Convert between bits, bytes, KB, MB, GB, TB, and PB — in both decimal (SI) and binary (IEC) standards.
1 Gigabyte = 1,000 Megabyte
A data storage converter translates digital storage quantities between units — from the single bit (the smallest unit of information) to petabytes and beyond (the scale of hyperscale cloud infrastructure). Whether you need to convert megabytes to gigabytes for a cloud storage plan, understand why your new hard drive shows less capacity than advertised, or translate a 100 Mbps internet speed into actual file-transfer MB/s, this converter handles it instantly. Crucially, this tool offers two modes: Decimal (SI) mode uses powers of 1,000 (the convention used by storage manufacturers and macOS), and Binary (IEC) mode uses powers of 1,024 (the convention used by Windows and most operating systems when displaying storage). Selecting the wrong mode can introduce errors of up to 9.95% at the terabyte level — significant when purchasing storage or provisioning servers.
The roots of binary computing trace back to the earliest electronic computers. Because transistors and logic gates naturally represent two states — on or off, high voltage or low — base-2 arithmetic was the inevitable foundation of digital hardware. John von Neumann's 1945 architectural blueprint, which underpins virtually every modern processor, organized memory into addressable byte-sized cells to simplify addressing logic. Since memory chips were manufactured in power-of-two capacities (4 KB, 8 KB, 16 KB, and so on), engineers informally stretched the SI prefix "kilo" to mean 1,024 rather than 1,000 — the nearest power of two to 1,000. A 64 KB RAM module therefore held exactly 65,536 bytes (2¹⁶), not 64,000. This informal convention persisted for decades, and the kilobyte meaning 1,024 bytes (2¹⁰) became deeply embedded in software and operating system documentation long before any standards body intervened. The IEC did not introduce the unambiguous kibibyte (KiB = 2¹⁰ bytes) until 1998 — by which point the confusion had calcified across an entire industry.
The practical fallout of that naming collision is visible every time a user moves files between Windows and macOS. Copy a folder that Windows reports as "14.9 GB" to a Mac and Finder displays the same folder as "16 GB." No data was gained or lost — Windows File Explorer reports storage in binary gibibytes (GiB) but labels them "GB," while macOS Finder switched to decimal gigabytes in 2009 with Snow Leopard. A 16 GB (decimal) folder is 16,000,000,000 bytes; divided by 1,073,741,824 bytes per GiB, Windows displays that as 14.9 GB. The same bytes, two different numbers, zero difference in actual content. For consumers buying storage, the asymmetry means a "1 TB" hard drive looks full-sized in macOS but appears to be missing roughly 69 GB in Windows — a perennial source of support calls and consumer frustration that persists today precisely because mainstream operating systems have never converged on a single display convention.
The decimal vs. binary distinction also has an increasingly direct impact on the economics of data infrastructure. Cloud providers — AWS, Azure, Google Cloud — price object storage in decimal gigabytes (1 GB = 10⁹ bytes), consistent with the SI convention used in their capacity planning and billing documentation. But Linux servers, which run the overwhelming majority of cloud workloads, report disk usage in binary units (GiB, MiB) in tools like df andlsblk. A developer who provisions a 500 GB (decimal) volume on AWS EBS and then checks df -h inside their EC2 instance will see approximately "466G" available — not because AWS shorted them, but because 500,000,000,000 bytes ÷ 1,073,741,824 bytes/GiB = 465.7 GiB. Multiply this confusion across hundreds of storage volumes in a production environment and it can lead to capacity planning spreadsheets that are systematically off by ~7%. This converter exists to make that translation explicit, auditable, and instant — enter any value, select decimal or binary, and get the precise equivalent in every unit from bits to exabytes.
The converter uses the byte as its base unit. In decimal (SI) mode, each prefix step multiplies by 1,000. In binary (IEC) mode, each prefix step multiplies by 1,024. Use the decimal/binary toggle at the top of the converter to switch between modes — the conversion table below shows both systems side by side. The formula is: Result = Input × (fromUnit_in_bytes ÷ toUnit_in_bytes). For bit-based units, 1 byte = 8 bits. The complete reference tables used internally:
DECIMAL (SI) — Powers of 1,000 BINARY (IEC) — Powers of 1,024 ───────────────────────────────────── ────────────────────────────────────── 1 bit = 0.125 bytes 1 bit = 0.125 bytes 1 byte = 1 byte 1 byte = 1 byte 1 KB = 1,000 bytes 1 KiB = 1,024 bytes 1 MB = 1,000,000 bytes 1 MiB = 1,048,576 bytes 1 GB = 1,000,000,000 bytes 1 GiB = 1,073,741,824 bytes 1 TB = 1,000,000,000,000 bytes 1 TiB = 1,099,511,627,776 bytes 1 PB = 10^15 bytes 1 PiB = 2^50 bytes ≈ 1.126 × 10^15 1 EB = 10^18 bytes 1 EiB = 2^60 bytes ≈ 1.153 × 10^18
Kilobyte / Kibibyte
Megabyte / Mebibyte
Gigabyte / Gibibyte
Terabyte / Tebibyte
Petabyte / Pebibyte
Exabyte / Exbibyte
FORMULA
Result = Input × (fromUnit_bytes ÷ toUnit_bytes) Decimal mode: KB=10³, MB=10⁶, GB=10⁹, TB=10¹², PB=10¹⁵ Binary mode: KiB=2¹⁰, MiB=2²⁰, GiB=2³⁰, TiB=2⁴⁰, PiB=2⁵⁰ Bits to bytes: divide by 8 (1 byte = 8 bits)
Worked Example: A "1 TB" hard drive in Windows — how many GiB does it show?
The decimal vs. binary storage confusion is not merely academic — it has triggered class action lawsuits against hard drive manufacturers. In 2005, a California court settlement required Western Digital to pay $250,000 to charity and provide backup software to customers who claimed they were misled by "gigabyte" labeling. The FTC ultimately ruled that manufacturers could use decimal SI prefixes, but the consumer expectation mismatch persists. In cloud computing, the distinction directly affects costs: AWS charges by the GB (decimal), but you are provisioning storage visible to your OS in GiB (binary). A cloud engineer who budgets 1 TB (1,000 GB decimal) but needs 1 TiB (1,099.5 GB decimal) of actual usable space will unexpectedly exceed their storage allocation by ~9.95%.
