Morning Brief
- NVIDIA unveils DLSS 4.5 Ray Reconstruction technology
- NVIDIA introduces RTX Spark PC processors and the Surface Laptop Ultra
- ASUS announces multiple new products
- Intel reveals Xeon 6+ server processors and other new offerings
- AMD launches new processors and graphics cards at Computex 2026
- Huawei unveils the nova 16 series and several other products
- The next-generation AV2 video codec standard is officially released
- MiniMax launches the M3 model
- Sennheiser begins sales of the MOMENTUM 5 over-ear headphones
- iFLYTEK introduces the Fika smartphone-shaped E Ink reader
- Bambu Lab launches the A2L 3D printer
- Dell announces the new XPS 13
- News Worth a Quick Look
NVIDIA unveils DLSS 4.5 Ray Reconstruction technology
On June 1, NVIDIA announced DLSS 4.5 Ray Reconstruction technology. The new technology will officially launch this August and will be available on all GeForce RTX GPUs, with support already confirmed for 38 games. DLSS 4.5 Ray Reconstruction replaces traditional hand-tuned denoisers with neural rendering, improving image quality in ray-traced and path-traced scenes. By integrating denoising and super-resolution into a single model trained on a significantly larger dataset, it can generate higher-quality pixels between sampled rays. Compared to previous versions, the DLSS 4.5 Ray Reconstruction model delivers 35% more computational capability and increases parameter count by 20%. Relative to DLSS 4.5 Super Resolution, the new model features deeper spatial awareness in every scene and makes more intelligent use of game engine pixel sampling and motion data, resulting in improved lighting accuracy, temporal stability, and motion clarity. The model is also accompanied by a dedicated developer masking tool, allowing developers to fine-tune parameters more precisely and further optimize image quality. Source
In addition, NVIDIA unveiled new RTX 50 Series graphics cards from partners including ASUS, Gigabyte, MSI, and PNY, alongside two RTX 50-powered laptops—the Acer Nitro 16 and Mechrevo Yaoshi 16 Air—as well as multiple new monitors, desktops, and other products. Source
NVIDIA introduces RTX Spark PC processors and the Surface Laptop Ultra
During its Computex 2026 keynote on June 1, NVIDIA announced the RTX Spark PC processor. The chip combines a 20-core Grace CPU developed in collaboration with MediaTek and an NVIDIA Blackwell RTX GPU featuring 6,144 CUDA cores, delivering up to 1 petaFLOP of AI performance. Manufactured using TSMC’s 3nm process, it includes 128GB of unified memory. RTX Spark is designed to enable creators, AI developers, and gamers to render massive 3D scenes exceeding 90GB, edit 12K 4:2:2 video, generate 4K AI videos, run local AI agents powered by 120-billion-parameter large language models with context windows of up to one million tokens, and play AAA games at over 100 FPS at 1440p resolution. Adobe is currently redesigning the architecture of Photoshop and Premiere to support RTX Spark, with the goal of delivering up to twice the AI and graphics performance. RTX Spark laptops feature an ultra-thin design measuring just 14mm thick and weighing only 3 pounds. They will be available in 14-inch and 16-inch sizes, with precision-machined aluminum chassis, OLED displays, and NVIDIA G-SYNC support. The first RTX Spark-powered products will include premium thin-and-light Windows laptops offering all-day battery life and high-end displays, as well as compact desktop systems. Products from ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI are expected to launch this fall, with Acer and Gigabyte models arriving later. Source
NVIDIA and Microsoft also jointly announced the Surface Laptop Ultra, powered by the RTX Spark PC processor. The device will be available in Platinum and Nightfall color options and features a 15-inch mini-LED PixelSense Ultra touchscreen with up to 2,000 nits of peak HDR brightness and a pixel density of 262 PPI. It will also feature the largest trackpad ever included in a Surface device, along with HDMI, USB-C, USB-A, SD card, and headphone ports. For power efficiency, Microsoft and NVIDIA have collaborated to implement the Microsoft Power and Thermal Framework on RTX Spark systems, maximizing performance while reducing power consumption during mobile use. This enables RTX Spark PCs to remain cool while delivering industry-leading efficiency. Microsoft is also enhancing Windows support for unified memory architectures, increasing the amount of system memory accessible to the GPU, allowing larger local AI models and more complex creative workloads to be loaded. The Surface Laptop Ultra is scheduled to launch later this year. Source
ASUS announces multiple new products
On June 1, ASUS unveiled a wide range of new products at Computex 2026.
