Intel Debuts Thunderbolt Share for Direct PC-to-PC Video Streaming

Linus Tech Tips

Intel's new Thunderbolt Share protocol justifies the interface's premium status by bypassing network compression entirely, while its Wi-Fi 7 hardware delivers real-world wireless speeds that finally surpass traditional gigabit Ethernet.

AI-assisted summary of This WAS under NDA... Until I Called in a Favor — original source. Check the original for full context.

Source published: 2024-01-12

Key ideas

Intel showcased Thunderbolt Share, a prototype software technology that enables uncompressed, zero-latency PC-to-PC display streaming and direct file transfers reaching sustained speeds of 800 MB/s. (0:00)
Thunderbolt Share avoids video compression artifacts by streaming directly between GPU frame buffers over a virtual network utilizing roughly 4 Gbps of bandwidth, at the cost of ~25% CPU overhead on an Intel Core Ultra 7 client. (3:00)
Wi-Fi 7 (802.11be) demonstrated real-world client throughput ranging from 2 Gbps to past 4 Gbps, easily outpacing standard 1 Gbps wired Ethernet connections found on consumer hardware. (6:00)
Key Wi-Fi 7 technical enhancements include Multi-Link Operation (MLO) for sub-1ms failover or band aggregation, 320 MHz channel widths in the 6 GHz spectrum, 4K-QAM modulation, and universal preamble puncturing. (8:32)

Detailed notes

## Background Published by Linus Tech Tips on YouTube on January 12, 2024, this video covers Intel's demonstrations from CES 2024. According to the page description, Linus showcases Thunderbolt Share as his most anticipated reveal of the show and explores how Wi-Fi 7 introduces practical performance upgrades for both everyday users and niche tech enthusiasts.

## TL;DR Linus examines Intel's prototype Thunderbolt Share technology, demonstrating how point-to-point connections between two Windows PCs enable 800 MB/s file transfers and zero-compression, low-latency display mirroring directly across GPU frame buffers. He also unpacks the real-world advantages of Wi-Fi 7 (802.11be), highlighting verified speeds exceeding 2 to 4 Gbps alongside technical leaps such as 320 MHz channels, 4K-QAM modulation, Multi-Link Operation (MLO), and preamble puncturing. While cautioning viewers against misleading router marketing like exaggerated antenna counts and unrealistic 16x16 MU-MIMO expectations, Linus concludes that even entry-level Wi-Fi 7 and Thunderbolt Share offer transformative networking and display capabilities.

## Atmosphere & style The video carries an energetic, tech-enthusiast vibe filled with Linus's signature blend of unscripted geeky excitement, dry skepticism, and hands-on demonstrations. The mood oscillates between genuine wonder—especially when witnessing uncompressed PC-to-PC game mirroring—and grounded consumer advocacy as he calls out deceptive router packaging and paywalled mesh networking gear. Watching the segment feels like getting an enthusiastic insider tour across a trade-show floor, pairing accessible analogies with sharp technical breakdowns.

## Bulletin - [00:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=0s) Intel showcased Thunderbolt Share, a prototype software technology that enables uncompressed, zero-latency PC-to-PC display streaming and direct file transfers reaching sustained speeds of 800 MB/s. - [03:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=180s) Thunderbolt Share avoids video compression artifacts by streaming directly between GPU frame buffers over a virtual network utilizing roughly 4 Gbps of bandwidth, at the cost of ~25% CPU overhead on an Intel Core Ultra 7 client. - [06:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=360s) Wi-Fi 7 (802.11be) demonstrated real-world client throughput ranging from 2 Gbps to past 4 Gbps, easily outpacing standard 1 Gbps wired Ethernet connections found on consumer hardware. - [08:32](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=512s) Key Wi-Fi 7 technical enhancements include Multi-Link Operation (MLO) for sub-1ms failover or band aggregation, 320 MHz channel widths in the 6 GHz spectrum, 4K-QAM modulation, and universal preamble puncturing.

## Key takeaways - [00:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=0s) Thunderbolt Share enables point-to-point connections between Windows PCs over Thunderbolt 4 and 5, bypassing traditional local area network routing bottlenecks. - [00:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=0s) Direct file transfers sustained approximately 800 MB/s during demonstrations, with theoretical ceilings reaching 20 Gbps on Thunderbolt 4 and up to 40 Gbps on Thunderbolt 5. - [03:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=180s) Unlike Steam Link or Moonlight, Thunderbolt Share bypasses video encoding and decoding steps, streaming raw frame buffer data with imperceptible latency and no compression artifacts. - [03:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=180s) Client-side processing of uncompressed remote video consumed roughly 25% CPU utilization on an Intel Core Ultra 7 155H processor. - [05:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=300s) Thunderbolt Share could allow outdated laptops or all-in-one PCs to serve as zero-latency external secondary monitors for primary workstations. - [06:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=360s) Wi-Fi 7 real-world performance delivers between 2 Gbps and over 4 Gbps, disproving expectations that its theoretical gains would not translate to consumer environments. - [06:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=360s) Elevating the theoretical Wi-Fi ceiling benefits real-world devices by preserving higher usable bandwidth across long distances and shortening airtime per transfer. - [08:32](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=512s) Multi-Link Operation (MLO) allows simultaneous connections across multiple radio bands, providing either split data streams for high throughput or low-latency routing down to 1 ms. - [08:32](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=512s) 4K-QAM packs 4,096 values per transmission symbol compared to 1024-QAM in Wi-Fi 6, securing an automatic 20% boost in raw data rates under clean RF conditions. - [10:49](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=649s) Preamble puncturing lets Wi-Fi 7 access points carve out narrow frequencies occupied by legacy hardware and aggregate the remaining non-contiguous bandwidth. - [11:38](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=698s) While Wi-Fi 7 raises the spatial stream limit to 16x16 MU-MIMO, consumer routers and client endpoints will rarely implement the full specification due to cost and hardware constraints. - [12:35](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=755s) External antenna counts on consumer routers do not indicate actual MIMO capabilities, making physical design an unreliable gauge of wireless stream performance.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=0s) Thunderbolt Share and Lossless PC-to-PC Streaming

Connecting two computers directly with a single Thunderbolt cable allows files to be dragged across at roughly 800 megabytes per second—enough throughput to transfer a modern *Call of Duty: Modern Warfare* install in under five minutes. While users traditionally cannot plug two Windows computers together for direct file transfers, Intel demonstrated a prototype software technology called Thunderbolt Share. Beyond raw file copying, the setup allows running a game on one PC and playing it completely losslessly on another screen with an input controller attached to the client machine. Linus highlights the broader industry context: “What you guys are looking at is, in my opinion, finally the answer to what exactly is it that makes Thunderbolt special now that Intel gave the spec to the USB-IF?”

Thunderbolt Share establishes a high-bandwidth, point-to-point network between two Thunderbolt 4 or Thunderbolt 5 Windows systems. Because Thunderbolt functions essentially as an external four-lane PCIe link bundled with DisplayPort, USB, and power delivery, it avoids the standard routing bottlenecks of local area networks. In a standard file transfer, data is pulled from the host SSD, packed into packets, routed through the Thunderbolt controller across the cable, unpacked on the receiving system, and written directly to the target SSD. Theoretical transfer speeds reach up to 20 Gbps over Thunderbolt 4 and up to 40 Gbps over Thunderbolt 5, which could slash game installation transfers to under sixty seconds. The only immediate trade-offs are elevated CPU usage during transfer bursts and the physical constraint of a one-meter passive cable limit.

```mermaid flowchart LR accTitle: Thunderbolt Share Data Pipeline accDescr: Diagram showing how Thunderbolt Share transfers game frames directly between GPU frame buffers without compression. subgraph Host["Host PC"] H_GPU["GPU Frame Buffer"] H_TB["Thunderbolt Controller"] end subgraph Cable["Physical Link"] TB_Wire["Thunderbolt Cable\n(20 to 40 Gbps)"] end subgraph Client["Client PC"] C_TB["Thunderbolt Controller"] C_GPU["GPU Frame Buffer"] C_Disp["Display Output"] end

H_GPU -->|"No Encoding / Uncompressed"| H_TB H_TB --> TB_Wire TB_Wire --> C_TB C_TB --> C_GPU C_GPU --> C_Disp ```

- Theoretical bandwidth: 20 Gbps on Thunderbolt 4, scaling up to 40 Gbps on Thunderbolt 5. - Direct file transfer demo achieved 800 MB/s sustained throughput. - Physical connection limit: passive cables are capped at one meter.

---

[03:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=180s) Eliminating Video Encoding and Display Latency

When testing PC-to-PC game mirroring over Thunderbolt Share, the latency is imperceptible to the naked eye. Traditional in-home streaming setups—such as Steam Link or Moonlight—force the rendering GPU to compress every frame using lossy video encoders before transmitting them across an Ethernet or Wi-Fi network, where the client device must decode and display the video. In contrast, Thunderbolt Share transmits uncompressed display data over the point-to-point link. Task Manager showed roughly 4 Gbps of virtual network traffic flowing over the connection with none of the compression artifacts typical of high-bitrate AV1 or HEVC video feeds.

To make this practical, Intel implemented a proprietary remote desktop protocol designed to package native display buffer signals directly into Thunderbolt data packets. Bandwidth is conserved by updating only the regions and elements of the screen that change between frames. System overhead remains noticeable; a receiving laptop equipped with an Intel Core Ultra 7 155H processor sat at approximately 25% CPU usage just to process the wide uncompressed incoming stream. Linus explains the core architectural advantage: “Intel says they are going straight from GPU frame buffer on this side to GPU frame buffer on the other side.” By avoiding system RAM dumps and disk buffering, the link provides zero-compression video mirroring alongside bidirectional keyboard, mouse, and game controller inputs.

- Streaming bandwidth utilization registered at around 4 Gbps in Windows Task Manager. - Client CPU overhead: ~25% utilization on an Intel Core Ultra 7 155H. - Direct frame buffer handoff eliminates both system RAM caching bottlenecks and video compression artifacts. - Bidirectional control channels pass display frames one way while returning input signals (gamepad, keyboard, mouse) the other way.

---

[05:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=300s) Practical Alpha Applications and Remote Compute Potential

Although Thunderbolt Share remains an early alpha project without a committed public release date, the ability to connect two PCs directly opens multiple functional use cases. Instead of carrying dedicated dual-screen hardware or external portable monitors, any secondary laptop or outdated all-in-one desktop with a Thunderbolt 4 or 5 port could operate instantly as an external, zero-latency monitor for a primary workstation. It also simplifies field collaboration, letting creators tether laptops together to offload high-resolution production footage or mirror workspaces on location without network routers.

Linus makes an open pitch to Intel to develop the underlying engine into a generalized remote desktop platform over standard optical connections. While setting up thin-client hardware tied to a central rack-mounted compute server over 25 Gbps fiber represents an extreme enthusiast setup, building native low-latency frame buffer streaming into Windows could replace cumbersome display capture and network bridging setups altogether.

> “This would turn every outdated all-in-one or laptop, as long as it has Thunderbolt 4 and up, into a monitor.” — Linus

---

[06:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=360s) Wi-Fi 7 (802.11be) Real-World Throughput and Bandwidth

Intel also demonstrated hardware running Wi-Fi 7 (802.11be), the industry standard formally branded as EHT (Extremely High Throughput). While previous Wi-Fi generations often promoted theoretical speed bumps that failed to translate into real-world gains under normal household conditions, early hardware tests using mesh kits and Intel client adapters demonstrated actual throughput ranging between 2 Gbps and past 4 Gbps. This performance exceeds the standard wired Gigabit Ethernet ports found on modern consumer motherboards and routers.