The bits-vs-bytes gap creates a persistent and costly mismatch between advertised internet speeds and the download rates users actually observe. ISPs market plans in megabits per second (Mbps) — a convention inherited from telecommunications, where transmission capacity has always been measured in bits. Download managers, web browsers, and file transfer utilities, however, display progress in megabytes per second (MB/s) because files are measured in bytes. The conversion is straightforward — divide Mbps by 8 to get MB/s — but it is non-obvious to most users. A household with a "500 Mbps" plan will never see "500 MB/s" in their download bar; the true ceiling is 62.5 MB/s under ideal conditions. Ethernet overhead, Wi-Fi contention, server-side throttling, and TCP congestion control reduce that further. Understanding this single conversion prevents the most common internet-speed complaint: "I'm paying for 500 Mbps but my downloads only hit 60 MB/s."
Cloud storage pricing introduces another layer of ambiguity that directly affects consumers and businesses. When a provider sells a "1 TB plan," that 1 TB is invariably measured in decimal SI terabytes — 1,000,000,000,000 bytes — because that convention maximizes the advertised number. Your operating system, however, may report available space in binary units. A Windows user who purchases a 1 TB cloud plan will see it displayed as approximately 931 GB in any Windows-native storage view. This creates the persistent and widespread illusion that the cloud provider has shortchanged them by nearly 70 GB. In reality, every purchased byte is present; it is the labeling mismatch that conjures the apparent deficit. For enterprise purchasers, this distinction compounds significantly: a 100 TB object storage allocation (decimal) yields only about 90.9 TiB of binary-reported space, and a storage architect who confuses the two when sizing a system can end up running out of space before the provisioned quota is officially exhausted.
Beyond storage, the MB-vs-Mb distinction matters acutely in data engineering pipelines and ETL workflows. Stream-processing frameworks like Apache Kafka measure throughput in MB/s (bytes per second) when reporting consumer lag and partition offsets, but network monitoring dashboards typically display interface utilization in Mbps (bits per second). A Kafka cluster producing "500 MB/s" of data needs 500 × 8 = 4,000 Mbps = 4 Gbps of network uplink just to move that data off the broker — a figure that catches many platform engineers off guard the first time they instrument their pipeline at scale. Database replication compounds the issue further: backup tools commonly report restore progress in MB/s while the replication link between primary and replica databases is provisioned in Mbps, requiring the divide-by-8 conversion at every handoff point to ensure the link will not become a bottleneck under peak write load. Capacity planners who skip this conversion routinely under-provision network links by a factor of 8 — a mistake that surfaces only during high-throughput events such as batch jobs, overnight ETL runs, or major traffic spikes, when saturated links cause cascading latency and data pipeline delays. This converter makes all those translations a single lookup rather than a mental arithmetic exercise performed under pressure during an incident.
A user purchases a '2 TB' external hard drive from a major manufacturer. The box lists '2,000,000,000,000 bytes (2 TB)' in fine print. When they plug it into Windows, File Explorer shows '1.81 TB' (or '1,862 GB'). Using the data converter in binary (IEC) mode: 2,000,000,000,000 bytes ÷ 1,073,741,824 bytes/GiB = 1,862.6 GiB. Windows rounds this to 1.81 TiB but calls it 'TB.' On macOS, the same drive shows '2 TB' because Apple uses decimal SI since 2009. Neither reading is wrong — they use different measurement conventions. The practical impact: when planning backup storage, always budget based on binary GiB if your OS is Windows, using the formula: Windows display GB = (Manufacturer GB × 1,000,000,000) ÷ 1,073,741,824.
An internet service provider advertises a '100 Mbps' broadband plan. A customer wants to know how long it will take to download a 4K movie (15 GB) and upload 50 GB of vacation photos to cloud backup. Download speed: 100 Mbps ÷ 8 = 12.5 MB/s. Download 15 GB movie: 15,000 MB ÷ 12.5 MB/s = 1,200 seconds = 20 minutes. Upload speed is typically asymmetric: a common 100/20 Mbps plan uploads at 20 Mbps = 2.5 MB/s. Upload 50 GB of photos: 50,000 MB ÷ 2.5 MB/s = 20,000 seconds = 5.56 hours. The data converter confirms: 100 Mbps in the Mbps-to-MB/s conversion = 12.5 MB/s. Knowing this conversion prevents frustration when downloads are 8× slower than the advertised megabit number implies.
A photographer has 50 GB of processed photos in JPEG format (approximately 12,500 photos at ~4 MB each) and wants to back them up to cloud storage. In decimal SI: 50 GB = 50,000 MB = 50,000,000,000 bytes. Monthly cloud storage cost: Google One at $2.99/month for 100 GB is sufficient; AWS S3 Standard at $0.023/GB/month = $1.15/month. Upload time on a 20 Mbps uplink: 50 GB × 8 bits/byte = 400 Gb; 400,000 Mb ÷ 20 Mbps = 20,000 seconds = 5.56 hours. At $0.023/GB on S3: 50 × $0.023 = $1.15/month. After 3 years: $1.15 × 36 = $41.40 total. If the photo library grows to 500 GB (5 years at 100 GB/year), the monthly cost scales to $11.50 — still cheaper than most consumer plans. The converter makes these calculations explicit and plannable.
Confusing Mb/s (megabits) with MB/s (megabytes) — an 8× difference
The most common and costly data conversion mistake. Internet speeds are advertised in megabits per second (Mbps), but file transfer progress bars and storage benchmarks use megabytes per second (MB/s). Since 1 byte = 8 bits, 100 Mbps = 12.5 MB/s. Expecting 100 MB/s downloads on a '100 Mbps' plan — 8× the actual speed — leads to incorrect time estimates for downloads, backups, and streaming capacity planning. Always divide Mbps by 8 to get MB/s. Similarly, 1 Gbps = 125 MB/s, not 1,000 MB/s.
Assuming KB always means 1,024 bytes — it doesn't
Before the IEC standardized binary prefixes in 1998, 'KB' was ambiguous — sometimes meaning 1,000 bytes (decimal, used in storage) and sometimes 1,024 bytes (binary, used in computing). Today, the technically correct distinction is: KB (kilobyte) = 1,000 bytes (decimal SI, used by manufacturers and macOS); KiB (kibibyte) = 1,024 bytes (binary IEC, used by Windows and many software tools). Using KB=1,024 everywhere introduces a 2.4% error per prefix step, compounding to 7.4% error at GB and 9.95% at TB. Always check whether a spec sheet or OS display uses decimal or binary.