Leading the announcements was the next-generation ProArt lineup of AI-focused creator PCs, including the ProArt P16 and P14 laptops as well as a new ProArt mini PC, all powered by NVIDIA RTX Spark processors. Both the ProArt P16 and P14 feature ASUS Lumina Pro OLED displays, while the ProArt mini PC delivers 140W of cooling performance within a compact 150 × 150 × 51 mm chassis and supports M.2 PCIe Gen 5 x4 expansion. Source
On the display side, ASUS and ROG jointly introduced more than ten new monitors. Updates include the ROG Swift 4K Dual-Layer OLED monitor, ROG Strix OLED gaming monitors, ROG Strix 5K gaming monitors, ProArt professional creator displays, TUF Gaming Series 5 monitors, and ASUS ZenScreen products. ASUS also announced the ZenScreen Duo MB14FCD dual-screen portable monitor, which combines two 14-inch 16:10 FHD IPS panels into a display area equivalent to approximately 20 inches. Another new product is the 13.3-inch ZenScreen Color ePaper MP13UC color E Ink display. Source
ROG also unveiled the ROG XBOX Ally X20 handheld gaming device. Finished in a black-and-gold color scheme, it features a 7.4-inch 120Hz ROG Nebula HDR OLED display with up to 1,400 nits of peak brightness. It is powered by the AMD Ryzen AI Z2 Extreme processor, paired with 24GB of LPDDR5X memory and 1TB of storage, and supports Auto SR image-enhancement technology. The device incorporates an Xbox-style eight-direction D-pad and TMR analog sticks, along with a native Xbox mode. The package also includes the ROG XREAL R1 (20th Anniversary Edition) gaming AR glasses. When connected to the handheld, the glasses can project a virtual 171-inch display at a viewing distance of four meters with a 240Hz refresh rate and support native 3DoF head tracking. Source
To celebrate the 20th anniversary of the ROG brand, ASUS also introduced the ROG Edition 20 Series featuring the same black-and-gold design language. The lineup includes the ROG Crosshair X870E Edition 20 motherboard, ROG Crosshair X870E Edition 20 graphics card, ROG Thor 3000W Titanium III Edition 20 power supply, ROG GR20 Edition 20 open-frame chassis, ROG NUC 16 Edition 20, ROG G1000 Edition 20 desktop PC, ROG Swift OLED PG27AQWP-G Edition 20 monitor, and the ROG Rapture GT-BE98 Pro Edition 20 router, alongside various accessories such as mouse pads, keycaps, mice, keyboards, gaming chairs, and backpacks. Source
Intel reveals Xeon 6+ server processors and other new offerings
At Computex 2026 on June 1, Intel unveiled the Xeon 6+ server processor family. Designed for cloud-native workloads and 5G core networks, the Xeon 6+ adopts the E-core architecture built on Intel’s 18A process technology. In terms of internal chip structure, each Clearwater Forest compute tile consists of six modules, with each module integrating four Darkmont E-cores, resulting in 24 E-cores per compute tile. By stacking 12 such compute tiles, Intel has created a flagship model with up to 288 cores, delivering extremely high core density and parallel processing capability. Xeon 6+ features up to 576MB of “enhanced low-latency” LLC cache and supports 12-channel DDR5 memory at speeds of up to 8000 MT/s. In terms of performance, Intel claims the flagship Xeon 6990E+ delivers up to 30% higher average per-thread performance than the AMD EPYC 9965. Under typical workloads, the Xeon 6990E+ can also provide up to 30% better performance per watt compared with similarly positioned EPYC processors. Compared to Intel’s own previous-generation Xeon 6780E, the company expects approximately 55% higher efficiency and an overall performance increase of around 126% in comparable scenarios. Source
Intel also previewed its next-generation Xeon processor, codenamed Diamond Rapids, scheduled for release in 2027. Diamond Rapids continues Intel’s large-die, modular design approach and adopts the enhanced Intel 18A-P process technology. It features a scalable SoC architecture while maintaining uniform memory latency. Core counts will increase by 50%, with strong single-threaded performance optimized for high-end IaaS workloads. The processor will deliver double the memory bandwidth through expanded channel counts and higher supported frequencies, while also adding support for PCIe Gen 6. Source
In addition, Intel introduced the next-generation Crescent Island GPU platform for AI data centers, emphasizing large memory capacity and energy-efficient inference performance. Crescent Island is based on Intel’s Arc Xe 3P architecture, which is also used in the integrated graphics of the current Panther Lake platform. It represents Intel’s latest high-performance graphics and compute platform for data centers. According to Intel, it is one of the company’s most powerful data center GPUs to date, offering up to 480GB of onboard memory per card, clearly targeting AI and high-performance computing (HPC) workloads that require large-scale models and datasets. For cooling, Crescent Island uses an air-cooled design and is rated for a 350W TDP. Intel says the card is designed to meet the demands of next-generation AI workloads and supports a wide range of data formats and low-precision quantization types, from native FP4 and MXFP4 to FP64, enabling both training and inference across different accuracy and performance requirements. Crescent Island is specifically optimized for AI inference while delivering higher performance density and lower overall data center operating costs. Source
AMD launches new processors and graphics cards at Computex 2026
At Computex 2026 on June 1, AMD announced several new products. The company first introduced two new 3D V-Cache gaming processors for desktop platforms: the Ryzen 7 5800X3D 10th Anniversary Edition for the AM4 socket and the Ryzen 7 7700X3D for the AM5 platform.