```mermaid flowchart TD accTitle: Wi-Fi 7 Throughput Factors accDescr: Breakdown of technologies that contribute to Wi-Fi 7's real-world transfer speeds exceeding 4 Gbps. WIFI7["Wi-Fi 7 (802.11be / EHT)"] WIFI7 --> C320["320 MHz Channel Width\n(Uncongested 6 GHz Band)"] WIFI7 --> QAM["4K-QAM Modulation\n(4096 Values per Symbol / +20% Data)"] WIFI7 --> MLO["Multi-Link Operation (MLO)\n(Aggregated Bands or Sub-1ms Failover)"] WIFI7 --> OFDMA["Enhanced OFDMA\n(Multiple Resource Units per Client)"] ```

Addressing skepticism that consumer internet service tiers cannot saturate multi-gigabit connections, Linus points out that peak wireless speeds fundamentally reshape network efficiency. Because real-world environments suffer from signal attenuation, distance drops, and environmental interference, a higher theoretical ceiling ensures usable connection rates remain elevated across difficult coverage areas. Furthermore, ultra-fast burst transmissions complete non-streaming transfers quicker, leaving airtime open for other devices. The massive throughput jump also enables wireless virtual reality headsets to stream uncompressed, low-latency PC-grade video feeds without tether cables.

- Real-world wireless speeds: verified at 2 Gbps, 3 Gbps, and over 4 Gbps on retail and test hardware. - Wireless transfers exceed standard 1 Gbps wired local connections. - Shortened airtime per transfer relieves overall spectrum congestion across active client devices.

---

[08:32](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=512s) Multi-Link Operation, 4K-QAM, and Resource Unit Allocation

A core technical pillar behind Wi-Fi 7's throughput is the expansion of the 6 GHz spectrum first introduced in Wi-Fi 6E. In addition to operating in clean, uncrowded airwaves, the 6 GHz band doubles the maximum channel width from 160 MHz to 320 MHz. The standout feature of the standard, however, is Multi-Link Operation (MLO), which allows a client to connect across multiple distinct frequency bands (such as 5 GHz and 6 GHz) at the same time:

> “MLO is sick. Depending on the radio configuration, it can enable either higher overall throughput as data can be broken up and transmitted over different bands simultaneously, or this is great for the gamers out there, lower latency as low as one millisecond by utilizing the first available band instead of being forced to wait its turn like it would on a single band.” — Linus

In addition to channel bonding and band aggregation, Wi-Fi 7 upgrades signal modulation to 4K-QAM (Quadrature Amplitude Modulation). By encoding digital bits onto radio signals through subtle shifts in signal phase and amplitude, 4K-QAM increases the density of each transmission symbol from 1,024 discrete values (in Wi-Fi 6's 1024-QAM) to 4,096 discrete values, providing an automatic ~20% raw throughput boost under clean RF conditions. Finally, Wi-Fi 7 refines Orthogonal Frequency Division Multiple Access (OFDMA). While Wi-Fi 6 established sub-carrier divisions grouped into Resource Units (RUs) assigned to single devices, Wi-Fi 7 permits assigning multiple Resource Units to a single client, ensuring unused spectrum capacity is not wasted when few devices are active.

- Maximum channel width doubled to 320 MHz within the 6 GHz frequency band. - Multi-Link Operation (MLO) supports packet splitting across concurrent bands or dynamic packet routing to achieve latency as low as 1 ms. - 4K-QAM encodes 4,096 values per symbol, generating a ~20% data rate improvement over 1024-QAM. - Multi-RU assignment allows individual clients to leverage spare sub-carrier spectrum across OFDMA channels.

[10:49](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=649s) Preamble Puncturing and Spectrum Efficiency

Linus highlights preamble puncturing as another crucial feature that is now enabled across the board, providing modern devices with a much smarter way to exploit radio spectrum even when legacy hardware is cluttering the airwaves. Under previous standards, if an older device was transmitting on a slice of a channel, modern hardware was severely hamstrung. Linus offers a concrete hypothetical: imagine an aging Wi-Fi security camera camping out on a 40 MHz chunk right in the middle of what would otherwise be a pristine, 160 MHz wide channel.