Hard drive capacity discrepancy: manufacturer decimal vs OS binary
A '1 TB' hard drive contains exactly 1,000,000,000,000 bytes (manufacturer, decimal). Windows reports this as approximately 931 GiB, displayed as '931 GB' (binary). The ~69 GB 'missing' capacity is not a defect — it is purely a labeling convention difference. macOS shows the same drive as '1 TB' because it uses decimal. When ordering storage for a server with a specific usable space requirement (e.g., 900 GiB minimum), always verify the raw byte count and convert it using the correct binary divisor (1,073,741,824 bytes/GiB) rather than trusting the marketing label.
Confusing RAM size with storage size
Both RAM and storage are measured in GB, but they serve fundamentally different purposes and have different cost-per-GB profiles. RAM is fast volatile working memory (~$4–8/GB); storage is slow persistent capacity (~$0.03–0.10/GB SSD). A computer with '16 GB RAM and 512 GB SSD' is very different from one with '512 GB RAM and 16 GB SSD' (the latter is a specialized database server configuration). In benchmarks, RAM bandwidth is measured in GB/s (DDR5 ~50–100 GB/s), while SSD throughput is MB/s (typically 500–7,000 MB/s). Mixing these metrics when evaluating system performance leads to entirely wrong conclusions.
Confusing RAM speed (MHz) with storage size (GB) — different metrics entirely
RAM specifications include both capacity (GB) and speed (MHz or MT/s). DDR5-6400 RAM has a transfer rate of 6,400 MT/s (megatransfers per second) ≈ 51,200 MB/s (with 64-bit bus). The '6400' is a clock rate, not a storage amount. Confusing RAM clock speed with storage capacity is a category error that appears frequently in PC build discussions. Storage size (256 GB, 512 GB, 1 TB) and RAM size (8 GB, 16 GB, 32 GB) are comparable in GB, but RAM frequency (3,200–8,400 MHz range for DDR5) has nothing to do with how many gigabytes of data can be stored.
Backup size estimation errors from ignoring compression
A 100 GB folder of mixed content (documents, photos, videos) does not necessarily require 100 GB of backup storage, nor does it compress to the same ratio as a 100 GB folder of text files. Photos (JPEG) and videos (H.264/H.265) are already compressed and achieve only 0–5% further reduction with ZIP. Documents and code achieve 50–80% compression. If your backup tool reports '60 GB backed up from 100 GB source,' the compression ratio was 1.67:1 — but this will not repeat for different content types. Always budget 100% of original size for backup storage unless you have measured your actual compression ratio on your specific data types.
RAID storage calculation errors: RAID 1 halves usable space, RAID 5 loses one drive
RAID (Redundant Array of Independent Disks) configurations affect usable capacity significantly. RAID 0 (striping): usable = N × drive capacity (no redundancy). RAID 1 (mirroring): usable = 1 × drive capacity (50% overhead regardless of drive count — two 4 TB drives give 4 TB usable). RAID 5 (striped with parity, min. 3 drives): usable = (N-1) × drive capacity — three 4 TB drives give 8 TB usable (25% overhead). RAID 6 (dual parity, min. 4 drives): usable = (N-2) × drive capacity — four 4 TB drives give 8 TB usable (50% overhead). Buying '8 TB' of drives for a RAID 1 NAS and expecting 8 TB of usable storage is a common and expensive mistake — the actual usable space is only 4 TB.
The ambiguity of "kilo," "mega," and "giga" in computing predates the internet era. In the 1960s–80s, engineers informally used "kilo" to mean 1,024 (the nearest power of 2 to 1,000) because computer memory sizes were naturally powers of 2. A "64 kilobyte" RAM chip had exactly 65,536 bytes (2¹⁶). Meanwhile, telecommunications and storage engineers applied the standard SI meaning (1,000). By the 1990s, a "gigabyte" meant different things in different technical contexts, fueling consumer confusion and legal disputes. In 1998, the International Electrotechnical Commission (IEC) published IEC 80000-13, introducing unambiguous binary prefixes: kibi- (Ki, 2¹⁰ = 1,024), mebi- (Mi, 2²⁰), gibi- (Gi, 2³⁰), tebi- (Ti, 2⁴⁰), pebi- (Pi, 2⁵⁰), exbi- (Ei, 2⁶⁰). NIST (National Institute of Standards and Technology) adopted these prefixes in NIST SP 811. Despite the standard, adoption remains inconsistent: Linux kernel uses KiB/MiB/GiB; macOS uses decimal GB; Windows uses binary values but decimal labels. The IEEE and ISO endorse the IEC prefixes, and technical publications increasingly use them, but marketing materials almost universally continue with SI prefixes — a situation unlikely to resolve until OS labeling converges.
Modern data storage spans five primary technology families, each with distinct capacity-per-dollar, speed, and endurance characteristics. NAND Flash SSDs(2026 pricing: ~$0.05–0.10/GB): NVMe PCIe 4.0 delivers 5,000–7,000 MB/s sequential read; PCIe 5.0 reaches 12,000–14,000 MB/s. 3D NAND stacking (up to 232+ layers in 2025) continues to increase density. Hard Disk Drives (HDDs) (~$0.015–0.025/GB): Seagate and Western Digital now ship 22–26 TB helium-filled HDDs using SMR (Shingled Magnetic Recording) and HAMR (Heat-Assisted Magnetic Recording) technology, with 30+ TB drives expected by 2026. Optical Storage: M-DISC archival Blu-ray (100–1,000 GB) offers 1,000+ year data retention; standard BD-R is 25–100 GB. Magnetic Tape(LTO-9: 18 TB native, 45 TB compressed): tape remains the lowest-cost long-term archival medium at ~$0.002–0.005/GB, used extensively by Netflix, Facebook, and national archives.Emerging Technologies: DNA storage (theoretically 215 PB/gram), holographic storage (Cerabyte, targeting 1 PB/rack at PB scale), and glass storage (Microsoft Project Silica: 75.6 GB per glass square, 10,000-year lifespan).
Storage and network performance is described by three distinct metrics that are often conflated. Bandwidth (throughput) is the maximum data transfer rate, measured in MB/s or Gbps — relevant for large sequential reads (video editing, backups, large file downloads). Latency is the delay before a transfer begins, measured in microseconds (μs) or milliseconds (ms) — NVMe: ~20–100 μs; SATA SSD: ~100–500 μs; HDD: ~5–15 ms; cloud storage S3: ~100–200 ms per request. Latency dominates performance for small random I/O operations (database queries, web server responses). IOPS (Input/Output Operations Per Second) measures how many small random read/write operations can be performed per second — NVMe SSDs: 500,000–2,000,000 IOPS (4K random read); SATA SSDs: 90,000–100,000 IOPS; HDDs: 75–200 IOPS. A system with 7,000 MB/s bandwidth but 200 IOPS (an HDD) will perform terribly as a database server despite impressive sequential speed. Conversely, a system with "only" 500 MB/s bandwidth but 1,000,000 IOPS (NVMe SSD) excels for transaction workloads. When evaluating storage for a specific use case, all three metrics must be considered in the context of workload characteristics.
