Thunderbolt 5 Claims 120 Gbps. Here's What That Actually Shaves Off a 500GB Transfer

Thunderbolt 5 has been shipping in laptops and external drives for about a year now, and the external-SSD makers building around it keep leading with the same number: up to 120 Gbps. That's triple what Thunderbolt 3 offered and more than double Thunderbolt 4's ceiling. Headline bandwidth numbers like that rarely translate one-to-one into real transfer times, though, so here's what the jump from USB 3.2 to Thunderbolt 4 to Thunderbolt 5 actually does to a real 500GB file transfer — and where the extra bandwidth stops mattering.

An external storage drive connected to a laptop via a USB-C cable
Illustration: an external drive connected via USB-C — a generic file photo, not the specific Thunderbolt 5 hardware discussed below. Photo by Augkun-ane, licensed CC BY 4.0. Original file and license.

The headline number, straight from Intel

Thunderbolt 5 is Intel's standard, and its own announcement is specific about what the number means. Intel's newsroom post describes "80 gigabits per second (Gbps) of bi-directional bandwidth" as the base spec, with a feature called Bandwidth Boost pushing one direction "up to three times the throughput for video-intensive usage, up to 120 Gbps." Intel frames the base figure as two times Thunderbolt 4's total bi-directional bandwidth. Thunderbolt 4 and the USB4 spec it's built on top one out at 40 Gbps, and the USB 3.2 Gen 2 standard most budget external drives still ship with runs at 10 Gbps. That's the full ladder: 10, 40, and 80–120 Gbps.

What a real drive actually delivers

None of those numbers are what you actually get in practice, because a drive's own flash and controller are usually the bottleneck long before the cable is. Three real, currently shipping products show where each tier actually lands. Samsung's Portable SSD T7, a USB 3.2 Gen 2 drive, is rated by Samsung at up to 1,050 MB/s sequential read. OWC's Envoy Pro, a Thunderbolt 4 drive, has been independently measured at 2,847 MB/s. OWC's newer Envoy Ultra, built for Thunderbolt 5, is advertised by OWC as exceeding 6,000 MB/s "under suitable conditions" — the company's own language, and a reminder that host hardware and cable quality both affect whether you see that figure. Converting Gbps to MB/s loses roughly 20–25% to protocol overhead in practice, which is why Thunderbolt 5's 120 Gbps (a theoretical 15,000 MB/s) shows up as a real-world 6,000 MB/s rather than anything close to the raw number.

There's a second gap worth knowing about on top of that protocol overhead: the rated number on any of these three drives is usually a burst figure, not a sustained one. Consumer SSDs write incoming data first into a fast SLC cache before moving it to slower long-term storage, and once that cache fills, write speed on cheaper drives can drop to a third of the rated figure or less. A short transfer of a few gigabytes will hit the rated number; a full 500GB project transfer, the scenario below, is exactly the kind of job long enough to expose whether a drive's real sustained speed matches its marketing number. The three products cited here were chosen because independent reviewers and the manufacturers themselves report sustained, not just burst, figures close to what's listed.

Converted into an actual transfer

Here's the part that matters for anyone actually moving files: how long it takes to copy a 500GB project — a reasonable size for a day's worth of 4K video footage — across each tier, assuming a sustained transfer with no other bottleneck.

InterfaceReal-world speedTime to transfer 500GB
USB 3.2 Gen 2 (Samsung T7)1,050 MB/s~7 minutes 56 seconds
Thunderbolt 4 / USB4 (OWC Envoy Pro)2,847 MB/s~2 minutes 55 seconds
Thunderbolt 5 (OWC Envoy Ultra)6,000 MB/s~1 minute 23 seconds

The math is simple division — 500,000 MB divided by the sustained MB/s figure — and it assumes the drive can sustain that speed for the full transfer rather than dropping off once an SLC cache fills, which happens on cheaper drives well before 500GB. The jump from USB 3.2 Gen 2 to Thunderbolt 4 saves about five minutes on this job. The jump from Thunderbolt 4 to Thunderbolt 5 saves only about ninety seconds more. The biggest single upgrade, in both percentage and absolute terms, is the first one — off the slowest tier — not the newest one.