Under the old rules designed to prevent interference, a router would only be permitted to latch onto the single largest contiguous block of unused spectrum—meaning it could only utilize 80 MHz, stranding the remaining 40 MHz chunk on the other side completely abandoned. Preamble puncturing overhauls that behavior entirely by letting Wi-Fi 6 and Wi-Fi 7 access points identify the exact slice being occupied and punch a hole around it, aggregating all remaining unused pieces. As Linus puts it: “Well, preamble puncturing allows a Wi-Fi 6 or 7 device to recognize that the channel is partially occupied and to utilize all the unoccupied portions of that channel.” In his 160 MHz channel example, that means harvesting a total of 120 MHz of usable bandwidth instead of being artificially throttled to 80 MHz.

```mermaid flowchart TD accTitle: Preamble Puncturing Spectrum Utilization accDescr: Comparison between legacy channel allocation and preamble puncturing when an older device occupies 40 MHz in the middle of a 160 MHz channel. subgraph Legacy["Legacy Channel Allocation (Max Contiguous: 80 MHz)"] L1["Usable Block: 80 MHz"] L2["Occupied: 40 MHz Legacy Device"] L3["Wasted Block: 40 MHz"] end subgraph Punctured["Wi-Fi 6/7 Preamble Puncturing (Total Usable: 120 MHz)"] P1["Usable Block: 80 MHz"] P2["Punctured Out: 40 MHz Legacy Device"] P3["Reclaimed Block: 40 MHz"] end ```

- **Legacy channel penalty:** To avoid radio interference, older Wi-Fi protocols strictly required contiguous blocks, sacrificing any non-contiguous free spectrum. - **Spectrum reclaimed:** In a 160 MHz slice with a 40 MHz legacy obstruction in the middle, preamble puncturing recovers 120 MHz across both flanks rather than discarding the isolated 40 MHz block.

---

[11:38](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=698s) Spatial Streams and MU-MIMO Capabilities

Turning from channel width to spatial streams—the concurrent conversations occurring between access points and client devices—Linus dives into the world of multi-user, multiple input, multiple output, or MU-MIMO. While these configurations are often branded on retail router packaging as 2x2, 4x4, or "8x8 if you're a real baller," Wi-Fi 7 scales the theoretical ceiling all the way up to an astonishing 16 simultaneous client streams: 16x16 MU-MIMO.

However, Linus injects a dose of realism into these bleeding-edge paper specifications. Just because the standard permits 16x16 does not mean consumer-grade networking equipment will actually ship with it. He points out that Wi-Fi 6 on paper allowed up to 8x8, yet even premium consumer equipment like Netgear's $600 Nighthawk Wi-Fi 6E router topped out at 4x4, while virtually all client endpoints—from everyday laptops to high-end smartphones—remain limited to 2x2 configurations. As Linus summarizes the market landscape: “The Wi-Fi 7 spec now allows for up to 16 simultaneous clients worth of input and output. That is 16x16 MU-MIMO. But the reality of it is, most devices aren't going to implement that.”

- **MU-MIMO specification jump:** Wi-Fi 7 theoretically doubles the spatial stream cap over Wi-Fi 6, jumping from 8x8 up to 16x16 multi-user streams. - **Client and router reality:** High-end routers such as the $600 Netgear Nighthawk 6E only implemented 4x4, while most modern laptops and mobile devices are strictly 2x2.

---

[12:35](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=755s) The Antenna Trap and Realistic Wi-Fi 7 Expectations

Linus cautions viewers against falling for common retail router marketing gimmicks, specifically the assumption that an imposing spider-like router bristling with external antennas correlates directly to higher spatial stream counts. While spatial streams do have baseline physical antenna requirements—a 2x2 setup requires at least two antennas and a 4x4 requires four—the reverse logic is entirely false. Pointing to the eight-antenna ASUS ROG Rapture GT-AX11000 sitting on his desk, Linus explains that having eight antennas does not mean a router can actually execute the maximum 8x8 MU-MIMO configuration, and the same marketing sleight-of-hand will be pervasive with Wi-Fi 7. As Linus bluntly warns: “More antennae sticking out of a device does not automatically equal better MIMO performance.”

He stresses that consumer hardware pushing the theoretical outer bounds of the Wi-Fi 7 specification will be exceptionally rare and accompanied by astronomical price tags. Despite that, Linus argues that the everyday upgrades delivered by more sensible, conservative Wi-Fi 7 designs remain a massive leap forward. Even a basic budget-tier router supporting Multi-Link Operation (MLO) across just two bands still doubles the simultaneous frequency bands available to a device compared to older single-band links.

Wrapping up the video, Linus tosses to the sponsor, Wicked Cushions, touting their dual-layer cooling gel headphone cushions designed to keep ears from overheating like "steamed broccoli" during gaming sessions. In his closing remarks, he points viewers toward a ShortCircuit episode where he unboxed an Amazon eero Max 7 mesh unit. While expressing frustration over its high price and subscription paywalls, he admits that Wi-Fi 7's raw speed is staggering: “It's super expensive, and some of the features are locked behind a paywall that I'm kind of not that into, but it's fast. It's really fast. Like even on a phone, like more than two and a half gigabit, what?”

- **Antenna correlation myth:** External antenna count is not a 1:1 indicator of MIMO capabilities; an eight-antenna chassis like the ROG Rapture GT-AX11000 does not inherently deliver 8x8 performance. - **Entry-level Wi-Fi 7 value:** Even standard two-band MLO implementations on entry-level hardware represent a doubling of active radio links for end users. - **Real-world throughput:** Hands-on testing of early Wi-Fi 7 hardware (such as the eero Max 7) demonstrated speeds exceeding 2.5 Gbps directly on a smartphone.

## Notable quotes > [12:35](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=755s) More antennae sticking out of a device does not automatically equal better MIMO performance. — Linus

> [10:49](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=649s) Preamble puncturing allows a Wi-Fi 6 or 7 device to recognize that the channel is partially occupied and to utilize all the unoccupied portions of that channel. — Linus

> [11:38](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=698s) The Wi-Fi 7 spec now allows for up to 16 simultaneous clients worth of input and output. That is 16x16 MU-MIMO. But the reality of it is, most devices aren't going to implement that. — Linus

## Mentioned Linus Tech Tips, Linus, Intel, CES 2024, Thunderbolt, Thunderbolt Share, Thunderbolt 4, Thunderbolt 5, Windows, Call of Duty: Modern Warfare, USB-IF, Steam Link, Moonlight, Intel Core Ultra 7 155H, Wi-Fi 7, Wi-Fi 6, Wi-Fi 6E, Netgear, Nighthawk, ASUS ROG Rapture GT-AX11000, Wicked Cushions, ShortCircuit, Amazon, eero Max 7

## Video description Check out Wicked Cushions at Amazon: https://lmg.gg/WickedCushionsAmazon or their website: https://wickedcushions.com/LTT

Intel showed us the goods on probably the single thing I'm most excited about from CES 2024, and it's...Thunderbolt? Seriously, this is so cool. Not only that, but Wi-Fi 7 seems like it will actually be an increase that actually matters to regular people- as well as niche corner-case weridos like me lol

Discuss on the forum: https://linustechtips.com/topic/1552428-this-was-under-nda-until-i-called-in-a-favor-intel-ces-2024/

► GET MERCH: https://lttstore.com ► GET EXCLUSIVE CONTENT ON FLOATPLANE: https://lmg.gg/lttfloatplane ► SPONSORS, AFFILIATES, AND PARTNERS: https://lmg.gg/partners ► EQUIPMENT WE USE TO FILM LTT: https://lmg.gg/LTTEquipment ► OUR WAN PODCAST GEAR: https://lmg.gg/wanset

FOLLOW US --------------------------------------------------- Twitter: https://twitter.com/linustech Facebook: http://www.facebook.com/LinusTech Instagram: https://www.instagram.com/linustech TikTok: https://www.tiktok.com/@linustech Twitch: https://www.twitch.tv/linustech

