Intel Messed Up - WAN Show June 24, 2022

Episode Summary

Executive Summary: The episode centers on several tech and internet-policy controversies: weak early reviews of Intel Arc GPUs, Google/YouTube’s uneven handling of HDR and product features, TikTok’s data-access concerns, and the risks of Amazon Alexa voice cloning. It also covers LTT Labs’ evolving audio methodology, major LTT Store promotions and launches, PCIe 7.0, and a wide-ranging merch-message Q&A touching smart home security, gaming, and hardware choices.

Main Topics: Intel Arc GPU Early Reviews (Priority: 5/5): The hosts react to poor early benchmarks for Intel’s Arc Alchemist A380, noting that real-game performance appears disappointing versus entry-level AMD/NVIDIA cards, though synthetic benchmarks look better. They also discuss Intel’s unusual entry-level-first launch strategy and its naming scheme. YouTube HDR and Google Product Friction (Priority: 5/5): They criticize YouTube/Google for prioritizing new features while basic functionality remains broken or inconsistent, using HDR brightness handling and broken mobile community/story features as examples. They also praise Apple’s color handling by comparison. TikTok Data Access and Digital Sovereignty (Priority: 5/5): BuzzFeed’s reporting on China-based employees accessing non-public U.S. TikTok user data leads to a broader discussion of cross-border social-media regulation, internet fragmentation, and national-security concerns around foreign software/services. Alexa Voice Cloning and Deepfake Risk (Priority: 5/5): The hosts express strong discomfort with Amazon’s ability to clone a voice from under a minute of audio, warning about consent, post-mortem use, social engineering, and the possibility of bypassing voice-based security or authentication. LTT Labs Audio Measurements Debate (Priority: 4/5): Crinical’s feedback on LTT’s headphone/testing approach sparks discussion about balancing rigorous objective data with practical user-facing explanations. They debate whether sound demos can still be useful if clearly labeled as relative or entertainment-focused. LTT Store, Merch, and Operational Updates (Priority: 4/5): They promote the 50% off stealth hoodie clearance, new processor and hard-drive hoodies, and two new notebook designs. They also discuss the business necessity of clearing inventory and giving customers gift cards if they recently bought discounted items. Hardware/Platform Notes: PCIe 7.0, Smart Home, and Devices (Priority: 3/5): The episode includes brief coverage of PCIe 7.0’s bandwidth leap, a discussion of smart-home security (Matter, Ubiquiti, port forwarding/VPNs), and personal device notes like foldable phones, Steam Deck/Aya Neo handhelds, and electric motorcycles.

Key Arguments: Intel Arc’s timing was unfortunate: launching during a GPU market correction meant the card needed to be strong, but early results still looked underwhelming in real games. Intel’s own reviewer guidance appears to corroborate poor gaming performance, making third-party criticism more credible. Launching an entry-level GPU first is unusual and makes it harder for Intel to define the generation with a flagship product. Google/YouTube often ships features and experiments while core functionality remains buggy or missing, which the hosts see as a misallocation of engineering effort. Platforms like TikTok, Facebook, and foreign-built software will face increasing data-localization and sovereignty pressures, fragmenting the internet over time. Voice cloning is now easy enough to be a practical social-engineering tool; authenticity systems are not ready for it. Objective audio measurements are valuable, but user-friendly demos and clear disclaimers may still help consumers compare products. LTT Labs should prioritize being excellent at specific tests rather than trying to cover everything superficially. Inventory clearance and aggressive discounting are sometimes necessary business moves, even if they upset recent buyers; customer support compensation can soften the impact.

Data Points: Intel Arc A380 relative performance: Comparable to GT 1650 Super / Radeon RX 6400 in many real-world tests - Early third-party reviews and Intel review guidance for the entry-level Arc GPU MyDrivers aggregate against GTX 1050 Ti: GTX 1050 Ti at about 77% of A380 performance - One cited benchmark comparison discussed on the show GTX 1650 D6 vs A380: About 10% faster than the A380 - Aggregate results mentioned during Arc discussion RX 6400 vs A380: About 10% faster than the A380 - Aggregate results mentioned during Arc discussion Naraka Bladepoint benchmark delta: A380 beat GTX 1650 Super by 2 FPS and RX 6400 by 6 FPS - One of the few internal Intel test cases where Arc performed better Nizhan benchmark result: 200 FPS across the board - Intel’s reviewer guide example where performance was unusually high AV1 support: Both decoding and encoding supported - Intel Arc feature highlighted as potentially important long-term PCIe 7.0 bandwidth: 512 GB/s - 16-lane bandwidth claim mentioned for the new standard PCIe 7.0 lane equivalence: One lane equals a full PCIe Gen 4 x16 slot - Illustrating the scale of the bandwidth increase PCIe 7.0 vs PCIe 5.0: 8x faster - Comparison stated during the PCI-SIG segment PCIe 7.0 vs PCIe 6.0: 2x faster - Comparison stated during the PCI-SIG segment Stealth hoodie stock reduction: 1800 units to 115 remaining after two days - Inventory clearance update during sponsor/merch segment Stealth hoodie sales volume: 1685 units moved - Based on the hosts’ running tally during the show Day-one stealth hoodie sales: 500 units - Initial surge after the discount announcement Later same-day sales: Another 330 units - Update given after the first sales burst Discount on stealth hoodie: 50% off - Clearance promotion for remaining stock Gusto promo: 3 months free - Sponsor offer for listeners Mac Weldon promo: 20% off first order - Sponsor offer and summer clothing push Nord Security promo: 1 month free on all products - Sponsor offer for listeners Alexa voice cloning sample length: Less than one minute of recorded audio - Feature described in discussion of Amazon’s new voice model

Pivotal Quotes: "Everything is seen in China." — TikTok trust and safety employee (quoted in BuzzFeed report): Used to underline the privacy concerns around ByteDance/TikTok data access "Why do you need a high-profile influencer to complain about something that is obviously broken?" — Linus: Frustration over Google/YouTube fixing obvious product issues only after public pressure "We are commercializing deep fakes of voices. That is literally what's happening." — Linus: Reaction to Amazon Alexa voice cloning and its consent/security implications

Implications: Expect more pressure on platforms to localize data, improve core product quality, and defend against deepfake-enabled fraud. For consumers, hardware and software choices are increasingly shaped by privacy, security, and long-term support concerns.