Networking professionals deal with data measurement at every layer of the protocol stack, and the units shift depending on the context. At the IP layer, the standard Ethernet Maximum Transmission Unit (MTU) is 1,500 bytes — the largest payload a single IPv4 packet can carry without fragmentation on most networks. Larger frames are possible with Jumbo Frames (up to 9,000 bytes, common in data center 10 GbE environments), but routers on the public internet enforce the 1,500-byte MTU, meaning any larger packet is silently split. The bandwidth-delay product (BDP) is a critical metric for high-throughput long-distance links: BDP = bandwidth × round-trip latency. A 10 Gbps link with 100 ms RTT has a BDP of 10 × 10⁹ × 0.1 = 10⁸ bits = 125 MB — meaning up to 125 MB of data can be "in flight" simultaneously before the sender receives confirmation. TCP window sizes must be tuned to match the BDP or throughput will be throttled far below the link's physical capacity.
At the optical fiber layer, modern DWDM (Dense Wavelength Division Multiplexing) systems carry hundreds of wavelengths simultaneously, with each wavelength supporting up to 400 Gbps using coherent optics — yielding total fiber capacities in the tens of terabits per second per strand. Terrestrial and submarine fiber cables now routinely operate at 100+ Tbps aggregate capacity; the 2Africa submarine cable (around Africa and connecting to Europe and Asia) carries a design capacity of 180 Tbps. At the wireless edge, 5G NR (New Radio) defines a theoretical peak downlink throughput of 20 Gbps per cell under ideal millimeter-wave (mmWave) conditions, though real-world deployments typically deliver 1–4 Gbps in dense urban mmWave deployments and 100–400 Mbps in sub-6 GHz macro-cell coverage. Wi-Fi 7 (802.11be), the current generation of indoor wireless, achieves throughput up to 46 Gbps using 320 MHz channels and 4096-QAM modulation across the 2.4, 5, and 6 GHz bands simultaneously. Protocol headers consume a measurable fraction of raw capacity: a minimal IPv4 header is 20 bytes; TCP adds another 20 bytes minimum; TLS 1.3 adds a further 5-byte record header plus authentication overhead. On a 1,500-byte MTU frame, protocol overhead alone consumes roughly 3% of available capacity — a rounding error at modest scale but a meaningful throughput tax at 400 Gbps line rate. Understanding these numbers in both bits and bytes — and knowing where each unit system applies — is essential for network architects, telecom engineers, and anyone sizing bandwidth for data-intensive applications.
The rise of large-scale machine learning has introduced storage and data transfer requirements that dwarf conventional computing workloads, making fluency with large data units practically necessary for anyone working in the field. Large language model (LLM) weights are typically stored in 16-bit or 32-bit floating-point format. GPT-4 is widely estimated to contain roughly 1.8 trillion parameters; at 2 bytes per parameter (FP16), that equates to approximately 3.6 TB of raw weight data — far exceeding the VRAM of any single GPU. Even more compact open-source models illustrate the VRAM constraint sharply: Llama 3 70B requires approximately 140 GB in FP16, necessitating two or more 80 GB H100 GPUs for inference, or quantization to 4-bit (INT4) which reduces the footprint to roughly 35 GB at the cost of some output fidelity. NVIDIA's H100 SXM5 offers 80 GB of HBM3 high-bandwidth memory with 3.35 TB/s memory bandwidth; the H200 doubles capacity to 141 GB with 4.8 TB/s HBM3e bandwidth. Even so, frontier models require NVLink-connected multi-GPU nodes (8× H100 = 640 GB combined VRAM) or multi-node tensor parallelism to hold weights entirely in GPU memory during inference. The GPU memory bandwidth figure — measured in TB/s — is the primary throughput bottleneck for autoregressive token generation, which is memory-bound rather than compute-bound; a model generating one token at a time must stream the entire weight matrix through VRAM for each token, making high-bandwidth memory more valuable than peak FLOP counts for inference latency.
Training datasets compound the storage challenge considerably. Common Crawl, the web-scale text corpus used to pre-train most modern LLMs, accumulates roughly 400 TB of compressed raw HTML per annual crawl — over 3 petabytes of uncompressed text. After deduplication, language filtering, and quality scoring, a usable training set might shrink to 5–20 TB of cleaned text tokens, but the intermediate storage and processing pipeline must handle the full petabyte-scale input. Image and video corpora are even larger: LAION-5B (5.85 billion image-text pairs used to train Stable Diffusion and CLIP) required approximately 240 TB of storage at roughly 40 KB average per image. Video foundation models such as Sora operate on corpora of hundreds of petabytes of video. Checkpointing alone adds significant overhead: saving optimizer states, gradients, and activations for a trillion-parameter model mid-training requires tens to hundreds of terabytes per checkpoint, and production training runs save checkpoints every few hundred steps as insurance against hardware failure. The implication for infrastructure engineers is that storage cost, I/O throughput between storage and GPU nodes, and inter-node network bandwidth (400 Gbps InfiniBand or RoCE is now standard in AI clusters) are all first-class design constraints — and being able to convert accurately between GB, TB, and PB is a baseline competency for anyone sizing or operating ML infrastructure.
What is the difference between decimal (SI) and binary (IEC) storage units?
Decimal (SI) prefixes use powers of 1,000: 1 kilobyte (KB) = 1,000 bytes, 1 megabyte (MB) = 1,000,000 bytes, 1 gigabyte (GB) = 1,000,000,000 bytes. Binary (IEC) prefixes use powers of 1,024: 1 kibibyte (KiB) = 1,024 bytes, 1 mebibyte (MiB) = 1,048,576 bytes, 1 gibibyte (GiB) = 1,073,741,824 bytes. Hard drive manufacturers use decimal prefixes (so a '1 TB' drive contains exactly 1,000,000,000,000 bytes), while operating systems like Windows report storage in binary (GiB) but label it 'GB.' This mismatch causes the well-known confusion where a 1 TB hard drive appears as only ~931 GB in Windows.
Why does my new hard drive show less storage than advertised?