MUSIC CREDIT --------------------------------------------------- Intro: Laszlo - Supernova Video Link: https://www.youtube.com/watch?v=PKfxmFU3lWY iTunes Download Link: https://itunes.apple.com/us/album/supernova/id936805712 Artist Link: https://soundcloud.com/laszlomusic

Outro: Approaching Nirvana - Sugar High Video Link: https://www.youtube.com/watch?v=ngsGBSCDwcI Listen on Spotify: http://spoti.fi/UxWkUw Artist Link: http://www.youtube.com/approachingnirvana

Intro animation by MBarek Abdelwassaa https://www.instagram.com/mbarek_abdel/ Monitor And Keyboard by vadimmihalkevich / CC BY 4.0 https://geni.us/PgGWp Mechanical RGB Keyboard by BigBrotherECE / CC BY 4.0 https://geni.us/mj6pHk4 Mouse Gamer free Model By Oscar Creativo / CC BY 4.0 https://geni.us/Ps3XfE

CHAPTERS --------------------------------------------------- 0:00 Intro 1:21 Thunderbolt Share 3:00 Latency 3:45 It's not encoded! 5:00 Why I'm excited 6:00 Wi-Fi 7 is legit. 9:00 "MLO. Is. Sick." 10:43 Preamble Puncturing 11:40 Spatial Streams 12:30 Don't get fooled again! 14:03 outro

[Source](https://www.youtube.com/watch?v=GqCwLjhb4YY)

中文简报

英特尔推出Thunderbolt Share实现PC间直接视频流式传输英特尔全新的Thunderbolt Share协议通过彻底绕过网络压缩证明了该接口的高端价值，同时其Wi-Fi 7硬件提供了最终超越传统千兆以太网的现实无线速度。 Linus考察了英特尔的Thunderbolt Share原型技术，展示了两台Windows PC之间的点对点连接如何实现800 MB/s的文件传输，以及直接跨GPU帧缓冲区的零压缩、低延迟显示镜像。他还解析了Wi-Fi 7（802.11be）的实际优势，重点介绍了经实测超过2至4 Gbps的速度，以及320 MHz信道、4K-QAM调制、多链路操作（MLO）和前导码打孔等技术飞跃。尽管Linus告诫观众警惕夸大天线数量和不切实际的16x16 MU-MIMO预期等误导性路由器营销，但他总结认为，即便入门级Wi-Fi 7和Thunderbolt Share也能带来颠覆性的网络与显示功能体验。英特尔展示了Thunderbolt Share，这是一种原型软件技术，可实现无压缩、零延迟的PC到PC显示串流，以及持续速度达800 MB/s的直接文件传输。 (0:00) Thunderbolt Share通过利用约4 Gbps带宽的虚拟网络在GPU帧缓冲区之间直接串流，从而避免了视频压缩失真，代价是在英特尔酷睿Ultra 7客户端上产生约25%的CPU开销。 (3:00) Wi-Fi 7（802.11be）展示了2 Gbps至超过4 Gbps的真实客户端吞吐量，轻松超越了消费级硬件上常见的标准1 Gbps有线以太网连接。 (6:00) Wi-Fi 7的关键技术升级包括用于实现1毫秒以下故障转移或频段聚合的多链路操作（MLO）、6 GHz频段内的320 MHz信道带宽、4K-QAM调制以及通用前导码打孔。 (8:32)

Reading summary

Intel demonstrated prototype Thunderbolt Share software for uncompressed, point-to-point PC-to-PC streaming and high-speed data transfers, alongside real-world hardware implementations of the emerging Wi-Fi 7 standard. Linus Tech Tips evaluates how direct frame-buffer connections and multi-gigabit wireless protocols solve persistent network latency and bandwidth bottlenecks. Direct Point-to-Point Interconnects and Lossless Display Mirroring (0:00) Thunderbolt Share establishes an ad-hoc, point-to-point local connection between two Thunderbolt 4 or Thunderbolt 5 Windows PCs over a single cable. In a live demonstration, direct file copying achieved sustained throughput of roughly 800 megabytes per second, bypassing standard local area network routing. Theoretical connection caps reach 20 Gbps over Thunderbolt 4 and 40 Gbps over Thunderbolt 5, though current setups require passive cables capped at one meter and show brief spikes in host CPU utilization. Beyond basic storage transfers, the software mirrors display output and peripheral inputs with imperceptible latency. Unlike standard remote desktop software or in-home streaming tools like Steam Link that rely on lossy HEVC or AV1 video encoding, Thunderbolt Share routes native display buffer data directly from one GPU frame buffer to the target system's GPU frame buffer. The prototype transmitted an uncompressed, dynamic feed drawing approximately 4 Gbps in Task Manager while generating around 25% CPU overhead on an Intel Core Ultra 7 155H client laptop. Hardware Repurposing and Remote Compute Potential (5:00) Because the protocol operates without video compression artifacts or perceptible input lag, Linus highlights its utility for repurposing older hardware. Any secondary laptop or outdated all-in-one desktop equipped with a Thunderbolt 4 or 5 port can function as a zero-latency external monitor for a primary workstation, eliminating the need to purchase dedicated portable screens. The architecture also facilitates high-bandwidth field transfers, enabling creators to dump large video assets directly between mobile workstations without local network switches. Linus pitches this direct frame-buffer streaming model to Intel as a potential foundation for centralized compute environments, where lightweight thin-client laptops could interface with centralized, rack-mounted hardware across high-speed optical connections. Wi-Fi 7 Throughput Gains and Modulation Upgrades (6:00) Intel also showcased hardware operating on Wi-Fi 7 (802.11be / Extremely High Throughput). Early hardware benchmarks demonstrated real-world throughput ranging between 2 Gbps and beyond 4 Gbps, exceeding standard 1 Gbps wired Ethernet connections. Linus argues that high theoretical bandwidth floors remain valuable in consumer environments because they maintain usable speeds through wall attenuation, signal decay, and environmental noise, while faster burst transmissions free up airtime for concurrent devices. Wi-Fi 7 achieves these gains by expanding 6 GHz channel widths from 160 MHz to 320 MHz and upgrading radio modulation to 4K-QAM. By packing 4,096 discrete signal states per transmission symbol—compared to 1,024 in Wi-Fi 6—the specification yields an immediate ~20% increase in raw data rates under clean RF conditions. Additionally, refined OFDMA allows multiple Resource Units to be allocated to a single client, preventing unused spectrum from sitting idle. Band Aggregation and Spectrum Preamble Puncturing (8:32) A central feature of the Wi-Fi 7 specification is Multi-Link Operation (MLO). Rather than binding a device to a single frequency, MLO allows clients to connect across multiple bands (such as 5 GHz and 6 GHz) concurrently. Depending on router configuration, MLO can divide and stream data packets across bands to maximize aggregate throughput, or dynamically select the first available band to drive transmission latency down to approximately one millisecond. Spectrum utilization is further improved through preamble puncturing. Under legacy rules, an active older device occupying a narrow slice—such as a 40 MHz channel in the middle of a 160 MHz band—forced routers to fall back to the largest single contiguous block, effectively discarding remaining spectrum. Wi-Fi 6 and 7 devices can carve out the occupied slice and aggregate the non-contiguous free blocks, reclaiming up to 120 MHz of usable bandwidth in that same scenario. Consumer Hardware Limits and Router Antenna Marketing (11:38) While Wi-Fi 7 scales its theoretical multi-user specification up to 16x16 MU-MIMO—doubling the 8x8 ceiling of Wi-Fi 6—Linus cautions that retail consumer hardware will rarely hit these ceilings. Most client devices like laptops and smartphones remain capped at 2x2 configurations, and premium consumer routers typically top out at 4x4. Linus also warns against correlating external antenna count with actual spatial stream performance, pointing out that an eight-antenna router does not inherently offer 8x8 MU-MIMO. Despite marketing hyperbole and expensive initial hardware releases, Linus notes that practical baseline implementations—such as basic two-band MLO on affordable routers—deliver meaningful connectivity upgrades, with retail mesh systems already demonstrating real-world speeds exceeding 2.5 Gbps directly on smartphones. Intel's Thunderbolt Share demonstrates viable lossless display mirroring and high-bandwidth transfers over direct cable connections, though it remains confined to an early alpha state with one-meter passive cable constraints. Meanwhile, while bleeding-edge Wi-Fi 7 metrics like 16x16 MU-MIMO will remain largely theoretical for consumer clients, standard features like MLO, 320 MHz channels, and preamble puncturing reliably push multi-gigabit performance beyond conventional wired home networking.

精炼阅读摘要

