It's common for devices rated to USB 3.2 Gen 2 not to hit the advertised 10 Gbit/s speed.
It's common for devices rated to USB 3.2 Gen 2 not to hit the advertised 10 Gbit/s speed.
I own a Sandisk Extreme Portable and an Extreme Pro Portable SSD. The first model supports up to 1050 MB/s while the second reaches 2000 MB/s. My ASUS ROG Strix X570-E Gaming board and Corsair 680X RGB case both feature USB 3.2 Gen 2 ports, expected to deliver around 1200 MB/s (10 Gbit/s). Testing showed the Extreme Portable achieved about 1022MB/s for reads and 940MB/s for writes on both the motherboard and case USB-C ports. The Extreme Pro Portable recorded roughly 1034MB/s read and 950MB/s write, indicating it should perform well but only slightly exceeds the port speeds.
These bandwidth figures are just theoretical limits—what you'd get if you imagined perfect conditions like spherical cows in a vacuum. In reality, no interface reaches the advertised top speed. SATA SSDs operate at around 6 Gbps, translating to roughly 750 MB/s, but actual performance is closer to 550 MB/s. PCIe x4 Gen 3 NVMe drives claim up to 3.938 GB/s, yet most deliver only about 3.5 GB/s. Portable SSDs using a 5 Gbps USB connection might reach 625 MB/s, but typically hit around 500 MB/s. So while a 10 Gbps connection suggests 1.25 GB/s, the actual speed is often much lower—around 8 Gbps or 1000 MB/s.
Remember that 6 Gbit/s equals a theoretical speed of about 600 MB/s, not 750 MB/s. In SATA transfers, each bit corresponds to one byte.
USB 3.2 Gen 1 also suffers from the same problem, employing 8b/10b encoding which lowers the actual speed to 500 MB/s. USB 3.2 Gen 2 utilizes 128b/132b encoding, offering a theoretical maximum of about 1200 MB/s. This difference reflects what sets them apart—Gen 2 uses a more efficient scheme, yet both are marketed as "10 Gbps." The discrepancy could explain why the expected speed isn't reaching that figure; the motherboard or case port might actually be Gen 1x2 instead of Gen 2. Even though they're labeled similarly, the real performance often falls short. This inconsistency between label and actual output is what matters here.