When you buy a '2 TB' hard drive, the manufacturer uses decimal gigabytes: 2 TB = 2,000,000,000,000 bytes. Windows reports storage in binary gibibytes but labels them 'GB': 2,000,000,000,000 ÷ 1,073,741,824 = 1,862.6 GiB — displayed as '1.86 TB' or '1,862 GB.' This is not a fraud; the manufacturer is technically correct in using SI prefixes. macOS (since version 10.6 Snow Leopard in 2009) switched to reporting storage in decimal GB, so a 2 TB drive shows as '2 TB' on a Mac. Linux typically reports in binary GiB. The IEC 80000-13 standard (1998) introduced KiB/MiB/GiB to resolve this ambiguity, but adoption has been slow.
What is the difference between a bit and a byte?
A bit is the smallest unit of digital information — it can hold a value of either 0 or 1. A byte is a group of 8 bits and can represent 256 different values (2⁸ = 256), sufficient to encode a single ASCII character (letter, digit, or symbol). The 8-bit byte became the industry standard in the 1960s–70s, though early computers used 6-bit and 7-bit 'bytes.' Modern computing universally uses 8-bit bytes. Storage capacity is measured in bytes (GB, TB), while data transfer speeds are typically measured in bits per second (Mbps, Gbps). This difference is critical: a 100 Mbps internet connection transfers 100 megabits per second = 12.5 megabytes per second (÷ 8).
What is the difference between Mbps and MBps?
Mbps (megabits per second) and MBps (megabytes per second) differ by a factor of 8. Network speeds (internet plans, Wi-Fi, Ethernet) are almost always advertised in megabits per second (Mbps or Mb/s). File sizes and download progress bars are shown in megabytes (MB or MBps). So a '100 Mbps' internet plan downloads at 100 ÷ 8 = 12.5 MB/s maximum throughput. A '1 Gbps' (gigabit) fiber connection delivers up to 125 MB/s. This distinction matters enormously: if your plan says 50 Mbps and you expect to download a 5 GB file in about 2 minutes (5,000 ÷ 50 = 100 seconds seems fast), the correct calculation is 5,000 MB ÷ 6.25 MB/s = 800 seconds ≈ 13.3 minutes.
How many GB are in a TB?
In decimal (SI) notation: 1 TB = 1,000 GB = 1,000,000 MB = 1,000,000,000 KB = 1,000,000,000,000 bytes. In binary (IEC) notation: 1 TiB = 1,024 GiB = 1,048,576 MiB = 1,073,741,824 KiB = 1,099,511,627,776 bytes. The difference between 1 TB and 1 TiB is about 9.95% — roughly 99.5 GB. For consumer storage devices sold as '1 TB,' the actual byte count is 1,000,000,000,000 bytes. In Windows, this appears as approximately 931 GiB (displayed as 'GB'). In macOS (post-2009), it correctly shows as approximately 1,000 GB. When purchasing storage, always check the byte count rather than relying on the marketing label.
What is the difference between a petabyte and a pebibyte?
A petabyte (PB) = 10¹⁵ bytes = 1,000,000,000,000,000 bytes (decimal SI). A pebibyte (PiB) = 2⁵⁰ bytes = 1,125,899,906,842,624 bytes (binary IEC). The difference is approximately 12.6%: 1 PiB is about 1.126 PB. To put petabytes in perspective: the entire text of the US Library of Congress is estimated at about 15 TB (0.015 PB). Facebook generates approximately 4 petabytes of data per day. The entire human genome (30 billion base pairs, compressed) is about 1.5 GB — so 1 PB could store roughly 667,000 complete human genomes. Data center operators typically measure storage in petabytes, while internet backbone providers track traffic in exabytes per month.
What is an exabyte?
An exabyte (EB) = 10¹⁸ bytes = 1,000,000 terabytes = 1,000 petabytes (decimal SI). An exbibyte (EiB) = 2⁶⁰ bytes ≈ 1.153 × 10¹⁸ bytes. As of the mid-2020s, global internet traffic exceeds 400 exabytes per month. A single exabyte could store approximately 36,000 years of high-definition video, or the entire content of the internet circa 2004–2005. Hyperscale data centers operated by companies like Google, Amazon, Microsoft, and Meta each store tens to hundreds of exabytes of data. The NSA's Utah Data Center, built in 2013, was estimated to store between 3 and 12 exabytes. Total global data created and replicated in 2024 was estimated at approximately 147 zettabytes (147,000 exabytes).
What is a zettabyte and yottabyte?
A zettabyte (ZB) = 10²¹ bytes = 1,000 exabytes = 1,000,000 petabytes. A yottabyte (YB) = 10²⁴ bytes = 1,000 zettabytes. IDC's 'Data Age' report estimates that the global datasphere (all data created, captured, copied, and consumed) will reach 291 ZB by 2027. The entire observable universe, if filled with hard drives at current storage densities, would hold approximately 10⁷⁸ to 10⁸⁰ bits — roughly 10⁶⁹ to 10⁷¹ YB — so yottabytes represent the practical ceiling of any foreseeable human data need. In 2022, the IEC added ronnabyte (RB, 10²⁷) and quettabyte (QB, 10³⁰) to the SI prefix system to future-proof against data growth that may eventually reach yottabyte scale per year.
What is the storage capacity of a CD, DVD, and Blu-ray disc?
CD (Compact Disc): standard single-layer CD holds 700 MB (approximately 80 minutes of audio). DVD (Digital Versatile Disc): single-layer DVD holds 4.7 GB; dual-layer DVD holds 8.5 GB. DVD-R DL (dual layer) stores approximately 8.54 GB. Blu-ray: single-layer BD holds 25 GB; dual-layer BD holds 50 GB; triple-layer BD holds 100 GB; quad-layer UHD BD holds 100 GB (used for 4K UHD movies). These figures use decimal SI gigabytes as specified by their standards bodies. Optical media has largely been supplanted by streaming and solid-state storage for consumer use, but remains relevant for archival storage (M-DISC can last 1,000+ years) and physical media distribution in regions with slow internet.
How much storage is on a typical smartphone?
Smartphone storage in 2025–2026 typically ranges from 128 GB to 1 TB for flagship devices. The iPhone 16 lineup starts at 128 GB and goes up to 512 GB. Android flagships like the Samsung Galaxy S25 Ultra offer up to 1 TB. In decimal SI terms: 128 GB = 128,000,000,000 bytes; in Windows-style binary display, this would appear as approximately 119 GiB. A 128 GB phone can realistically store approximately 32,000 photos (at 4 MB per HEIF photo), 8,000 minutes of 720p video (at 300 MB per hour), or about 25,600 songs (at 5 MB per MP3). Apps, the OS, and system files typically consume 10–15 GB, leaving 113–118 GB for user content.
How many photos fit in 1 GB?