The Ryzen 7 7700X3D belongs to the Zen 4-based Ryzen 7000 series and is an 8-core, 16-thread desktop processor positioned below the highly popular Ryzen 7 7800X3D. In terms of specifications, the two processors share the same core count, cache configuration, and 3D V-Cache stacking design, with the primary differences being clock speeds. The 7800X3D features a 4.2GHz base clock and boosts up to 5.0GHz, while the 7700X3D runs at a 4.0GHz base clock with a maximum boost frequency of 4.5GHz. Both processors carry a default TDP of 120W. The 7700X3D retains 96MB of L3 cache, along with 512KB of L1 cache and 8MB of L2 cache, and uses TSMC’s 5nm FinFET process for the CPU cores alongside a 6nm I/O die. Memory support includes dual-channel DDR5 with capacities up to 128GB. AMD officially rates support at up to DDR5-5200 with two single-rank or two dual-rank DIMMs, and DDR5-3600 with four single-rank or dual-rank DIMMs installed.
The Ryzen 7 5800X3D 10th Anniversary Edition is AMD’s tribute to the 10th anniversary of the AM4 socket. Over the years, AMD has continued extending the lifespan of the AM4 platform with products such as the Ryzen 7 5700X3D and Ryzen 5 5500X3D, both featuring 3D V-Cache technology. The newly unveiled Ryzen 7 5800X3D 10th Anniversary Edition retains identical specifications to the original version, including its 8-core, 16-thread design, cache capacity, and clock speeds, meaning performance is expected to be unchanged.
The Ryzen 7 7700X3D will retail for $329 and launch on July 16, while the AM4-based Ryzen 7 5800X3D 10th Anniversary Edition is priced at $349 and will become available on June 25.
AMD also introduced EXPO ULL (Ultra Low Latency), a new memory overclocking technology. Built upon the existing EXPO memory profile overclocking system, EXPO ULL aims to further reduce DDR5 memory timings to deliver higher frame rates and smoother gaming performance on Ryzen platforms. The technology primarily focuses on lowering DDR5 CAS latency values. According to AMD’s demonstration, compared to standard JEDEC-spec DDR5-5600 CL40 memory, a DDR5-6000 CL28 EXPO ULL configuration can improve average gaming frame rates by approximately 13%, representing about four percentage points more performance than a conventional EXPO configuration on the same platform.
AMD also officially announced that it will extend the lifespan of its current mainstream desktop socket, AM5, through 2029. This means the platform is expected to support at least one more generation of Zen processors. While AMD did not explicitly name future products, industry observers widely interpret the announcement as confirmation that AM5 will support Zen 6 and related Ryzen 10000-series processors.
