actual capacity from this hint; check your operating system for installed RAM. Privacy protections can also limit processor counts and mask graphics names. Compare runs in the same browser and configuration.
RAM verification
Check a chosen amount of memory with repeated writes and read-back verification.
Set up your memory test
Choose a memory budget and duration, then start.
Large allocations can freeze or crash your device and lose unsaved work. Available memory is not detected. Start with a modest target.
Test options · Standard
Standard uses up to four workers. Parallel uses more workers where possible. Both verify the same patterns.
Ready
Live measurements
- Write bandwidth
- —
- Compared read bandwidth
- —
- STREAM composite
- Unavailable
- 64 MiB pointer latency
- Unavailable
- Elapsed duration
- 0.0 s
Browser workload rates include verification overhead. Pointer latency uses a fixed working set; cache levels are not detected.
Coverage & verification
Coverage unavailable
- Unique Bytes Checked
- Awaiting verification
- Confirmed allocation
- —
- Observed mismatches
- —
Repeated traffic & passes
- Bytes written · repeated traffic
- —
- Bytes read · repeated traffic
- —
- Complete passes · minimum across workers
- —
Unique coverage includes compared bytes with mismatches. Repeated pass traffic is not additional unique coverage. Full requested coverage is not a certification of physical memory health.
Allocation limits & safety
No application RAM ceiling. Targets round down to 128 MiB chunks. Choose your own budget; available memory is not detected. Large requests can freeze or crash the browser or system and lose unsaved work. A crash is not proof of verification. Short runs may end before allocation or coverage completes; use Until stopped for longer runs. Browser and OS limits still apply. Stop terminates workers; memory reclamation is browser-controlled.
Measurements & method
RAM method & calculations
Workers allocate disjoint buffers in 128 MiB chunks. Each 32-bit word receives an index-dependent pattern, XORed with alternating 0x55555555 / 0xAAAAAAAA masks on successive passes, then is read back and compared. Mismatches count observed unequal words, not diagnosed faulty physical cells.
Unique coverage counts compared bytes once; repeated write/read traffic counts every pass. Live rates use changes in byte counters divided by elapsed seconds over a recent window, converted using 1 GiB = 1,073,741,824 bytes. They include elapsed time spent on both writing and verification, not separately isolated read/write timings. The manager sums worker rates. JavaScript execution and browser scheduling influence these values; they are not raw memory-controller bandwidth.
Finite runs also attempt a STREAM-style suite over three Float64 arrays: copy and scale count two array transfers; add and triad count three. Each rate is logical bytes × iterations / elapsed seconds / 2³⁰. The composite is the arithmetic mean of the four rates. This is this app's JavaScript workload, not a certified native STREAM result.
Pointer diagnostics time dependent accesses through shuffled chains at fixed 32 KiB, 512 KiB, 16 MiB and 64 MiB working sets. Nanoseconds/access = elapsed milliseconds × 1,000,000 / access count. Those sizes do not detect cache boundaries or guarantee DRAM-only access. Continuous runs skip these extra diagnostics.
Stopped coverage comes from the last received worker snapshots and may be a lower bound. Browser buffers are not all installed memory; available capacity is not detected. No RAM performance score or hardware-health certification is produced.
Worker policy & optional installed capacity
Standard: up to 4 workers, independent of target size, for lower-contention sequential pattern verification. Planned pool: 4 workers. Both modes use the same verified patterns; finite runs also attempt bandwidth and latency diagnostics. Until stopped skips those diagnostics.
Windows, the browser and other applications also need memory. Large allocations may trigger OS disk paging, so throughput may include virtual-memory pressure rather than physical-RAM speed. Available free memory and physical residency are not detected.