The number of photos that fit in 1 GB (1,000 MB) depends heavily on file format and resolution. JPEG photos from a modern smartphone (12–50 MP) typically compress to 3–8 MB each, fitting about 125–333 photos per GB. RAW photos from professional DSLRs (24–61 MP) range from 20–80 MB each, yielding only 12–50 per GB. Apple's HEIF format (used on iPhones since iOS 11) achieves similar quality to JPEG at roughly half the file size (~2–4 MB), fitting 250–500 photos per GB. A standard 12 MP JPEG averages about 4 MB: 1,000 MB ÷ 4 MB = 250 photos per GB. Cloud storage services like Google Photos and iCloud offer free tiers of 15 GB and 5 GB respectively.
How many songs fit in 1 GB?
An average MP3 song at 128 kbps bitrate is approximately 3–4 MB for a 4-minute track. At 320 kbps (near-CD quality), the same song is about 9.5 MB. At 128 kbps: 1,000 MB ÷ 3.75 MB = ~267 songs per GB. At 320 kbps: 1,000 MB ÷ 9.5 MB = ~105 songs per GB. Apple AAC at 256 kbps (iTunes default): about 7.5 MB per song = ~133 songs per GB. Lossless FLAC audio (CD quality, ~1,411 kbps): about 25–40 MB per song = 25–40 songs per GB. Streaming services like Spotify and Apple Music offer 24-bit lossless at 3,730 kbps, which would be ~100 MB per song if downloaded, fitting only about 10 songs per GB. Spotify Premium's cache for offline playback uses approximately 10–150 MB per album.
How many hours of video fit in 1 TB?
Video file size depends on resolution, frame rate, codec, and bitrate. Approximate sizes per hour: SD (480p, H.264): ~500 MB/hr; HD (1080p, H.264): ~1.5–4 GB/hr; HD (1080p, H.265/HEVC): ~0.75–2 GB/hr; 4K (2160p, H.264): ~10–20 GB/hr; 4K (2160p, H.265): ~5–10 GB/hr; 4K RAW (cinema): 100–300 GB/hr. On a 1 TB drive using H.265 compression: 1080p: ~500–1,333 hours; 4K: ~100–200 hours; 4K RAW: ~3–10 hours. Consumer camcorders typically record at 50 Mbps (≈22.5 GB/hr), fitting about 44 hours per TB. Broadcast-quality ProRes 4K records at 400–800 Mbps (~180–360 GB/hr), fitting only 3–5 hours per TB.
What is Netflix's 4K streaming bitrate?
Netflix streams 4K Ultra HD content at approximately 15–25 Mbps (megabits per second) using H.265/HEVC compression. Standard HD (1080p) streams at 5–25 Mbps depending on the plan and content. SD (480p) streams at 1–3 Mbps. To convert to MB/s: 25 Mbps ÷ 8 = 3.125 MB/s. A 2-hour 4K movie at 25 Mbps uses: 25 Mbps × 8 = 200 Mb/min × 120 min = 24,000 Mb = 3,000 MB = 3 GB of data. Netflix recommends a minimum 15 Mbps connection for 4K streaming. By contrast, Disney+ streams 4K at up to 25 Mbps, Apple TV+ at up to 29 Mbps, and Amazon Prime Video at up to 25 Mbps. Physical 4K Blu-ray discs contain H.265 video at up to 100 Mbps — 4–7× higher quality than streaming.
How fast is a 1 Gbps fiber internet connection in MB/s?
A 1 Gbps (gigabit per second) fiber connection theoretically delivers 1,000 Mbps ÷ 8 = 125 MB/s of download speed. In practice, overhead from TCP/IP protocol headers, ISP throttling, server speed limits, and Wi-Fi bottlenecks reduce real-world speeds to 80–110 MB/s over a wired Ethernet connection, and 50–90 MB/s over Wi-Fi 6 (802.11ax). At 125 MB/s theoretical max: a 50 GB game downloads in 400 seconds (6.7 minutes); a 4K movie (15 GB) downloads in 120 seconds (2 minutes); a 1 TB backup transfers in 8,000 seconds (2.22 hours). 2.5 Gbps and 10 Gbps (multi-gigabit) fiber is now available in some markets, delivering 312 MB/s and 1,250 MB/s respectively — fast enough to saturate even NVMe SSD write speeds (~3,000–7,000 MB/s for PCIe 4.0/5.0).
What is the difference between RAM and storage, and why are both measured in GB?
RAM (Random Access Memory) and storage (HDD/SSD) are both measured in gigabytes because both store digital data in bytes, but they serve entirely different functions. RAM is the computer's working memory — fast (DDR5 RAM transfers at ~50–100 GB/s), volatile (data is lost when power is removed), and expensive per GB (~$3–8/GB for DDR5). Storage (SSD/HDD) is for persistent data — slower (NVMe SSDs at 5–12 GB/s, HDDs at 100–200 MB/s), non-volatile, and much cheaper per GB (~$0.03–0.10/GB for SSDs, ~$0.01–0.02/GB for HDDs). A typical 2026 computer might have 16–32 GB of RAM and 512 GB–2 TB of SSD storage. Confusing the two is common: '16 GB' is ambiguous — 16 GB RAM means fast multitasking; 16 GB storage means very limited file space.
What is the difference between SSD and HDD storage?
SSDs (Solid State Drives) use NAND flash memory chips with no moving parts: sequential read speeds of 500 MB/s (SATA) to 12,000 MB/s (PCIe 5.0 NVMe), random access latency of 20–100 microseconds, and typical lifespans of 1–5 million hours MTBF. HDDs (Hard Disk Drives) use spinning magnetic platters with a read/write head: sequential speeds of 100–250 MB/s, random access latency of 5–15 milliseconds (100–750× slower than SSDs), but much higher capacity per dollar ($0.01–0.02/GB vs SSD $0.05–0.10/GB). HDDs remain cost-effective for bulk archival storage (1–20 TB at low cost). SSDs dominate as the system boot drive and application storage. Enterprise data centers use both: SSDs for hot data (frequently accessed) and HDDs for cold data (archival), a tiered storage architecture.
How is cloud storage priced?
Cloud storage pricing varies significantly by provider and service tier. Consumer cloud storage (2026 prices): Google One: free 15 GB, $2.99/mo for 100 GB, $9.99/mo for 2 TB. Apple iCloud+: free 5 GB, $0.99/mo for 50 GB, $2.99/mo for 200 GB, $9.99/mo for 2 TB. Amazon Photos (Prime members): unlimited photo storage free with Prime, 5 GB free for videos. Enterprise/developer cloud object storage: AWS S3: ~$0.023/GB/month for standard storage, dropping to $0.004/GB/month for Glacier deep archive. Azure Blob Storage and Google Cloud Storage have similar tiered pricing. Enterprise SAN/NAS storage on-premises: $0.05–0.20/GB/month including hardware amortization, power, cooling, and administration — often more expensive than cloud for small deployments.