Intel 展示的原型软件 Thunderbolt Share 通过点对点直连绕过局域网与视频编码损耗，实现了跨 PC 的无损帧缓冲区串流；与此同时，Wi-Fi 7（802.11be）借助 320 MHz 频宽、MLO 多链路聚合与前导码打孔等物理层革新，使消费级实际无线吞吐突破 4 Gbps，但在多空间流等极端参数上仍面临终端支持与营销宣称的现实脱节。雷电直连与无损低延迟串流 (0:00) 根据 Intel 展示的 Thunderbolt Share 技术，两台配备雷电 4 或雷电 5 的 Windows PC 仅通过单根雷电线缆即可建立高带宽点对点连接。现场实测文件持续传输速率约为 800 MB/s，理论带宽在雷电 4 下可达 20 Gbps，在雷电 5 下可达 40 Gbps。由于雷电接口实质上是集成了 DisplayPort、USB 和供电的外置四通道 PCIe 链路，该方案避开了传统局域网的网络路由限制。不过，该技术的物理局限在于被动线缆长度被限制在 1 米之内，且突发传输会伴随明显的 CPU 占用。在串流机制上，Thunderbolt Share 摒弃了 Steam Link 或 Moonlight 等依赖有损视频编解码的传统方案，转而采用专有远程协议直接提取原生显示缓冲信号，且仅传输发生变动的屏幕区域。Linus 解释其核心架构在于数据直接在两端 GPU 帧缓冲区之间传递，完全规避了内存转储与压缩噪点。实测 Windows 任务管理器显示约 4 Gbps 的虚拟网络流量，画面无压缩损失，接收端配备 Intel Core Ultra 7 155H 处理器的笔记本 CPU 占用率约为 25%，并原生支持手柄、键盘与鼠标的双向控制输入。 Wi-Fi 7 物理层与实测吞吐突破 (6:00) 在无线网络方面，Intel 展示了 Wi-Fi 7（802.11be / EHT）的实机运行表现。测试中基于网状套件与 Intel 客户端网卡，实测吞吐量达到 2 Gbps 至 4 Gbps 以上，超越了消费级硬件常见的千兆有线以太网口。Linus 指出，超高理论无线带宽的核心价值在于增强抗衰减能力，并大幅压缩单次传输的空口占用时间（Airtime），从而缓解整网拥堵，并为无线 VR 设备的无压缩超低延迟串流提供支撑。 Wi-Fi 7 吞吐跃升依赖于多项核心技术：首先，在 6 GHz 频段中将最大信道宽度翻倍至 320 MHz；其次，调制模式升级为 4K-QAM（每个符号编码 4096 个离散值），相比 Wi-Fi 6 带来约 20% 的原始吞吐增益。此外，多链路操作（MLO）允许客户端跨 5 GHz 与 6 GHz 等频段并发传输以聚合吞吐，或动态选择首个可用频段将延迟降低至最低 1 毫秒；增强型 OFDMA 则支持向单个客户端分配多个资源单元（Multi-RU），确保空闲频谱得到充分调度。前导码打孔与频谱利用优化 (10:49) 为了在杂乱无线环境中提升频谱利用率，Wi-Fi 7 全面启用了前导码打孔（Preamble Puncturing）技术。传统 Wi-Fi 规范为避免干扰，强制要求信道必须使用连续频段；一旦中间频段被老旧设备切断，路由器便只能退化使用单侧最大的一段连续空闲频谱。 Linus 以具体情境说明了该技术的作用：若一台老旧监控摄像头占用了 160 MHz 信道正中间的 40 MHz 频宽，旧规则会导致设备只能退回至单侧连续的 80 MHz，白白废弃另一侧孤立的 40 MHz 频宽。而前导码打孔技术允许设备精准避让该被占用的 40 MHz 切片，将两侧空闲部分重新聚合，从而在 160 MHz 信道中提取出 120 MHz 的可用带宽，显著降低了老旧设备对高速频宽的惩罚性限制。空间流落地现实与天线营销误区 Wi-Fi 7 规范在理论上将多用户多输入多输出（MU-MIMO）上限推至 16x16，支持多达 16 条并发空间流，相较 Wi-Fi 6 的 8x8 上限翻倍。然而 Linus 强调，纸面规格与消费级现实存在巨大断层。回顾 Wi-Fi 6 时代，即便售价 600 美元的 Netgear Nighthawk 6E 等高端路由器也仅实现 4x4，而绝大多数主流笔记本与智能手机受体积功耗所限仍维持在 2x2 配置，16x16 规格在消费端很难实际普及。 Linus 同时提醒消费者警惕路由器的多天线营销话术。虽然 2x2 和 4x4 配置具有最少天线数量的硬性门槛，但外部天线数量与空间流性能并非等价对应；例如配备 8 根天线的华硕 ROG Rapture GT-AX11000 并不具备 8x8 MU-MIMO 性能。不过，即便是不追求极端规格的入门级双频 MLO 路由器，其实际活跃链路也较传统单频实现了翻倍；正如他在开箱售价高昂的 Amazon eero Max 7 网状路由时发现，手机端实测无线速率已突破 2.5 Gbps，基础升级效益依然显著。本期内容展示了两项传输技术的演进潜能与现实边界：Intel Thunderbolt Share 验证了 GPU 帧缓冲区直连在消除延迟与屏幕扩展上的可行性，但目前受制于 Alpha 阶段定位、1 米线长以及 CPU 开销；Wi-Fi 7 则凭借 320 MHz、MLO 与打孔机制实现了超越千兆有线的无线速率，但消费者在面对 16x16 等顶配参数与多天线营销时，仍需看清终端适配能力的实际局限。

Detailed notes — zh

## Background 该视频由知名科技频道 Linus Tech Tips 于 2024 年 1 月 12 日发布在 YouTube 平台上。视频简介指出，英特尔在 CES 2024 展会上向他们展示了最令人兴奋的技术成果，涵盖了颠覆认知的全新 Thunderbolt 功能，以及为普通用户和极客人群带来实质性速度与体验提升的 Wi-Fi 7 技术。 ## TL;DR Linus 实地体验并深度剖析了英特尔的两大互连前沿技术：Thunderbolt Share 与 Wi-Fi 7。Thunderbolt Share 允许两台 PC 仅凭一根雷电线缆建立点对点直连，通过 GPU 显存直通和无损切片实现高达 800 MB/s 的文件拷贝及肉眼零延迟的双向游戏串流。在无线网络领域，Wi-Fi 7 现场实测吞吐量突破 4 Gbps，视频深入解析了 320 MHz 超大频宽、MLO 多链路聚合、4K-QAM、前导码打孔等核心升级，同时也提醒消费者警惕天线营销陷阱与 MU-MIMO 实际落地规格的缩水。 ## Atmosphere & style 整期视频充斥着 Linus 标志性的极客狂热与技术探究欲，开场即用“线缆直连跨设备秒传与零延迟游戏”带来极强的冲击感。叙述节奏明快紧凑，既有对突破性黑科技的大声惊呼与热情赞美，也有面对厂商营销话术时犀利冷静的泼冷水与理性拆解。观众在观看过程中不仅能直观感受到硬件跃升带来的爽快感，还能在深入浅出的技术图解中获得扎实硬核的科普收获。 ## Bulletin - [01:21](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=81s) 英特尔在 CES 2024 展示了名为 Thunderbolt Share 的点对点直连技术，支持两台 PC 间通过雷电线缆实现最高 40 Gbps 的极速文件互传，以及基于 GPU 显存直通且肉眼零延迟的无损游戏串流。 - [03:45](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=225s) Thunderbolt Share 采用定制远程桌面协议与局部更新机制，彻底舍弃传统有损视频编码，串流吞吐高达 4 Gbps，实现双向屏幕共享与手柄低延迟反向控制。 - [06:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=360s) 现场实测 Wi-Fi 7 传输吞吐达到 2 Gbps 至 4 Gbps，配合 320 MHz 频宽、多链路聚合（MLO）、4K-QAM 及前导码打孔等新技术，带来空口延迟低至 1 毫秒的实质代际跨越。 - [11:37](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=697s) 尽管 Wi-Fi 7 规范支持高达 16x16 MU-MIMO，但现实中移动终端普遍仅支持 2x2，路由器外置天线数量也并不等同于 MIMO 流数，消费者无需盲目追求顶级规格即可从基础双频 MLO 中获益。 ## Key takeaways - [00:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=0s) Thunderbolt Share 是英特尔开放雷电规范给 USB-IF 后，重新定义雷电接口独特价值的关键专属软件技术。 - [01:21](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=81s) 雷电 4 与雷电 5 理论传输速率分别达 20 Gbps 与 40 Gbps，实测大文件互传可达 800 MB/s，大幅简化了离线、出差等特殊场景下的跨 PC 高速传输流程。 - [03:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=180s) 不同于 Steam Link 或 Moonlight 等依赖 GPU 有损压缩编码的传统串流方案，Thunderbolt Share 彻底舍弃了有损编码。 - [03:45](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=225s) Thunderbolt Share 的串流数据直接在发送端与接收端的 GPU 帧缓冲区之间流转，全程绕过主机内存中转并以 4 Gbps 带宽传输无损画面。 - [03:45](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=225s) 除无损画面传输外，Thunderbolt Share 还支持极低延迟的双向控制，可直接将闲置的雷电设备转变为高性能便携外接屏。 - [06:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=360s) Wi-Fi 7 实测吞吐稳定在 2 Gbps 到 4 Gbps，超越了绝大多数普通家庭的有线网络配置，显著减少空中信道占用时间并赋能高清无线 VR 等场景。 - [08:32](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=512s) Wi-Fi 7 在 6 GHz 频段下将信道频宽上限翻倍扩展至 320 MHz，彻底摆脱了传统 2.4 GHz 与 5 GHz 频段的拥挤瓶颈。 - [09:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=540s) 多链路聚合（MLO）技术允许终端设备跨频段并发传输以拔高总吞吐，或择优选择空闲频段将空口通信延迟压缩至 1 毫秒。 - [09:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=540s) Wi-Fi 7 升级至 4K-QAM 调制方式，单符号数据承载能力显著提升，在相同物理射频条件下带来约 20% 的纯理论速率增益。 - [10:49](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=649s) 前导码打孔（Preamble Puncturing）技术允许设备跳过老旧设备占用的局部信道，整合剩余碎片频宽，大幅提升频谱利用率。 - [11:37](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=697s) Wi-Fi 7 标准虽支持最高 16x16 MU-MIMO，但由于手机与笔记本受限于体积功耗普遍维持 2x2，用户不应指望硬件能跑满纸面理论流数。 - [12:34](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=754s) 路由器天线数量与 MU-MIMO 性能无绝对对应关系，即便选购仅支持双频 MLO 的入门级 Wi-Fi 7 路由，相较以往依然能获得翻倍的并发体验升级。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=0s) 令人震惊的雷电直连体验 Linus 在现场直接拿出一根普通的 Thunderbolt 线缆，将两台电脑连在一起并直接拖拽超大文件，传输速率瞬间飙到了 800 MB/s。Linus 兴奋地惊呼这种速度相当于在五分钟内装好最新款的《使命召唤：现代战争》（Modern Warfare）。他调侃观众此刻肯定激动得从椅子上跳起来想亲自去试，但大家都清楚，通常情况下两台 Windows 电脑根本无法仅凭一根线缆就做到如此无缝的文件互传，更不用说在其中一台机器上运行游戏，还能完全无损地在另一台上进行游玩。 Linus 认为这项技术完美解答了长久以来的疑问：在英特尔（Intel）将 Thunderbolt 技术规范开放给 USB-IF 之后，雷电到底还有什么独特之处？正如 Linus Sebastian 所总结的：“What you guys are looking at is, in my opinion, finally the answer to what exactly is it that makes Thunderbolt special now that Intel gave the spec to the USB-IF?” 这一演示不仅展示了极高带宽，还引发了对无损编码实现方式与超低延迟的技术探讨。 ### [01:21](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=81s) Thunderbolt Share 的工作机制与文件传输这项黑科技被正式命名为 Thunderbolt Share。本质上，这是一套在两台搭载 Thunderbolt 4 或 Thunderbolt 5 的 Windows PC 之间建立点对点网络连接的专用软件。雷电接口本身就相当于一个外置的 4 通道 PCIe（four-lane-wide PCIe）通道，不仅整合了外设 USB 数据、显示器 DisplayPort 信号以及单线供电扩展能力，还兼具极高带宽与超低延迟特性。在最基础的文件传输场景中，数据流极为清晰直接：第一台设备的 SSD 读取文件并打包成数据包，通过雷电主控芯片和雷电线缆传输至对端，再由目标设备解包并直接写入 SSD。尽管 Linus 本人一直崇尚中心化的局域网络方案，但他坦言在特殊场景下（如在酒店急需向搭乘早班机离开的同事备份视频素材，或在局域网联机派对上需要快速分享游戏安装包），这种点对点直连有着不可替代的便利性。 - Thunderbolt 4 理论传输速率最高可达 20 Gbps。 - Thunderbolt 5 理论传输速率最高可达 40 Gbps，足以将《现代战争》级别的庞大游戏拷贝时间压缩至一分钟以内。 - 该方案的物理限制主要是无源（passive）雷电线缆长度通常被限制在 1 米之内，且传输时伴随较高的 CPU 占用率。 ```mermaid flowchart LR accTitle: Thunderbolt Share 文件传输流程图 accDescr: 展示文件从一台电脑的 SSD 经雷电通道传输至另一台电脑的过程 subgraph PC1["发送端 PC"] A["本地 NVMe SSD"] -->|"读取并打包数据"| B["Thunderbolt 芯片"] end B -->|"雷电 4/5 线缆 (超高带宽通道)"| C["Thunderbolt 芯片"] subgraph PC2["接收端 PC"] C -->|"解包并直接写入"| D["目标 NVMe SSD"] end ``` ### [03:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=180s) 颠覆常规的零延迟游戏串流体验如果说高速传文件只是常规操作，那么接下来的游戏串流演示则完全超出了预期。Linus 在两台机器间启动串流后惊叹不已，指出即便是在 10G 甚至 25G 的万兆光纤网络下，也从未见过延迟如此之低的跨电脑游戏串流。正如 Linus Sebastian 描述的那样：“I mean, I obviously haven't measured it with a high-speed camera yet. I don't have one in my pocket, but to the eye, I would describe the latency as none.” 造成这种质的飞跃的原因在于数据处理逻辑的彻底变革。传统的串流方案（如 Steam Link、Moonlight 等）高度依赖本地 GPU 渲染每一帧画面，将其压缩编码为有损画质的数据流，传输至接收端后再解压输出到显示屏，整个过程层层损耗且不可避免地增加延迟。相比之下，Thunderbolt Share 在传输链路中完全舍弃了有损编码。 ### [03:45](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=225s) “真·无损”传输与显存直通黑科技为了验证是否真的没有经过传统编码压缩，Linus 仔细观察了屏幕细节。即便与画质极高的 AV1 编码相比，雷电镜像画面中也看不到任何压缩伪影。打开任务管理器后可以发现，该虚拟链路正以高达 4 Gbps 的惊人带宽实时吞吐未压缩数据。英特尔在此处研发了一套定制的远程桌面协议，专门负责将显示输出切片为雷电数据包。为了节省部分带宽，该协议采用了仅刷新远端屏幕发生变动区域的局部更新机制。不过巨大的吞吐量也带来了硬件负载：接收端搭载的 Core Ultra 7 155H 处理器被占用了约 25% 的 CPU 使用率，但真正的技术杀手锏在于 GPU 的协同运作。英特尔透露，数据直接从发送端的 GPU 帧缓冲区（frame buffer）直连到接收端的 GPU 帧缓冲区，既避开了本地存储瓶颈，甚至全程绕过了主机内存（RAM）中转。 ```mermaid flowchart TD accTitle: Thunderbolt Share 显存直通流架构 accDescr: 对比传统串流方式与 Thunderbolt Share 的显存直通路径 subgraph Traditional["传统串流 (Steam Link / Moonlight)"] T1["GPU 渲染"] --> T2["视频压缩编码 (有损/延迟)"] T2 --> T3["网络传输"] T3 --> T4["客户端解码并显示"] end subgraph TBShare["Thunderbolt Share 方案"] S1["源端 GPU 帧缓冲区"] -->|"定制协议切片"| S2["Thunderbolt 极速物理链路"] S2 -->|"4 Gbps 局部更新"| S3["目标端 GPU 帧缓冲区"] end ``` - 该软件演示时仍处于 Alpha 测试阶段，尚未公布具体发布时间。 - 该连接支持完全的双向控制：一台机器的屏幕画面可以投射过来，接收端连接的手柄也可以无感反向操控对端电脑。 - 适用场景广泛：无需额外购买笨重的双屏笔记本，插上线即可与同事完全共享屏幕操作；任何带有 Thunderbolt 4 及以上接口的旧体机或闲置笔记本都可以一键化身为外接便携显示器。 - Linus 甚至隔空向英特尔建言，希望他们单独推出这款远程桌面应用，好让他通过全屋 25G 光纤网络配合机柜服务器打造终极瘦客户端（thin client）环境。 ### [06:00](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=360s) Wi-Fi 7：真正带来代际跨越的无线革命将目光从有线连接转向无线技术，英特尔现场展示的 Wi-Fi 7 同样带来了实质性的代际飞跃。Linus 坦承自己对无线网络一贯持怀疑态度，但此前在 Short Circuit 频道开箱体验 Eero Mesh Wi-Fi 7 套件时，其性能表现彻底颠覆了他的认知。展台现场测得的 Wi-Fi 7 实际传输吞吐量稳定在 2 Gbps 到 3 Gbps，甚至一度突破了 4 Gbps。这样的实际无线速率已经彻底超越了当前 99% 的普通家庭有线千兆网，完全对得起 Wi-Fi 7 代号“EHT”（Extremely High Throughput，超高吞吐量）的称号。针对很多人认为“宽带速率赶不上 Wi-Fi、所以速度再高也没用”的质疑，Linus 指出两点关键现实： 1. 无线信号在实际穿墙和远距离传输中必然会衰减，只有基础吞吐上限大幅跃升，衰减后的“残存速度”才会保持在相当高的水平。 2. 非流媒体类的数据传输如果能在极短时间内瞬间完成，就能大幅减少空中信道占用时间（air time），从而从根本上缓解多设备信道拥堵。 3. 这种级别的带宽将解锁全新的应用生态，比如摆脱线缆束缚、达到接近原生画质的高清 PC VR 头显体验。 ### [08:32](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=512s) 320 MHz 频宽、MLO 与 4K-QAM 技术解析过去几代 Wi-Fi 技术的演进重心往往放在增加设备并发接入数量上，而普通用户单机体验到的网速提升有限。Wi-Fi 7 则在单机速率和降低延迟上实现了真正的实质突破。这一代全面继承并释放了 6 GHz 频段的潜力，不仅避开了 2.4 GHz 与 5 GHz 频段的严重拥挤，更将最大信道频宽（channel width）直接翻倍扩大至 320 MHz。更具革命性的是名为多链路聚合（Multi-Link Operation，简称 MLO）的技术。正如 Linus 兴奋喊出的：“MLO is sick.” 依托 MLO 技术，终端设备不再局限于单一频段通信： - **聚合高吞吐模式**：数据可拆解后跨不同频段（如 5 GHz 和 6 GHz）同时并行传输，大幅拔高总带宽。 - **超低延迟模式**：电竞玩家的数据包无需像以往那样在单个频段排队等候，而是哪个频段空闲就走哪个，能将空口延迟压缩至惊人的 1 毫秒（1 ms）。此外，Wi-Fi 7 将调制方式升级为 4K-QAM（4096 正交振幅调制）。传统的数字信号转换为模拟射频信号时，Wi-Fi 6 的 1024-QAM 每个符号（symbol）最多只能编码 1024 个离散数值，而 4K-QAM 将每个符号的数据承载上限提升至 4096 个数值。在相同物理条件下，仅此一项改动就让传输速率纯粹提升了约 20%。最后在多设备分配方面，Wi-Fi 6 引入的 OFDMA 机制将 Wi-Fi 信道拆分为多个子载波（sub-carriers）并组合为资源单元（Resource Unit, RU）指派给各终端；而 Wi-Fi 7 进一步支持为单个客户端同时分配多个资源单元（Multiple RUs），从而确保在网络空闲、连接客户端较少时，剩余带宽资源不会白白闲置浪费。 ### [10:49](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=649s) 前导码打孔（Preamble Puncturing）解决信道碎片化现代无线网络中最令人头疼的问题之一，就是老旧设备占着信道导致带宽浪费。Linus 解释道，如今全行业全面启用的前导码打孔（Preamble Puncturing）技术彻底改变了这一现状，它允许现代设备直接利用那些部分被老旧设备占用的信道。举个典型的场景：假设一个老旧的 Wi-Fi 监控摄像头占用了某条原本空闲的 160 MHz 超宽信道正中间的 40 MHz。在旧标准的规则下，设备为了避免相互干扰，只能退而求其次去使用信道中“最大且连续的未占用部分”，这意味着用户最终只能调用 80 MHz 带宽，剩下的 40 MHz 空间则被白白浪费掉。而前导码打孔技术允许 Wi-Fi 6 或 Wi-Fi 7 设备智能识别出信道已被局部占用，并直接跳过被占用的频段，将剩余所有未被占用的碎片带宽通通压榨出来。在上述案例中，设备就能完整用上 120 MHz 的可用带宽，让无线频谱的利用效率大幅提升。 ```mermaid flowchart TD accTitle: 前导码打孔信道利用对比 accDescr: 对比传统 Wi-Fi 与启用前导码打孔后在部分信道被占用时的带宽利用情况 subgraph Channel["160 MHz 信道结构"] A1["未占用频段 (80 MHz)"] A2["旧设备占用频段 (40 MHz)"] A3["未占用频段 (40 MHz)"] end subgraph Legacy["传统标准处理方式"] L1["仅能利用最大连续频段: 80 MHz"] L2["浪费剩余孤立频段: 40 MHz"] end subgraph Puncturing["前导码打孔技术"] P1["打孔绕过占用频段"] P2["聚合所有剩余频段: 120 MHz (80 + 40)"] end A1 -.-> L1 A3 -.-> L2 A1 & A3 ==> P1 --> P2 ``` - 传统处理机制：遇干扰时只能退回最大连续未占用频段（如 160 MHz 信道被居中切断时只能用 80 MHz）。 - 前导码打孔机制：跨过被占用的 40 MHz 频段，同时聚合两侧的 80 MHz 与 40 MHz，榨取出整整 120 MHz 带宽。 ### [11:37](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=697s) 空间流与 MU-MIMO 的现实差距无线路由器（AP）与客户端设备之间的通信通道通常被称为空间流（Spatial Streams）。现代 AP 具备同时处理多条空间流的能力，而包装盒上往往印着一个听起来十分拗口的专业术语——MU-MIMO（多用户多输入多输出，Multi-User, Multiple-Input, Multiple-Output）。市面上常见的规格通常标示为 2x2、4x4，如果是顶级土豪级设备，甚至能看到 8x8 的身影。在 Wi-Fi 7 的技术规范中，MU-MIMO 的上限被一举拉高到了惊人的 16x16，意味着理论上能同时承载多达 16 个客户端的并发吞吐。然而 Linus 毫不留情地给消费者的狂想泼了一盆冷水，指出现实中大部分终端设备根本不会实装这项满血规格。Linus 直言：“Wi-Fi 6 supported 8x8, but even a $600 Wi-Fi 6E router like Netgear's Nighthawk only implemented 4x4, and your laptop or smartphone is probably only 2x2.”（Wi-Fi 6 早就支持 8x8 了，但即使是一台售价高达 600 美元的 Netgear Nighthawk Wi-Fi 6E 路由器也仅仅做了 4x4，而你的笔记本电脑或智能手机大概率只有 2x2）。 - MU-MIMO 规范上限：Wi-Fi 7 理论最高支持 16x16 并发流。 - 硬件落地现实：主流手机与笔记本受限于体积与功耗通常维持 2x2，昂贵的高端路由器常见配置也仅到 4x4。 ### [12:34](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=754s) 谨防天线营销陷阱与保守版 Wi-Fi 7 的价值针对路由器外观上的消费误区，Linus 特别提醒大家切勿被路由器的外观参数迷惑，因为外置天线数量和实际的 MIMO 性能根本没有必然联系。Linus 明确警告道：“Don't let manufacturers fool you. More antennae sticking out of a device does not automatically equal better MIMO performance.”（别被厂商给骗了，设备上插着的天线多，绝不自动等同于更好的 MIMO 性能）。虽然 2x2 确实需要两根天线、4x4 需要四根天线，但仅仅因为一台路由器外表伸出 8 根天线（例如华硕 ROG Rapture AX11000 那样），并不代表它真能跑满 Wi-Fi 6E 的 8x8 极限规格，这套物理外观把戏在 Wi-Fi 7 时代依然会屡见不鲜。能够真正榨干 Wi-Fi 7 极限性能的设备即便真的面世，也必定凤毛麟角且售价不菲。但 Linus 强调，普通消费者完全不用为此气馁，因为即使是极具性价比的保守版 Wi-Fi 7 方案，也能带来切切实实的网络飞跃。哪怕你买的平价入门路由器只支持两频段 MLO（多链路聚合操作），那也意味着能同时调用的频段数量比起以往整整翻了一倍，日常体验的升级依然十分显著。 - 天线误区：天线数量往往是多频段分流或波束成形设计，外插 8 根天线并不等于具备 8x8 MU-MIMO。 - 体验下限高：即便入门级路由器仅实装双频 MLO，依然能让客户端同时并发使用两个无线频段，直接突破以往单频段连接的瓶颈。 ### [13:27](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=807s) 赞助商转场与结尾花絮在熟练地借着“频段翻倍”双重双关引出赞助商 Wicked Cushions 耳机替换耳垫（主打双层冷却凝胶告别耳部蒸桑拿、记忆海绵与透气运动面料，附带 365 天无忧保障）之后，Linus 在视频片尾推荐观众去看 ShortCircuit 频道的另一期评测。在那期节目中，他开箱了亚马逊旗下 Eero 的 Wi-Fi 7 路由器。Linus 吐槽那玩意儿价格贵得离谱，而且部分核心功能居然恶心人地锁在付费订阅墙后面，但他也不得不承认其实际速度确实强得不可思议——即便仅用一台普通智能手机连接，实测无线握手与传输速率竟然也能飙到 2.5 Gbps 以上，展现了 Wi-Fi 7 真正落地的强悍实力。 ## Notable quotes > [12:34](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=754s) Don't let manufacturers fool you. More antennae sticking out of a device does not automatically equal better MIMO performance. — Linus Sebastian > [11:37](https://www.youtube.com/watch?v=GqCwLjhb4YY&t=697s) Wi-Fi 6 supported 8x8, but even a $600 Wi-Fi 6E router like Netgear's Nighthawk only implemented 4x4, and your laptop or smartphone is probably only 2x2. — Linus Sebastian ## Mentioned Linus Sebastian, Intel, Thunderbolt 4, Thunderbolt 5, Thunderbolt Share, USB-IF, Wi-Fi 7, Core Ultra 7 155H, Steam Link, Moonlight, Eero Mesh Wi-Fi 7, Netgear Nighthawk, ASUS ROG Rapture AX11000, ShortCircuit, Wicked Cushions, 使命召唤：现代战争 ## Video description Check out Wicked Cushions at Amazon: https://lmg.gg/WickedCushionsAmazon or their website: https://wickedcushions.com/LTT Intel showed us the goods on probably the single thing I'm most excited about from CES 2024, and it's...Thunderbolt? Seriously, this is so cool. Not only that, but Wi-Fi 7 seems like it will actually be an increase that actually matters to regular people- as well as niche corner-case weridos like me lol Discuss on the forum: https://linustechtips.com/topic/1552428-this-was-under-nda-until-i-called-in-a-favor-intel-ces-2024/ ► GET MERCH: https://lttstore.com ► GET EXCLUSIVE CONTENT ON FLOATPLANE: https://lmg.gg/lttfloatplane ► SPONSORS, AFFILIATES, AND PARTNERS: https://lmg.gg/partners ► EQUIPMENT WE USE TO FILM LTT: https://lmg.gg/LTTEquipment ► OUR WAN PODCAST GEAR: https://lmg.gg/wanset FOLLOW US --------------------------------------------------- Twitter: https://twitter.com/linustech Facebook: http://www.facebook.com/LinusTech Instagram: https://www.instagram.com/linustech TikTok: https://www.tiktok.com/@linustech Twitch: https://www.twitch.tv/linustech MUSIC CREDIT --------------------------------------------------- Intro: Laszlo - Supernova Video Link: https://www.youtube.com/watch?v=PKfxmFU3lWY iTunes Download Link: https://itunes.apple.com/us/album/supernova/id936805712 Artist Link: https://soundcloud.com/laszlomusic Outro: Approaching Nirvana - Sugar High Video Link: https://www.youtube.com/watch?v=ngsGBSCDwcI Listen on Spotify: http://spoti.fi/UxWkUw Artist Link: http://www.youtube.com/approachingnirvana Intro animation by MBarek Abdelwassaa https://www.instagram.com/mbarek_abdel/ Monitor And Keyboard by vadimmihalkevich / CC BY 4.0 https://geni.us/PgGWp Mechanical RGB Keyboard by BigBrotherECE / CC BY 4.0 https://geni.us/mj6pHk4 Mouse Gamer free Model By Oscar Creativo / CC BY 4.0 https://geni.us/Ps3XfE CHAPTERS --------------------------------------------------- 0:00 Intro 1:21 Thunderbolt Share 3:00 Latency 3:45 It's not encoded! 5:00 Why I'm excited 6:00 Wi-Fi 7 is legit. 9:00 "MLO. Is. Sick." 10:43 Preamble Puncturing 11:40 Spatial Streams 12:30 Don't get fooled again! 14:03 outro [Source](https://www.youtube.com/watch?v=GqCwLjhb4YY)

Source quotes

[12:34] Linus Sebastian: Don't let manufacturers fool you. More antennae sticking out of a device does not automatically equal better MIMO performance. [11:37] Linus Sebastian: Wi-Fi 6 supported 8x8, but even a $600 Wi-Fi 6E router like Netgear's Nighthawk only implemented 4x4, and your laptop or smartphone is probably only 2x2.

Source quotes