Finally, AMD unveiled the Radeon RX 9070 GRE graphics card. Positioned in the mid-range segment, the RX 9070 GRE features 48 Compute Units, 3,072 stream processors, 48 third-generation ray accelerators, and 96 second-generation AI accelerators. The card operates at a game clock of 2070MHz and can boost up to 2790MHz. Memory specifications include 12GB of GDDR6 on a 256-bit memory bus running at up to 20Gbps, providing 640GB/s of bandwidth, along with 48MB of AMD Infinity Cache. The card has a typical board power rating of 220W, requires a recommended 650W power supply, and uses dual 8-pin power connectors. It also features 96 ROPs and 192 texture units. Compared with the RX 9060 XT’s 128-bit memory interface, the RX 9070 GRE doubles memory bus width, although its 12GB memory capacity is lower than the 16GB found on the RX 9070 and RX 9070 XT. The Radeon RX 9070 GRE is priced at $549. Source
Huawei unveils the nova 16 series and several other products
On June 1, Huawei held its nova 16 Series and All-Scenario New Product Launch Event. Source
The nova 16 series includes the nova 16 Ultra, nova 16 Pro, and nova 16, all powered by the Kirin 9010S processor. Both the nova 16 Pro and Ultra feature 6.84-inch displays and a quad-camera rear setup. The Pro model includes a 200MP RYYB ultra-high-resolution large-sensor camera (F1.8 aperture, OIS), a 50MP RYYB periscope telephoto camera (F2.6 aperture, OIS), a 50MP ultra-wide macro camera (F2.2 aperture), and a Red Maple Color camera. The Ultra model upgrades the telephoto lens aperture to F2.2. On the front, both models feature a 50MP camera paired with a Red Maple Color camera, with an F2.4 aperture on the Pro and F2.2 on the Ultra. The nova 16 Pro offers IP65 dust and water resistance, a 7000mAh battery, and starts at RMB 3,899 (12GB+256GB). It is available in Sky Blue, Horizon White, Starry Black, and Iridescent Pearl. The nova 16 Ultra features IP68 and IP69 dust and water resistance, a 7000mAh battery, and starts at RMB 4,699 (12GB+256GB). Color options include Sky Blue, Horizon White, and Starry Black.
The nova 16 features a 6.68-inch display and a triple-camera rear setup consisting of a 50MP ultra-high-resolution camera (F1.9 aperture), a 50MP RYYB periscope telephoto camera (F2.6 aperture, OIS), and a Red Maple Color camera. The front camera does not include a Red Maple Color sensor. It is equipped with a 7000mAh battery, supports IP65 dust and water resistance, and starts at RMB 2,999 (12GB+256GB).
Huawei also introduced the nova 16z, powered by the Kirin 8020 processor. It features a 6.7-inch display and a rear camera system consisting of a 50MP ultra-high-resolution camera, a 12MP RYYB telephoto portrait camera with OIS, and a Red Maple Color camera. The device is equipped with a 6000mAh battery, supports IP65 dust and water resistance, and starts at RMB 2,699 (12GB+256GB). Available colors include Horizon White, Starry Black, and Iridescent Pearl.
The HUAWEI MatePad Pro Max is available in Glow Blue, Moonlight Silver, Deep Space Gray, and Obsidian Gray. It features a 13.2-inch display, with an optional Soft Light Edition equipped with a flexible OLED CloudClear Soft Light display. The tablet is powered by the Kirin T93 Pro processor, while the Enjoy Edition uses the Kirin T93. It packs a 10,400mAh battery and supports up to 40W wired reverse SuperCharge. The device weighs 499g and is only 4.7mm thick. The Enjoy Edition starts at RMB 5,999 (12GB+256GB), the standard Wi-Fi version starts at RMB 6,199 (12GB+256GB), and the Soft Light Edition starts at RMB 7,399 (12GB+256GB).
Huawei also announced the HUAWEI FreeClip 2 Collector’s Edition clip-on earbuds, the Huawei Lingxiao Q7 Powerline Mesh Router, the HUAWEI Watch GT Runner 2 Track Legend Edition, and the Huawei Nova Watch X1 and X1 Pro.
The next-generation AV2 video codec standard is officially released
The Alliance for Open Media (AOMedia) officially released version 1.0 of the next-generation AV2 video coding standard on May 29. Like AV1, AV2 continues to follow an open-source model centered on being efficient, royalty-free, and open. According to AOMedia, AV2 builds upon AV1 with further optimizations, enabling higher-quality video delivery at lower bitrates for evolving use cases such as streaming, broadcasting, and real-time video conferencing. It also enhances support for AR and VR applications, improves split-screen multi-program transmission and screen-content coding, and operates effectively across a broader range of visual quality targets. In terms of compression efficiency, official testing shows that AV2 reduces bitrate by approximately 28.63% compared to AV1 under the PSNR-YUV metric, while achieving a reduction of 32.59% under the VMAF metric, with virtually no loss in image quality. It is worth noting that AV2’s reference software, AVM (AOMedia Video Model), currently runs very slowly, with encoding speeds on mainstream hardware typically reaching only around one frame per second, meaning practical deployment is still a long way off. Source
MiniMax launches the M3 model
