Terabyte per second to Megabyte per second
TBps
MBps
Conversion History
| Conversion | Reuse | Delete |
|---|---|---|
| No conversion history to show. | ||
Quick Reference Table (Terabyte per second to Megabyte per second)
| Terabyte per second (TBps) | Megabyte per second (MBps) |
|---|---|
| 0.001 | 1,000 |
| 0.01 | 10,000 |
| 0.1 | 100,000 |
| 1 | 1,000,000 |
| 3.35 | 3,350,000 |
| 10 | 10,000,000 |
About Terabyte per second (TBps)
A terabyte per second (TB/s or TBps) equals 8 terabits per second and represents the bandwidth scale of GPU memory systems, high-performance computing interconnects, and the fastest data center storage fabrics. The HBM3 memory stacks on high-end AI accelerators provide 3–4 TB/s of internal bandwidth. InfiniBand NDR connections used in supercomputers reach 400 Gbps per link, with multiple links aggregated to TB/s totals. At 1 TB/s, the entire contents of a 1 PB data store could transfer in about 17 minutes.
The NVIDIA H100 GPU features 3.35 TB/s of HBM3 memory bandwidth. Top-tier supercomputers like Frontier aggregate over 75 TB/s of storage I/O bandwidth.
About Megabyte per second (MBps)
A megabyte per second (MB/s or MBps) equals 8,000,000 bits per second and is the practical unit that most users encounter when watching a download progress bar. A 100 Mbps broadband connection downloads at up to 12.5 MB/s; a USB 3.0 drive transfers at 50–100 MB/s; an NVMe SSD reads at 3,000–7,000 MB/s. Understanding MB/s alongside Mbps resolves the common frustration of seeing a "1 Gbps" plan deliver "only" 125 MB/s — the two figures are consistent, not contradictory.
A 100 Mbps home broadband plan delivers up to 12.5 MB/s in a download manager. A USB 3.2 flash drive typically writes at 50–200 MB/s.
Terabyte per second – Frequently Asked Questions
Why do AI chips need TB/s of memory bandwidth?
Large language models have billions of parameters that must be read from memory for every inference pass. An LLM with 70 billion parameters at 16-bit precision needs 140 GB of data read per forward pass. At 3 TB/s, the H100 can perform roughly 20 inference passes per second — bandwidth directly determines tokens-per-second output.
Why is memory bandwidth the main bottleneck for large language model inference?
During LLM inference each token requires reading all model weights from memory. A 70-billion-parameter model at 16-bit precision means 140 GB read per forward pass. At 30 tokens per second, that is 4.2 TB/s of memory reads — right at the limit of an H100's HBM3. This is why AI inference is "memory-bound": the GPU's compute cores sit idle waiting for data. Quantising weights to 8-bit or 4-bit halves or quarters the bandwidth demand, directly increasing tokens per second.
What is the fastest memory bandwidth ever achieved in a commercial chip?
The NVIDIA B200 GPU with HBM3e achieves approximately 8 TB/s of memory bandwidth as of 2025. Each generation roughly doubles bandwidth — from 2 TB/s (A100) to 3.35 TB/s (H100) to 4.8 TB/s (H200) to 8 TB/s (B200). The trajectory suggests 16+ TB/s within a few years.
How long would it take to transfer a petabyte at 1 TB/s?
About 16.7 minutes. A petabyte is 1,000 terabytes, so at 1 TB/s, the math is simple division. For context, the Library of Congress contains roughly 10–20 petabytes of data. Transferring it all at 1 TB/s would take about 3–6 hours.
Is there anything beyond TB/s?
Yes — petabytes per second (PB/s). Experimental optical interconnects and photonic computing architectures are pushing toward PB/s-class bandwidth. Some supercomputer storage systems already aggregate into the PB/s range when all nodes operate simultaneously. It is the next frontier for AI training clusters.
Megabyte per second – Frequently Asked Questions
Why does copying files to a USB drive slow down partway through?
Many USB drives use a small SLC cache for initial writes at high MB/s, then slow dramatically once the cache fills and data writes to slower TLC/QLC NAND. A drive that starts at 200 MB/s might drop to 20–30 MB/s after the first few gigabytes. Check sustained write speed reviews, not just peak numbers.
What MB/s do I need for video editing?
Editing 4K ProRes footage requires about 200–400 MB/s of sustained read speed. 8K RAW can demand 1,000+ MB/s. A SATA SSD (550 MB/s) handles 4K fine, but 8K workflows really need NVMe drives at 3,000+ MB/s. The timeline scrubbing experience directly correlates with MB/s.
How do I tell if a speed test result is in MB/s or Mbps?
Look at the capitalisation: lowercase "b" (Mbps) means megabits, uppercase "B" (MB/s) means megabytes. Most speed test websites (Speedtest by Ookla, fast.com) default to Mbps. If your result seems 8× lower than expected, you are probably reading MB/s where you expected Mbps.
What is the fastest consumer storage in MB/s as of 2026?
PCIe 5.0 NVMe SSDs hit 12,000–14,000 MB/s sequential read speeds. That is fast enough to load an entire 50 GB game in about 4 seconds. PCIe 6.0 drives, expected soon, will double this again to roughly 25,000 MB/s.
Why is network file sharing slower than local disk speed?
Network transfers add latency, protocol overhead (SMB, NFS), and are limited by the network link speed. A file on a local NVMe SSD reads at 7,000 MB/s, but sharing it over a 1 Gbps network caps throughput at 125 MB/s. Even 10 GbE only gives 1,250 MB/s — a fraction of modern SSD capability.