[12:35] Linus: More antennae sticking out of a device does not automatically equal better MIMO performance. [10:49] Linus: Preamble puncturing allows a Wi-Fi 6 or 7 device to recognize that the channel is partially occupied and to utilize all the unoccupied portions of that channel. [11:38] Linus: The Wi-Fi 7 spec now allows for up to 16 simultaneous clients worth of input and output. That is 16x16 MU-MIMO. But the reality of it is, most devices aren't going to implement that.

Original transcript

[0:00] Watch, I'm gonna take this Thunderbolt cable, [0:03] connect these two computers, [0:05] drag these giant files across, [0:07] and whoo-hoo-hoo, mama, 800 megabytes per second. [0:13] That is like installing the new Modern Warfare [0:16] in less than five minutes. [0:18] And I can practically hear your chair scraping the floor [0:20] as you jump up to go and try this, [0:22] but hold on a second. [0:24] You can't just plug two computers together [0:27] and transfer files between them like that. [0:30] And you especially can't run a game on one PC [0:34] and then play it completely losslessly on another. [0:38] And yet, here I am doing it. [0:40] What you guys are looking at is, in my opinion, [0:44] finally the answer to what exactly is it [0:46] that makes Thunderbolt special [0:48] now that Intel gave the spec to the USB-IF? [0:51] And if you're anything like me, [0:52] your head is absolutely swimming with questions right now. [0:56] How are they doing lossless encoding? [0:59] What is the latency like? [1:01] Do I really have to find out after the segue to our sponsor? [1:06] Headphone juice is not something you should be proud of. [1:09] These are your typical headphones [1:10] that accumulate heat while you game. [1:12] Ew. [1:13] Wicked Cushions add a cooling gel to their cushions [1:15] to keep your ears sweat-free. [1:17] You can easily replace them on your headset. [1:19] Check out Wicked Cushions at the link below. [1:21] What you were just looking at is called Thunderbolt Share, [1:24] and it's, in essence, a piece of software [1:27] that establishes a point-to-point network connection [1:29] between two Thunderbolt 4 or Thunderbolt 5 Windows PCs. [1:33] It features exceptionally high bandwidth [1:36] and hyper-low latency, which kind of makes sense, [1:39] since you may recall that Thunderbolt [1:41] is essentially a four-lane-wide PCIe connection [1:45] with extras like USB for peripherals, [1:47] DisplayPort for monitors, and power for one-cable docking. [1:51] Let's start by revisiting that file transfer demo [1:54] that I just showed you guys, [1:55] because it's A, the most obvious use case for this, [1:59] and B, the easiest to understand in terms of the data flow. [2:02] Our files get pulled off of the SSD of the first machine, [2:04] they get packed in two packets, [2:06] they flow through the Thunderbolt chip [2:07] through the Thunderbolt cable, [2:08] and then they get unpacked on the other side [2:10] and dumped onto the SSD. [2:12] Now, I'm still a strong advocate for networked [2:16] rather than point-to-point data management solutions, [2:18] but it's a pinch, guys. [2:20] You're at the hotel trying to back up some footage [2:23] to a colleague who's flying home in the morning, [2:25] or you're at a LAN [2:26] and your buddy needs to grab a game install. [2:28] This will allow data transfers at theoretical speeds [2:32] of up to 20 gigabit per second for Thunderbolt 4 [2:35] and 40 gigabit for Thunderbolt 5, [2:38] which would bring that modern warfare install [2:40] that I talked about before down to under a minute. [2:43] And other than high CPU usage during the transfer [2:46] and Thunderbolt's one-meter passive cable length limit, [2:49] there is no downside. [2:51] It's just ripping fast. [2:54] Now, let's push this button. [2:56] This is incredible. [2:59] Even with a 10 or 25 gigabit network connection, [3:03] I have never in my life seen a PC-to-PC game stream [3:08] with this little latency. [3:10] I mean, I obviously haven't measured it [3:12] with a high-speed camera yet. [3:13] I don't have one in my pocket, [3:14] but to the eye, I would describe the latency as none. [3:18] And that's because every other solution, [3:21] Steam Link, Moonlight, et cetera, [3:24] relies on the GPU to render each frame, [3:27] encode it with a significant loss of fidelity, [3:29] and then transfer it to where it needs to be, [3:31] where it is received, unpacked, [3:33] and then finally output to the display. [3:37] By contrast, this is not encoded at all. [3:42] And before you call BS, because believe me, [3:45] I was thinking the same thing, [3:47] have a look at this pudding because the proof is in it. [3:51] While mirroring, we find none of the visual anomalies [3:54] that I would normally associate [3:55] with even a very high-quality AV1 stream. [3:58] And task manager over here [4:00] is picking up some very serious bandwidth [4:03] that is coming over that virtual link, about four gigabit. [4:07] There is, of course, some optimization taking place here. [4:09] Intel created their own remote desktop protocol [4:12] that handles turning your display's output [4:14] into Thunderbolt packets, [4:16] and that allows them to reduce bandwidth [4:18] by updating only changed elements on the remote display, [4:22] but it does basically what it says on the tin. [4:26] We're a little cagey about the details, [4:28] but as far as I can tell, that high CPU usage, [4:31] oh, 25% on a Core Ultra 7 155H, [4:36] is just the cost of doing business [4:38] when you're receiving a data stream this wide, [4:40] and the GPU usage is the secret sauce [4:44] that makes this work so well. [4:47] Intel says they are going straight [4:50] from GPU frame buffer on this side [4:52] to GPU frame buffer on the other side. [4:54] No storage bottleneck, they aren't even dumping it into RAM. [4:58] The software is still in alpha [5:00] with no committed release date for now, [5:02] so we're just scratching the surface [5:04] of what a link like this could unlock in the future, [5:06] but I'm already extremely impressed by what I've seen [5:10] and imagining countless applications for this. [5:13] I mean, I showed you guys, it goes both ways, right? [5:15] We can see the screen from the one, [5:16] we can use the controller from this one on the other one. [5:18] I mean, get this built into Windows, [5:21] and you wouldn't need a silly dual-screen laptop, [5:25] no offense, no offense, dual-screen laptops are great, [5:28] just to share content with a colleague. [5:30] You just plug their computer into yours and boop, flawless. [5:34] Oh, here's another one. [5:35] This would turn every outdated all-in-one or laptop, [5:39] as long as it has Thunderbolt 4 and up, into a monitor. [5:42] Also, by the way, while you're at it, Intel, [5:44] didn't talk to them about this ahead of time, [5:46] please release the remote desktop application. [5:50] I know I'm an edge case, [5:51] but I would totally wire up my house for 25 gig fiber [5:55] if I could just have all my compute in a rack [5:57] and then use thin clients for everything. [5:58] Of course, wires are passe now anyway. [6:02] Let's move a couple of paces this way, [6:04] where Intel is showing off [6:06] their other next big thing for connectivity. [6:09] I already got my first taste of Wi-Fi 7 [6:11] when I unboxed an Eero Mesh Wi-Fi kit over on Short Circuit. [6:15] And being the cynic that I am, [6:18] I was blown away when for once it actually delivered [6:23] on the kind of generational performance upgrade [6:26] that they claimed. [6:27] Check this out. [6:28] These are real world transfer speeds at two, three, [6:33] even in excess of four gigabit per second. [6:37] I mean, that is faster than what 99% of people [6:41] are getting with a wire right now. [6:43] And this easily lives up to Wi-Fi 7's EHT [6:48] or extremely high throughput moniker. [6:51] Now, your first reaction is probably, why should I care? [6:54] I mean, no one's internet is that fast. [6:56] And okay, well, first of all, there are some people, [6:59] princes