MiniMax officially launched its next-generation model, MiniMax M3, on June 1. The model combines state-of-the-art coding capabilities, up to 1 million tokens of context length, and native multimodal support, including image input, video input, and desktop computer operation. It is the first model in China to offer all three capabilities simultaneously and currently the only open-source model to do so. According to official benchmarks, M3 achieved a score of 59.0% on the SWE-Bench Pro programming benchmark, surpassing GPT-5.5 and Gemini 3.1 Pro while approaching Opus 4.7. It also achieved the highest score on the Claw-Eval agent benchmark and outperformed Gemini 3.1 Pro on the multimodal benchmark OmniDocBench. M3 is built on a new sparse attention architecture called MSA (MiniMax Sparse Attention), reducing per-token computation at a 1-million-token context length to just one-twentieth of the previous generation. The architecture delivers more than 9× acceleration during the prefilling stage and over 15× acceleration during decoding. Alongside the model launch, MiniMax updated its Agent product, MiniMax Code, and introduced a Token Plan subscription service, with Plus priced at RMB 49/month, Max at RMB 119/month, and Ultra at RMB 469/month. The M3 API is available immediately, while model weights and the technical report will be open-sourced within 10 days. Source
Sennheiser begins sales of the MOMENTUM 5 over-ear headphones
The Sennheiser MOMENTUM 5 over-ear headphones went on sale on May 31 with the slogan “Lossless Audio, Even Wirelessly,” priced at RMB 3,299. The new model features Sennheiser’s self-developed 42mm dynamic driver units. It supports major audio codecs including SBC, AAC, aptX, aptX HD, and aptX Adaptive, while also incorporating Snapdragon Sound TT technology and aptX Lossless support. This enables CD-quality lossless audio transmission (16-bit/44.1kHz) over Bluetooth. The headphones are Hi-Res Audio certified and support Dolby Atmos spatial audio. For noise cancellation, the number of microphones has doubled to eight. With active noise cancellation enabled, battery life can reach up to 57 hours, while a 10-minute charge provides up to 7 hours of playback. Source
iFLYTEK introduces the Fika smartphone-shaped E Ink reader
iFLYTEK introduced the Fika e-book reader on June 1. Designed in a smartphone-like form factor, the device is priced at RMB 2,399. It is available in “Aussie White” and “Iced Americano” color options, weighs 140g, and measures 6.7mm thick. The reader features a 6.13-inch 300 PPI Carta 1300 E Ink display, paired with a 32-level dual-tone front light system and the IFR 2.0 fast-refresh algorithm. It is powered by an octa-core MediaTek processor, with 6GB of RAM and 128GB of storage. Buyers receive 2GB of mobile data per month free during the first year. The device includes a 2,580mAh battery and offers up to 21 days of standby time. Source
Bambu Lab launches the A2L 3D printer
On June 1, Bambu Lab introduced the Bambu A2L 3D printer. The A2L features a single-nozzle design and supports a maximum build volume of 330 mm × 320 mm × 325 mm, representing a 105% increase compared to the A1. It incorporates an adaptive vibration compensation algorithm and built-in granular dampers within the frame to reduce print artifacts caused by vibrations. The printer supports quick switching between a blade-cutting module and a brush module. It also features active motor noise reduction, allowing operating noise to drop as low as 49 dB in Silent Mode. The standalone Bambu A2L is priced at RMB 2,549 and is eligible for government subsidies, with limited-time launch promotions available. Source
Dell announces the new XPS 13
On May 31, Dell unveiled the new XPS 13 laptop. Positioned against the MacBook Neo, the new XPS 13 starts at $699, weighs 1kg, and measures 12.7mm thick. It is powered by the Intel Core 5 320 processor, with Core Ultra 7 355 variants planned for a later release. Memory configurations include LPDDR5x-7467 MT/s, with 8GB to 16GB single-channel options at launch and future higher-end models offering 16GB to 32GB dual-channel configurations. The laptop features a 13.4-inch InfinityEdge touchscreen display. It supports Wi-Fi 7 and Windows Hello, and comes with a backlit keyboard. The device will go on sale in June. Source
News Worth a Quick Look
- According to 36Kr, Doubao is expected to officially launch a paid subscription service in late June. In the third quarter, the platform is also expected to further integrate with Douyin E-commerce features and expand its monetization scenarios. Source
- Leaker Ice Universe posted what appears to be a mock-up image of the rumored iPhone Ultra/Fold on Weibo. Source
- Microsoft has announced that Fable has been delayed and is now scheduled for release in February 2027. Source
- On June 1, iReader and WIKO respectively introduced AI-powered collectible companion devices called CreMoMo and Xingzai. Xingzai is powered by Huawei’s Xiaoyi large language model, while CreMoMo focuses on a growth-oriented, personality-driven AI companion experience. Source 1 2
- Randy Pitchford, founder of Gearbox Software, the studio behind the Borderlands series, posted on X that a friend of his found what appears to be an unreleased Google Pixel Watch 5 while diving near Saint Martin, recovering it from the ocean floor. Source
