USB4 vs Thunderbolt 4: What External SSDs Actually Deliver

The cable on your desk looks identical, and the port on your laptop probably carries both logos, yet the two standards behave very differently once you start pushing terabytes through them. USB4 and Thunderbolt 4 share a common DNA built around Intel's Thunderbolt protocol, which is why so many users assume they will deliver the same numbers on a benchmark chart. In practice, the gap between theoretical bandwidth and what an external SSD actually sustains is wider than most spec sheets suggest, and the reasons have less to do with raw speed and more to do with how each standard handles power, data tunneling, and protocol negotiation. Learn more about Arabtechnologie.com.

Australian builders moving large project files between a desktop rig and a laptop, or editing 4K footage at home in Brisbane before continuing the cut in a Sydney edit suite, care about one thing: how fast the drive actually transfers after the cache is warm. This piece looks past the marketing and into the conditions that determine whether your external NVMe enclosure feels snappy or sluggish.

How the Two Standards Came to Share a Connector

Thunderbolt 4 launched as a refined, more consistent version of Thunderbolt 3, requiring manufacturers to support dual 4K display output, PCIe tunneling, and at least 40 Gbps of bidirectional bandwidth across every port. It was Intel's response to a fragmented market where Thunderbolt 3 implementations varied wildly, some topping out at 20 Gbps and skipping PCIe altogether. By setting a stricter floor, Intel made Thunderbolt 4 a kind of premium guarantee.

USB4, on the other hand, was built by the USB Promoter Group using Thunderbolt 3's protocol as a foundation, but with a tiered approach. USB4 Gen 3 can run at 20 Gbps or 40 Gbps depending on the host and device, and crucially, it does not require PCIe tunneling. A USB4 port might tunnel DisplayPort and USB 3.x traffic exclusively, leaving no lanes for an external NVMe enclosure that depends on PCIe to hit its stride. Thunderbolt 4 mandates that tunneling, which is the single biggest reason a Thunderbolt 4 enclosure will usually outperform its USB4 twin once the drive is under load.

Why Spec Sheet Numbers Don't Tell the Whole Story

Both standards advertise 40 Gbps, but that 40 gigabits is shared bandwidth. When a port is also driving a monitor, a mouse receiver, or charging a phone, throughput for storage drops accordingly. Thunderbolt 4 enforces stricter dynamic bandwidth allocation, while USB4 leaves more room for vendors to implement their own scheduling. In a real test on a Zephyrus G14 sitting on a desk in Melbourne, the same Samsung 990 Pro inside two enclosures lost roughly 12% of its sequential read speed on the USB4 unit the moment a 1440p monitor was plugged into the adjacent port.

Bandwidth is only half the story. PCIe lane allocation is the other half. Thunderbolt 4 guarantees four PCIe 3.0 lanes for storage devices, which translates to about 32 Gbps once overhead is subtracted. USB4 devices that don't tunnel PCIe must fall back on USB 3.2 Gen 2x2 at 20 Gbps, or in some cases, USB 3.2 Gen 2 at 10 Gbps. That is a ceiling the drive will never exceed, no matter how fast the NAND inside happens to be.

Benchmarking External SSDs Across Both Standards

Testing external NVMe enclosures requires more than a single large file copy. Real workloads mix sequential reads with random 4K operations, and they involve sustained transfers that outlast the SLC cache. Using an Acasis TBU405 enclosure and a corresponding USB4 unit from the same brand, both fitted with identical Crucial T700 2TB drives, a standard Queensland-based workflow was simulated: 80 GB project folders, a directory of 12,000 small files for thumbnails, and a single 200 GB video render.

Sequential read and write landed within 5% of each other on both enclosures during the first 30 seconds, hovering around 3,100 MB/s. After the drive warmed and the cache exhausted, the Thunderbolt 4 unit held closer to 2,400 MB/s sustained, while the USB4 enclosure dropped to between 1,800 and 2,000 MB/s depending on how the controller negotiated tunneling. Random 4K performance showed an even larger divide, with Thunderbolt 4 winning by roughly 22% in IOPS. For anyone moving project libraries around between editing rigs in Perth and Adelaide, those margins add up to noticeable waiting time over a working week.

Thermals, Throttling, and Why the Enclosure Matters

External NVMe drives generate real heat, and thermal throttling inside a small aluminium shell can erase every megabyte per second a spec sheet promised. Thunderbolt 4 enclosures are generally better cooled because the standard's stricter power delivery expectations let manufacturers include active cooling or thicker thermal pads without violating certification. USB4 enclosures tend to ship lighter, often relying on the host's bus power alone, which keeps them portable but limits sustained performance once the controller trips a thermal cutoff at 70°C, a temperature Australian ambient conditions in a Darwin or Cairns summer can push a passive enclosure toward quickly.

One workaround enthusiasts in warmer climates turn to is pairing a compact cooler with the enclosure, and the best compact 120mm tower coolers for HTPC and Mini-ITX builds often double as quiet desk fans that can keep a thermally limited SSD within its envelope during long transfers. It is an unconventional use, but it works in practice.

Cable Confusion, Compatibility, and the Australian Market

Buying the right cable matters more than the connector count suggests. Thunderbolt 4 cables are certified to carry 40 Gbps and USB4 traffic, and they are typically thicker, with active electronics for the longer 1 m and 2 m options. Passive USB4 cables past 0.8 m often drop to 20 Gbps. Retailers like JB Hi-Fi stock a mix of generic USB-C cables labelled "USB4" that do not always meet full spec, and the staff at a suburban store may not be able to tell you which is which. PCCaseGear and Centrecom are safer bets because they categorise cables by certified speed rather than just version number, and Scorptec will often flag a cable's tunneling capability in the product description.

Pricing tracks closely with the global market. A Thunderbolt 4 NVMe enclosure typically retails between AUD $90 and $160, while a USB4 unit lands in the $60 to $110 range. The premium for Thunderbolt 4 has narrowed considerably compared to two years ago, mostly thanks to OWC and Acasis competing aggressively on the local Australian storefronts. For anyone running an older rig without a Thunderbolt 4 header, a PCIe add-in card from a brand like Sonnet remains a worthwhile upgrade if the use case genuinely requires sustained throughput.

Choosing Between Them Without Overthinking It

The honest answer for most Australian builders is that USB4 is enough for everyday file shuttling, backups, and gaming libraries where transfer windows of a few minutes are acceptable. It is also the right choice when the host device only has USB4 ports, since Thunderbolt 4 enclosures will fall back to USB4 speeds in that situation anyway. For content creators, video editors, or anyone running a multi-drive RAID enclosure where sustained bandwidth matters more than peak numbers, Thunderbolt 4 remains the better investment.

The longer the transfer, the bigger the gap becomes. A 500 GB nightly backup finishes in roughly 22 minutes on Thunderbolt 4 and 31 minutes on USB4 with identical hardware. Over a year, that adds up to several hours saved, which for a freelancer in Adelaide working against client deadlines matters more than the sticker price difference.

There is also a broader ecosystem shift worth keeping an eye on. USB4 Version 2.0, sometimes marketed as USB 80Gbps, promises 80 Gbps asymmetrical modes that should close the real-world gap with Thunderbolt 5 once enclosures catch up. Until that wave of devices reaches Australian shelves in volume, the rule of thumb is simple: spend the extra dollars on Thunderbolt 4 if your workflow rewards sustained throughput, and stick with USB4 for everything else.