NVIDIA Fake Frames Scandal - WAN Show January 10, 2025

Episode Summary

Executive Summary: The episode centered on CES 2025, with extensive discussion of NVIDIA’s RTX 50-series “fake frames” controversy, Intel ARC B580’s compatibility/performance quirks on older platforms, and the unexpectedly important AMD Ryzen AI Max chip. The hosts also covered major CES hardware highlights, Meta’s moderation changes, Amazon TV ad behavior, and a long tangent about personal/creative ambitions like TV, boat, and plane projects, plus an upcoming Tonight Show appearance.

Main Topics: NVIDIA RTX 50-series and “fake frames” debate (Priority: 5/5): The hosts argued that DLSS/frame generation is useful but that NVIDIA’s marketing claim that the RTX 5070 matches RTX 4090 performance is misleading. They emphasized that generated frames are not equivalent to fully rendered frames, especially for latency, benchmarks, and user experience. How to evaluate modern GPU performance (Priority: 5/5): They discussed that testing 50-series cards will be unusually complex because performance depends heavily on DLSS mode, frame-gen level, monitor refresh, and whether the comparison is visual fidelity, latency, or frame rate. They urged viewers to read multiple reviews. AMD Ryzen AI Max as a standout CES product (Priority: 5/5): A major highlight was Ryzen AI Max/Max+, described as a breakthrough for integrated graphics and unified memory. The hosts felt it fulfilled AMD’s long-promised APU vision and could meaningfully disrupt thin-and-light laptops and handhelds. Intel ARC B580 on older platforms (Priority: 4/5): They reviewed reporting that the ARC B580 can underperform badly on older CPUs/chipsets despite meeting official requirements like Resizable BAR. The discussion focused on CPU overhead, the limits of “official support,” and how far back GPU vendors reasonably should support hardware. CES product highlights and oddities (Priority: 4/5): The crew shared favorite or memorable CES finds, including Stern’s pinball displays, the Sense Robot chess machine, an exoskeleton, Xreal One Pro glasses, Thunderbolt-based streaming/USB advances, and Thermaltake’s surprising immersion cooling comeback. Business/media, sponsorships, and process (Priority: 4/5): The hosts commented on NVIDIA’s openness with reviewers, the difficulty of testing early-access products, limited sponsor inventory for WAN Show, the value of good benchmarking tools, and the need to preserve access and openness rather than overreacting to marketing claims. Personal and future projects (Priority: 3/5): A long after-dark segment explored ambitious side projects: a techified boat, possibly a private plane, a tech RV, and even a possible hotel. The idea was framed as content-first, charterable luxury-tech experiments, though with many practical and financial caveats.

Key Arguments: Frame generation and upscaling are legitimate technologies, but they should not be counted as equivalent to native rendered performance when comparing GPUs. NVIDIA’s 5070-to-4090 claim is misleading because the result depends on AI-assisted frame generation rather than raw rendering horsepower. The best way to test modern GPUs is with multiple reviews and careful attention to settings, not a single benchmark or a simple FPS figure. Ryzen AI Max is more important than most people realize because it finally makes AMD’s APU/fusion vision real with unusually strong integrated graphics. Intel ARC B580’s performance drop on older CPUs suggests driver/CPU overhead matters and may require revised support expectations. Modern products and platforms are increasingly “system” products, meaning software, drivers, settings, and refresh-rate context matter as much as the chip itself. Creators and reviewers should preserve openness with vendors and avoid overreacting in ways that could make future access worse. The most compelling CES products were not always the flashiest; the standout items were the ones that materially changed what was possible in a device class.

Data Points: RTX 5070 announced price: $550 - Discussed during the NVIDIA RTX 50-series keynote criticism. DLSS multi-frame gen example FPS: 28 FPS native vs. 230+ FPS with DLSS features - NVIDIA’s Cyberpunk 2077 example showing the scale of frame generation effects. Frame generation ratio discussed: 3 fake frames for every 1 real frame - Explaining 4X multi-frame gen behavior. RTX gamers turning on DLSS: 80% - A NVIDIA claim used to argue that many users already accept AI-assisted rendering. ARC B580 testing platform: i5-9600K on Z390 with ReBAR - Hardware Canucks’ older-platform test setup. CPU generations affected by ARC support issues: 9th gen Intel / 300-series chipset - Intel does not officially support these older platforms for ARC, despite ReBAR. Ryzen AI Max graphics core count: Up to 40 graphics cores - Highlighted from AMD’s new integrated graphics platform. Ryzen AI Max CPU core count: Up to 16 CPU cores - Top-end configuration mentioned for the new platform. Thermaltake mini-ITX case radiator support: Up to 280mm AIO - One of Thermaltake’s upcoming cases shown at CES. Thermaltake GPU support: Up to 360mm GPU - TR100 mini case specifications. Thermaltake immersion cooling cases: IX300 Mini ITX and IX600 ATX - Thermaltake’s immersion-cooling announcement. Sense Robot price: $1,000 USD - The chess robot discussed as cool but expensive and slow. Exoskeleton lifting assistance: 30 kilograms - The CES exoskeleton demo’s claimed support capacity. Exoskeleton price: $9,000 USD - Used to argue that the device is too expensive for its practical limitations. Xreal One Pro display size claim: ~155-inch screen at 4 meters - The personal theater glasses were described as projecting a huge virtual screen. Wi-Fi/USB bandwidth mention: USB 4 version 2.0 / 80 Gbps class - Used in the discussion of future eGPU and single-cable workflows. Hot potato hoodie availability: 69 units - Limited-edition merch drop announced on the show. Booth scale remark: Some CES booths are multi-story and effectively disposable - Used to illustrate the scale and waste of major trade show builds. Private jet annual operating cost example: $1.7M to $2.6M per year - Discussed while joking about teching out a jet/charter business.

Pivotal Quotes: "“I called it a red flag in the intro.”" — Linus: Defending his early skepticism of NVIDIA’s RTX 50-series marketing claims. "“I’m not buying a $2,000 GPU to turn on DLSS.”" — Linus: Explaining why AI-assisted performance can be useful but is not how he personally evaluates premium GPUs. "“This is what AMD fusion was supposed to be all those years ago.”" — Linus: Describing Ryzen AI Max as the real fulfillment of AMD’s long-standing APU vision.

Implications: Expect GPU reviews to become more nuanced and harder to compare, with AI-assisted rendering, support matrices, and platform age all mattering more. Ryzen AI Max could shift expectations for integrated graphics, while vendor openness with reviewers remains crucial.