Where the extra bandwidth stops mattering

That shrinking return is the real story, and it comes with real costs working against it. A Thunderbolt 5 external SSD costs meaningfully more than a comparable Thunderbolt 4 drive — OWC's own lineup prices its Thunderbolt 5 Envoy Ultra from $449.99 for an entry capacity up to $1,699.99 for 8TB, well above Thunderbolt 4 drives of similar capacity. Thunderbolt 5 also requires a host port that actually supports it; plugging a Thunderbolt 5 drive into an older Thunderbolt 3 or USB4 laptop port caps you at that port's bandwidth, and OWC itself notes performance "varies depending on the host hardware" even when the drive is rated for the faster tier. Cable choice matters too: passive Thunderbolt 5 cables that carry the full 80–120 Gbps are limited to shorter lengths, and longer runs need active, more expensive cables.

There's also a ceiling that has nothing to do with the cable at all. If the files being moved are coming off a phone, an SD card, or a spinning hard drive rather than another NVMe SSD, that source device's own read speed becomes the real limit long before Thunderbolt 5's bandwidth does — a typical SD card tops out well under 300 MB/s, which makes a 120 Gbps destination port irrelevant for that specific job.

Then there's the host side. As of this writing, native Thunderbolt 5 support is still a short list: Apple's M4 Pro and M4 Max MacBook Pro chips, announced in October 2024, support it, and on the Windows side it currently shows up mainly in high-end gaming and workstation laptops — models like Lenovo's Legion 9i, Dell's Pro Max 16, Asus's ROG Strix SCAR 16, and similarly specced machines from Acer, Alienware, Gigabyte, Maingear, and MSI, according to laptop-tracking roundups. Plug a Thunderbolt 5 drive into anything outside that list — a Thunderbolt 4, Thunderbolt 3, or plain USB4 port — and OWC's own documentation confirms the drive will simply run at whatever that older port supports, not its rated speed. Checking your own laptop's actual port spec before buying a Thunderbolt 5 drive matters more than the drive's box does.

Who this actually changes things for

Scale the per-transfer savings up and the picture gets clearer. An editor moving one 500GB project a day saves roughly 6.5 minutes a day going from USB 3.2 Gen 2 to Thunderbolt 4, and another 1.5 minutes a day going from Thunderbolt 4 to Thunderbolt 5 — assuming a five-day work week, call it 33 minutes a week from the first jump and 7.5 minutes a week from the second. Across a year, the first upgrade saves something like 28 hours; the second saves about 6.5. That's the honest way to compare them: the first tier jump is the one that actually changes your week, and the second is a smaller, compounding bonus on top of it.

Our take: Thunderbolt 5 is a genuine upgrade for a specific, fairly narrow group — video editors and photographers moving large RAW or raw-video projects between an SSD and a workstation on a daily basis, where even ninety seconds saved per transfer adds up across dozens of transfers a week, and for anyone who already owns one of the small number of M4 Pro/Max or high-end gaming laptops that can actually use the full bandwidth. For everyday backups, photo libraries, or moving a handful of documents, the jump from a budget USB 3.2 Gen 2 drive to any Thunderbolt tier is the upgrade that actually shows up in your day; the further jump from Thunderbolt 4 to Thunderbolt 5 is a smaller, more specialized win that's easy to justify for professional video work and hard to justify for general backup duty, especially once the higher drive cost and host-compatibility requirements above are factored in.

Related reading

Tip: If most of your transfers run off a laptop parked on a desk, the bottleneck is usually heat and cable clutter rather than the interface — a laptop stand keeps airflow under the chassis during long copies, and a wireless charger frees up the ports you actually need for storage. (These are Amazon Associate links — we may earn a small commission on qualifying purchases.)