What is data compression and what are typical compression ratios?
Data compression reduces file size by encoding information more efficiently. Lossless compression (no data lost): ZIP (DEFLATE) achieves 2:1 to 5:1 compression on text, spreadsheets, and code; PNG reduces image files 20–50% vs. BMP; FLAC reduces audio 30–50% vs. PCM WAV. Lossy compression (some data discarded): JPEG compresses photos 10:1 to 40:1 vs. raw sensor data; MP3 at 128 kbps is roughly a 10:1 compression of CD audio; H.265 video achieves 40–50% smaller files than H.264 at equal quality. Video compression ratios vs. raw: H.264 compresses 4K RAW by approximately 100:1; H.265 by 200:1; AV1 (open-source, used by Netflix/YouTube) achieves similar quality to H.265 at 20–30% smaller file sizes. Compression ratios vary enormously with content type — highly repetitive data compresses better than random or already-compressed data.
What is a bit?
A bit (binary digit) is the most fundamental unit of information in computing and digital communications. Coined by John Tukey in 1947 and popularized by Claude Shannon in his 1948 paper 'A Mathematical Theory of Communication,' a bit can represent exactly one of two values: 0 or 1 (false or true, off or on, negative or positive voltage). Shannon's information theory defines the bit as the minimum amount of information needed to distinguish between two equally likely alternatives. Modern processors handle 64 bits simultaneously (64-bit architecture). A byte is 8 bits; a nibble is 4 bits. In network communications, the bit per second (bps) is the base transmission rate unit, scaled to kilobits (Kbps), megabits (Mbps), gigabits (Gbps), and beyond.
How many bytes does ASCII vs Unicode use per character?
ASCII (American Standard Code for Information Interchange, 1963) encodes 128 characters (A–Z, a–z, 0–9, punctuation, control codes) in exactly 7 bits, stored in 1 byte with 1 bit unused. Extended ASCII uses all 8 bits for 256 characters including accented letters. Unicode (Universal Character Standard) is designed to encode every character in every human writing system — over 149,000 characters across 161 scripts. UTF-8 (most common Unicode encoding): 1 byte for ASCII characters, 2 bytes for Latin, Greek, Hebrew, Arabic (128–2,047 range), 3 bytes for most Asian scripts and emoji (2,048–65,535), 4 bytes for supplementary characters. A plain English 1,000-word text in ASCII is about 6,000 bytes; the same text in UTF-8 is also ~6,000 bytes (since all ASCII characters map 1:1). A 1,000-character Chinese text in UTF-8 is ~3,000 bytes (3 bytes per CJK character).
What is a typical blockchain transaction size in bytes?
Bitcoin transaction sizes: a simple transaction with 1 input and 2 outputs (one recipient + change back to sender) is approximately 225–250 bytes. A transaction with multiple inputs (e.g., consolidating UTXOs) can be 500–1,000 bytes. The Bitcoin block size limit is 1 MB for legacy transactions (with SegWit enabling up to ~4 MB effective block weight). Each block is mined approximately every 10 minutes, so Bitcoin's theoretical throughput is about 3,500–7,000 transactions per block ÷ 10 minutes = ~7 transactions per second (TPS). Ethereum: transactions are 100–500 bytes; blocks process ~15 TPS. Visa processes ~1,700 TPS with theoretical peak capacity of ~24,000 TPS. The data storage implications: Bitcoin's entire blockchain as of 2026 exceeds 600 GB; Ethereum's full archive node exceeds 12 TB.
What are common email attachment size limits?
Email attachment size limits vary by provider: Gmail: 25 MB per message (sends larger files via Google Drive link). Outlook/Microsoft 365: 20–25 MB for consumer accounts, up to 150 MB for Exchange Online (configurable by admins). Yahoo Mail: 25 MB per message. iCloud Mail: 20 MB (larger files sent via Mail Drop, stored on iCloud for 30 days, up to 5 GB). When sending a 10 MB file, the actual email size is typically 13–14 MB because MIME base64 encoding inflates binary attachments by ~33%. Corporate email servers may have tighter limits (5–10 MB) enforced by IT policy. For large file transfers, alternatives like WeTransfer (2 GB free), Dropbox, Google Drive, OneDrive, or SFTP are standard in professional contexts.
What is the theoretical data storage density of DNA?
DNA is the most information-dense storage medium known. Each nucleotide base (A, T, G, C) in a DNA strand encodes 2 bits of information. DNA can be packed at densities of approximately 10¹⁸ bits per cubic centimeter — equivalent to about 215 petabytes per gram of DNA. In 2019, researchers at the University of Washington demonstrated storing and retrieving 1 gigabyte of data in DNA. In 2021, Microsoft and Twist Bioscience achieved random-access retrieval from a DNA pool containing 1 TB of data. Key challenges: writing (synthesis) speed is slow (~200–400 bases/second, improving with enzymatic methods), reading (sequencing) is faster with nanopore technology, and the cost per GB remains approximately $1,000–10,000 compared to $0.02/GB for flash storage. DNA storage could theoretically store all of humanity's data in a space smaller than a shoebox.
What is the estimated storage capacity of the human brain?
The human brain's storage capacity is estimated at approximately 2.5 petabytes (2,500 terabytes) by neuroscientists at the Salk Institute (2016 study), based on the number of synapses (~100 trillion) and their estimated precision (synaptic strength can take approximately 26 distinct values, encoding ~4.7 bits each). Total: 10¹⁴ synapses × 4.7 bits ≈ 4.7 × 10¹⁴ bits ≈ 58.8 TB, scaling up to 2.5 PB with more generous estimates. However, the brain is not a digital storage device — it encodes information in patterns of neural activity, chemical gradients, and synaptic weights simultaneously, making direct byte comparisons imprecise. The brain also continuously rewrites and reinterprets memories, unlike static digital storage.
What is Moore's Law as it applies to storage?
Moore's Law (Gordon Moore, 1965) originally described the doubling of transistors per chip roughly every two years. Applied to storage: HDD areal density has historically doubled every 18–24 months, driving cost per GB from ~$1,000/GB in 1980 to ~$0.02/GB in 2026 — a 50,000-fold reduction. Flash (NAND) storage density has followed a similar trajectory: 3D NAND stacking (64, 96, 128, 176, 232+ layers) has maintained density growth even as 2D shrinking approached physical limits. However, Moore's Law for compute is slowing — transistor density is approaching atomic scales (~1 nm process nodes in 2025). Storage density growth continues faster than compute, meaning data generation consistently outpaces our ability to process it — a fundamental challenge in big data, AI training, and scientific computing.