need internet too, but that is a valid point. [7:04] I mean, the services that you use online [7:07] are likely to be a bottleneck [7:08] before your Wi-Fi is at these kinds of speeds. [7:11] Except that Wi-Fi rarely performs at peak efficiency [7:14] due to range limitations. [7:17] It's not always gonna be like this. [7:19] Not to mention network congestion. [7:21] So you're gonna get some fraction of that maximum speed. [7:25] And given that it's a lot higher than it used to be, [7:28] it's gonna be a lot higher than it used to be. [7:30] Also, this is perhaps more important, [7:32] once everyone makes the switch, [7:34] non-streaming data transfers will be completed [7:38] so much faster that it will free up air time [7:40] for everyone else, which will help to aid with congestion. [7:43] That's just a small part of the justification [7:46] for the speed though. [7:47] This will enable Wi-Fi to be used [7:48] for completely new applications. [7:50] I don't know about you, but I am tired [7:52] of seeing 4K video touted as a use case [7:55] for the next gen Wi-Fi. [7:57] Didn't you just tell me the old one was good for that? [8:00] And I mean, 8K, okay, fair enough, [8:02] but seems to have pretty much completely petered out. [8:05] But a PC grade wireless VR headset [8:08] with nearly native video quality? [8:10] Okay, that is really exciting. [8:13] And there is so much more. [8:15] It's felt for a long time, like each generation of Wi-Fi [8:19] was all about support for more devices, [8:22] rather than making the one PC that I have [8:24] in my basement go faster. [8:27] Not this time. [8:28] One of the big keys to this is a carryover from Wi-Fi 6. [8:32] Reduced congestion, thanks to support [8:34] for the six gigahertz band. [8:36] Part of this is that it just hasn't gotten clogged up yet, [8:39] like the five gigahertz and especially 2.4 gigahertz bands, [8:43] but much more significant is the fact that six gigahertz [8:47] doubled the maximum channel width to 320 megahertz. [8:52] And there's the fact that Wi-Fi 7 [8:54] can actually utilize connections [8:56] on more than one band at the same time [8:59] using multi-link operation or MLO. [9:03] MLO is sick. [9:06] Depending on the radio configuration, [9:07] it can enable either higher overall throughput [9:11] as data can be broken up and transmitted [9:13] over different bands simultaneously, [9:15] or this is great for the gamers out there, [9:18] lower latency as low as one millisecond [9:21] by utilizing the first available band [9:23] instead of being forced to wait its turn [9:26] like it would on a single band. [9:27] Let's go gamers. [9:29] Another improvement for individual users [9:31] comes in the form of 4K QAM, [9:33] that's Quadrature Amplitude Modulation, [9:35] which is a method for translating digital packets [9:38] into analog signals by varying the amplitude [9:41] and phase of the radio signal. [9:43] Okay, that's a lot to digest, but in a nutshell, [9:46] instead of having a maximum of 1,024 values per symbol, [9:50] which is the smallest amount of data that can be transmitted, [9:53] we can have one of 4,096 values, [9:57] which results in about a 20% transmission increase [10:01] over Wi-Fi 6, all other things being equal. [10:04] And of course, there are some multi-client things as well. [10:07] Wi-Fi 6 added the ability for multiple clients [10:10] to transmit data to an access point simultaneously. [10:13] This is achieved through a technology called OFDMA, [10:16] which basically establishes multiple sub-carriers [10:20] within Wi-Fi channels. [10:21] These sub-carriers are grouped resource units, [10:24] which can each be assigned to a client device. [10:27] Well, Wi-Fi 7 expands on this by allowing a client [10:31] to be assigned multiple resource units. [10:34] So you're not leaving capacity that's just sitting there [10:37] unused when there aren't many clients connected. [10:39] Okay, this is a good thing for individuals as well. [10:42] Man, I love this stuff. [10:43] Speaking of connected clients, [10:45] preamble puncturing was an optional feature of Wi-Fi 6 [10:49] that is now enabled across the board [10:51] and allows modern devices to utilize channels [10:54] that are partly occupied by legacy users. [10:57] Let's say for example, an old Wi-Fi IP camera [11:01] is using 40 megahertz in the middle [11:03] of an otherwise unused 160 megahertz wide channel. [11:07] Well, preamble puncturing allows a Wi-Fi 6 or 7 device [11:11] to recognize that the channel is partially occupied [11:14] and to utilize all the unoccupied portions of that channel. [11:18] So in our example, that would be 120 megahertz of bandwidth [11:22] where under previous standards, [11:24] you'd be limited to the largest contiguous unused portion [11:28] of the channel, just 80. [11:30] That was done back then to avoid interference, [11:32] but it left 40 megahertz of bandwidth completely unused. [11:37] There's more. [11:38] Conversations between APs and client devices [11:41] are often called spatial streams. [11:43] And without boring you all completely to death, [11:46] modern APs can deal with more than one spatial stream [11:50] at a time. [11:50] Just how many they can deal with [11:52] is usually a spec on the box [11:53] with the ridiculous acronym of MU-MIMO. [11:57] That is multi-user, multiple input, multiple output. [12:01] And the values typically look like 2x2, 4x4, [12:04] or hey, maybe 8x8 if you're a real baller. [12:08] The Wi-Fi 7 spec now allows [12:10] for up to 16 simultaneous clients worth of input and output. [12:14] That is 16x16 MU-MIMO. [12:18] But the reality of it is, [12:19] most devices aren't going to implement that. [12:22] I mean, Wi-Fi 6 supported 8x8, [12:24] but even a $600 Wi-Fi 6E router like Netgear's Nighthawk [12:29] only implemented 4x4, [12:31] and your laptop or smartphone is probably only 2x2. [12:34] By the way, on that topic, [12:36] don't let manufacturers fool you. [12:39] More antennae sticking out of a device [12:42] does not automatically equal better MIMO performance. [12:46] While it is true that 2x2 requires two antennas [12:50] and 4x4 requires four, [12:53] just because it has eight sticking out of it, [12:55] like this ROG Rapture AX11000, [12:58] doesn't mean it can do the maximum 8x8 of Wi-Fi 6E, [13:02] and the same will go for Wi-Fi 7. [13:04] In fact, devices capable of max Wi-Fi 7 performance [13:08] are going to be rare if they exist at all, [13:10] and they're gonna come with a serious price tag. [13:13] But even conservative implementations of Wi-Fi 7 [13:16] are going to bring us significant improvements. [13:19] I mean, maybe your budget-friendly router [13:21] only implements two-band MLO, [13:23] but that is still double the bands [13:25] that you could use before. [13:27] And now here's double the segue, [13:29] double the segue to our sponsor. [13:31] Are your headphones cushions hot like a tiny steam room? [13:34] Stop steaming your ears like you're broccoli. [13:36] Wicked Cushions offer all sorts of cushion replacement [13:38] for your headphones. [13:39] And it's not just your regular cushions. [13:41] With added dual-layer cooling gel, [13:43] they ensure you keep a cool head [13:44] even your team is feeding. [13:46] Looking for a crack? [13:47] Wicked Cushion gives you a crack-free comfort. [13:51] They use top-notch materials [13:52] like breathable sport fabric and memory foam. [13:55] With over 25,000 reviews [13:57] and a 365 days to fall in love warranty, [13:59] don't listen to me, listen to your ears this time. [14:01] Check out Wicked Cushions at the link below. [14:03] If you guys enjoyed this video, [14:05] go check out that ShortCircuit [14:06] where I unboxed that Eero Wi-Fi 7 thing. [14:08] It's super expensive, [14:09] and some of the features are locked behind a paywall [14:11] that I'm kind of not that into, [14:13] but it's fast. [14:14] It's really fast. [14:15] Like even on a phone, [14:16] like more than two and a half gigabit, what?