Sources

Speeds are manufacturer-rated or independently measured figures as cited above, as of this writing (October 27, 2026); actual throughput depends on host hardware, cable, and source/destination storage. This is general technical information, not a recommendation to buy a specific product.

Start Here: every guide by topic · About · Privacy Policy

Comments

Post a Comment

Popular posts from this blog

Is Wi-Fi 7 Worth It in 2026? Your 314 Mbps Internet Plan Has an Opinion

Image
You are standing in the networking aisle a couple of weeks before Black Friday, holding a box with more antennas than seems strictly necessary. The front says Wi-Fi 7. The side lists a speed with far more digits than your internet bill. Somewhere at home, a perfectly functional router is blinking away, unaware that it is about to be judged. So, is Wi-Fi 7 worth it in 2026? For most homes the honest answer is: not for internet speed. The median U.S. home connection downloads at 314.35 Mbps, and a typical Wi-Fi 7 laptop radio tops out at 5.8 Gbps. The box in your hand is roughly eighteen times faster than the pipe it will be plugged into. Where Wi-Fi 7 does earn its keep is narrower, and more interesting, than the number on the box. A TP-Link Archer BE9700 tri-band Wi-Fi 7 router (shown as an example of the category, not a recommendation). Photo: TaurusEmerald, Wikimedia Commons , CC BY-SA 4.0 Key takeaways The median U.S. fixed broadband download speed was 314.35 Mbps in September ...
Read more

RTX 5060 vs. RTX 5070 Gaming Laptop: Which One Is Actually the Better Deal This Black Friday?

Image
Nvidia's GeForce RTX 5070 laptop GPU is consistently faster than the RTX 5060 laptop GPU in lab testing — that part is settled. What's less settled is whether that speed gap is worth paying for right now, because Black Friday 2026 pricing has, in at least a few live listings this week, squeezed the gap between a mid-spec RTX 5060 laptop and an RTX 5070 laptop down to a few hundred dollars, and in one case to barely $50 against an RTX 5070 Ti. Here's what the benchmark numbers actually show, why two laptops with the identical GPU name can perform differently, and which of this week's real listings comes out ahead on a dollars-per-frame basis. Illustration: a gaming laptop, a generic file photo and not a specific model discussed below. Photo by Kurt Kaiser, dedicated to the public domain (CC0), via Wikimedia Commons . Key takeaways In NanoReview's lab testing, the RTX 5070 laptop GPU averages about 13% higher frame rates than the RTX 5060 laptop GPU across re...
Read more

Your New Laptop's NPU Hits 80 TOPS, Double the Copilot+ Minimum. Here's What That Headroom Actually Changes

Image
Every new "AI PC" laptop this fall carries an NPU TOPS number on its spec sheet — 40, 50, 60, or 80 — and the number keeps climbing. Microsoft's original Copilot+ PC bar, set in 2024, was 40 TOPS of neural processing power. The chips shipping into stores right now clear that floor by a wide margin: Intel's Panther Lake (Core Ultra 300) NPU lands at roughly 50–60 TOPS, and Qualcomm's Snapdragon X2 Elite and X2 Elite Extreme, announced in September 2025 and now in retail Windows laptops, both carry an 80 TOPS Hexagon NPU — double the original minimum. The marketing treats bigger as simply better. The real question for a buyer is narrower: what does that extra TOPS headroom actually let a laptop do that a 40 TOPS machine couldn't, and what does it cost in return? Illustration: a generic backlit laptop keyboard, not a specific chip or device discussed in this article. Photo by Colin, via Wikimedia Commons , licensed CC BY-SA 4.0 . The number on the spec sheet...
Read more

Windows 10 Extended Security Updates: What $30 Actually Buys You Through October 2027