What are NAND flash cell types (SLC, MLC, TLC, QLC)?
NAND flash cells are classified by how many bits they store per cell: SLC (Single-Level Cell): 1 bit/cell — fastest, most durable (100,000 P/E cycles), most expensive, used in enterprise SSDs and industrial applications. MLC (Multi-Level Cell): 2 bits/cell — good balance of speed, endurance (10,000 P/E cycles), and cost; used in high-end consumer SSDs. TLC (Triple-Level Cell): 3 bits/cell — most common consumer SSD type (2,000–3,000 P/E cycles); good cost/GB, acceptable performance with DRAM cache. QLC (Quad-Level Cell): 4 bits/cell — lowest cost/GB, lowest endurance (300–1,000 P/E cycles), slower write speeds; used in value SSDs and high-capacity drives. The trade-off: more bits per cell = higher capacity = lower cost per GB = lower endurance and write speed.
What is write endurance and TBW (terabytes written)?
Write endurance measures how much data can be written to an SSD over its lifetime before cell degradation causes failure. TBW (terabytes written) is the manufacturer's endurance rating: a Samsung 990 Pro 2TB has a TBW of 1,200 TB (i.e., 1.2 PB of writes before warranted replacement). A typical consumer user writes 20–50 GB/day, meaning 1,200 TBW would last 24,000–60,000 days (65–164 years) — far beyond the drive's useful life. Enterprise SSDs have TBW ratings of 1–60 PB due to higher-endurance SLC NAND. DWPD (Drive Writes Per Day) normalizes TBW by warranty period: 1 DWPD means you can write the drive's full capacity once per day for the warranty period. A 2 TB SSD rated at 1 DWPD over 5 years = 5 × 365 × 2 TB = 3,650 TB = 3.65 PB TBW.
How much storage do hyperscale data centers operate?
Hyperscale data centers (Google, Amazon AWS, Microsoft Azure, Meta) collectively operate on the order of exabytes to zettabytes. Public estimates (2024–2025): Google stores approximately 15–20 exabytes of user data in Gmail alone; Google Photos stores 4+ trillion photos. Amazon S3 (object storage) holds trillions of objects totaling hundreds of exabytes. Microsoft's Azure and OneDrive collectively manage hundreds of exabytes. Meta (Facebook, Instagram, WhatsApp) generates 4+ petabytes of new data daily, adding ~1,500 PB per year. Apple's iCloud manages over 700 petabytes for its 1 billion+ device users. The entire Facebook photo archive (650 billion photos as of 2023) requires approximately 1.5 exabytes of storage, served from custom 'cold storage' systems using 30 TB hard drives.
What is the 3-2-1 backup rule?
The 3-2-1 backup rule is the gold standard for data backup strategy: keep at least 3 copies of data, on at least 2 different storage media types, with at least 1 copy offsite (or in the cloud). Example: original data on your computer (copy 1, medium 1: SSD), backup on an external hard drive at home (copy 2, medium 2: HDD), and cloud backup via Backblaze/iCloud/Glacier (copy 3, medium 3: cloud servers offsite). An extension, 3-2-1-1-0, adds: 1 immutable or air-gapped copy (ransomware protection) and 0 errors verified by testing restores. For large datasets, storage sizing: if original data is 2 TB, the 3-2-1 rule requires at minimum 6 TB of total storage spread across locations. Cloud backup pricing: Backblaze B2 charges $0.006/GB/month, so 2 TB = $12/month for the cloud component.
What are NVMe and SATA speeds compared in MB/s?
SATA III (the most common HDD/SSD interface through ~2020): maximum interface bandwidth of 600 MB/s; real-world SSD sequential reads: 500–560 MB/s; random 4K reads: 90,000–100,000 IOPS (~360 MB/s). NVMe PCIe 3.0 (×4 lanes): 3,500 MB/s sequential read, 3,000 MB/s sequential write. NVMe PCIe 4.0 (×4 lanes, mainstream 2021+): 7,000 MB/s sequential read, 6,500 MB/s write. NVMe PCIe 5.0 (×4 lanes, high-end 2023+): up to 14,000 MB/s sequential read, 12,000 MB/s write. HDD (7,200 RPM SATA): 120–250 MB/s sequential, 0.5–2 MB/s random (4K). The practical difference: copying a 100 GB file takes ~3.3 minutes at 500 MB/s (SATA SSD), ~23 seconds at 7,000 MB/s (PCIe 4.0 NVMe), and ~7–14 minutes at 120–250 MB/s (HDD). For OS booting, NVMe SSDs boot Windows in ~5–10 seconds vs. 30–60 seconds for SATA SSDs vs. 60–120 seconds for HDDs.
Binary prefixes (IEC mode): This converter's binary mode follows IEC 80000-13 (2008), defining kibibyte (KiB) = 2¹⁰ bytes, mebibyte (MiB) = 2²⁰ bytes, gibibyte (GiB) = 2³⁰ bytes, tebibyte (TiB) = 2⁴⁰ bytes, pebibyte (PiB) = 2⁵⁰ bytes, exbibyte (EiB) = 2⁶⁰ bytes. These definitions are also endorsed by NIST SP 811 and IEEE Std 1541.
Decimal prefixes (SI mode): This converter's decimal mode follows ISO/IEC 80000-13 SI prefixes: kilobyte (KB) = 10³ bytes, megabyte (MB) = 10⁶ bytes, gigabyte (GB) = 10⁹ bytes, terabyte (TB) = 10¹² bytes, petabyte (PB) = 10¹⁵ bytes, exabyte (EB) = 10¹⁸ bytes.
Bit-byte relationship: 1 byte = 8 bits (IETF standard, RFC 2119 and related; universally adopted since the 8-bit byte became standard in the 1970s). This converter uses the 8:1 bit-to-byte ratio for all conversions involving bits.
Disclaimer: This converter is provided for educational and informational purposes. Storage capacities reported by operating systems, manufacturers, and cloud providers may use different conventions and should be verified with the vendor. Data transfer speeds reflect theoretical maximums; real-world speeds depend on protocol overhead, hardware limitations, and network conditions. Financial Growth Hub accepts no liability for purchasing, provisioning, or infrastructure decisions based on these calculations.
Share your thoughts
Help us improve The Free Calculator Site