Image
Windows 10 support officially ended on October 14, 2025. If that sounds like old news, here is the part that did not get nearly as much attention: Microsoft quietly moved the goalposts in June 2026, extending the consumer Extended Security Updates (ESU) program's cutoff from October 13, 2026 to October 12, 2027. If you are one of the roughly one in six Windows PCs still running Windows 10 a year after "end of support," that change probably already shaped your decision about whether to pay, enroll for free, or finally replace the machine — whether you noticed it happening or not. The honest judgment call here: for most households with an otherwise-fine PC, the $30 (or free) ESU path is the right short-term answer, and buying new hardware purely to escape a security-update bill almost never pays for itself in year one. Illustration only, not a specific ESU enrollment screen. Photo: Donald Trung, Wikimedia Commons , licensed under CC BY-SA 4.0 . Wait — I thought Windows...
Read more

Microsoft's Recall Can Claim 150 GB of Your Drive. Here's What That Default Actually Costs You

Image
Windows 11's Recall feature — the one that screenshots your screen every so often so you can search your own activity later — doesn't just cost CPU cycles and an NPU fast enough to run it locally. It reserves a chunk of your drive, and Microsoft's own support documentation lays out exactly how big that chunk is by device tier. Run the numbers on that table and a few other figures Microsoft publishes alongside it, and the real cost of turning Recall on looks different from "it uses some storage." Illustration: a consumer SSD, representative of the fixed drive capacity Recall's storage allocation is carved out of. Photo by Jacek Halicki , via Wikimedia Commons , licensed under CC BY-SA 4.0 . The Four Numbers Microsoft Requires Before Recall Turns On Before any snapshot gets saved, a machine has to clear a hardware floor. Per Microsoft's own Recall management documentation, a device needs a Copilot+ PC meeting the Secured-core standard, an NPU rated a...
Read more

DDR5 RAM Prices Are Roughly 4x Higher Than a Year Ago. Here's What That Actually Does to a Laptop Price

Image
DDR5 memory prices have roughly quadrupled over the past year, and the common assumption is that a laptop's price should have gone up by something close to that much too. It hasn't — Dell's own commercial price increases have run 10% to 30%, not 300%, because of how little of a finished computer's cost is actually memory. A 16GB DDR5-4800 module (Crucial CT16G48C40U5). Photo: Rainer Knäpper, Free Art License, via Wikimedia Commons . Illustration of a DDR5 module generally, not a specific product endorsement. Key takeaways TrendForce's own tracker puts the spot price of a 16GB DDR5 UDIMM at $240 as of October 19, 2026, versus a pre-shortage baseline of roughly $50-70 for the same module in 2024 — a 3.5x to 4x increase. Buying RAM directly to upgrade a desktop or laptop yourself means eating close to that full increase. Buying a finished laptop or desktop means eating a much smaller slice of it — Dell's own commercial-line price increases were 10% to 30%,...
Read more

OpenAI's New Pricing Split: One Model Cuts Your Bill 80%, the Other Cuts Your Wait 8x for 6x the Price

Image
Correction October 12, 2026: the cached-input row described the discount as 90% off input for both models. It is 90% for Astra ($1.00 per 1M against $10.00) and 95% for Sol ($0.10 against $2.00); the prices listed in the table were already correct. Source: OpenAI API pricing page. Updated October 10, 2026. Every price and benchmark figure below was re-checked against the sources listed at the end on that date. The October update section near the end records what changed after this piece first ran on September 30. At DevDay 2026 on September 29, OpenAI announced two upgrades to its flagship API lineup that pull in opposite directions. GPT-6.1 Sol prices its tokens at roughly one-fifth of GPT-6 Astra's rate while scoring almost identically on independent benchmarks. A new "Ultrafast" tier for Astra generates tokens up to 8 times faster in Codex, but charges 6 times more per token to do it. Neither number means much on its own -- an 8x speed claim says nothing about wheth...
Read more