SK Hynix is flexing some future memory tech you will not be able to afford. The company showed off its next-level HBM4 with a 16-layer stack packing 48 gigabytes, claiming it as an industry first. This follows their existing 12-layer HBM3E product that is actually hitting the market soon. They also displayed a new server memory module called SOCAMM2, designed for better power efficiency in AI systems. For regular devices, they are working on LPDDR6 memory, which promises faster speeds and lower power use for on-device AI tasks.
Their NAND flash presentation included a 321-layer quad-level cell product aimed at massive data center solid-state drives. This thing holds two terabits per die, boasting better power and performance than older...
MSI built another tiny brick that yells about AI. The company showed off its Cubi NUC AI Plus 3MG mini PC, a half-liter box packing Intel's new Panther Lake processors, also called Core Ultra Series 3. This thing is a certified Copilot Plus PC thanks to a neural processor claiming fifty AI TOPS performance. It will come in three configurations, with a budget Core Ultra 5 model, a mid-tier Core Ultra 7 chip, and a flagship Core Ultra 9 386H variant sporting sixteen cores and integrated graphics.
The mini PC supports up to 32 gigabytes of memory across two slots and has one NVMe drive bay for storage. Connectivity is a highlight, featuring Wi-Fi 7, two Ethernet ports, and dual Thunderbolt 4 outputs that can run four monitors at once. A...
Gigabyte slapped a fancy name on slightly brighter OLED monitors. The company announced four new models featuring their HyperNits and AI Picture tech. The lineup includes a 34-inch ultra-wide, a 32-inch 4K screen, and two 27-inch QHD panels. The flagship MO34WQC36 is a curved display with a 360Hz refresh rate and WQHD resolution, using a Samsung sub-pixel layout. The 32-inch MO32U24 offers 4K at 240Hz and includes a special film to reduce the purple tint common on some OLED panels.
The two 27-inch options are the MO27Q28GR, which uses a glossy white OLED for deeper blacks, and a white-colored model called the MO27Q2A ICE. Both have a 280Hz refresh rate. The new HyperNits technology can boost peak brightness by up to thirty percent in...
Alienware just vomited out like eight new laptop models for next year. The company announced refreshes for its mainstream Aurora and flagship Area-51 lines, plus teased some cheaper and thinner future designs. The Alienware 16X Aurora is the new mid-range option, using Intel's Core Ultra 200HX series processors and Nvidia RTX 5060 or 5070 graphics. It features a 16-inch WQXGA OLED screen with a high refresh rate and good color coverage. The laptop supports up to 64 gigabytes of memory and has dual SSD slots. It uses a copper alloy cooling system and will be fairly heavy.
The flagship Area-51 series gets updates in both 18-inch and 16-inch sizes. These models use a new design language for the brand's anniversary and feature improved...
Alienware just shoved a fancy new CPU into the same overpriced box. The company updated its Area-51 desktop with AMD's Ryzen 7 9850X3D processor, which hits higher clock speeds and packs a massive cache for gaming. This pre-built system can be configured with top-tier parts like an Nvidia RTX 5090 graphics card, up to 64 gigabytes of DDR5 memory, and multiple terabytes of Gen5 SSD storage. The motherboard uses the X870E chipset and has decent expansion slots. Alienware claims the new chassis runs much cooler and quieter than before, supporting powerful cooling solutions and huge graphics cards. The power supply can be configured up to 1500 watts to handle the load. Connectivity includes the latest wireless standards and plenty of USB...
NVIDIA keeps juicing its little AI box with software updates. The DGX Spark, a mini PC for AI development, has gotten faster since launch through optimization. The company claims a two-and-a-half times performance increase for some large language models when using two units together. Other AI models for image generation also see significant speed bumps. New software playbooks have been added to help developers set up various AI tasks, including robotics simulations and local model inference.
Practical use cases show the device acting as a separate accelerator for other computers. One demo had it paired with a MacBook Pro, cutting AI video generation time from eight minutes down to one. Another example is offloading texture creation for...
AMD made a little box to try to one-up Nvidia's dev kit. The company revealed its own mini PC called the Ryzen AI Halo, designed as a competitor to Nvidia's DGX Spark for AI development work. This compact system uses AMD's highest-end Ryzen AI MAX processors, codenamed Strix Halo. It will support the full ROCm software suite and come pre-optimized for popular AI tools and models. The mini PC is meant to accelerate AI workflows for developers.
Internally, it packs up to sixteen CPU cores and a powerful integrated GPU with forty compute units. Memory support goes up to 128 gigabytes of fast LPDDR5x. The system uses a dual-fan cooler to handle the thermal load from these chips. AMD did not announce a price, positioning it as a more...
AMD's new laptop chips are just a mild refresh with a bigger number. The company launched its Ryzen AI 400 series, codenamed Gorgon Point, which is basically an optimized version of last year's Strix Point design. They squeezed out minor gains, like a ten to twelve percent performance bump, by tweaking clock speeds and supporting faster memory. The top model, the Ryzen AI 9 HX 475, still uses twelve Zen 5 cores and an RDNA 3.5 graphics unit, but now boosts up to 5.2 gigahertz. The integrated GPU can also hit higher clock speeds, and the NPU gets a small uplift to 60 AI TOPS.
The lineup includes seven different processors, with core counts ranging from four to twelve. Support for faster LPDDR5x memory is a noted upgrade. AMD claims...
AMD is really trying to flex on Nvidia with some monster AI chips. The company showed its first 2-nanometer processors, an EPYC server CPU called Venice, and an Instinct MI455X accelerator, built for a liquid-cooled system named the Helios AI rack. The CEO held up the new silicon, highlighting a configuration with one Venice CPU and four of the MI455X GPUs per node. That setup scales up to thousands of cores and an absurd amount of memory bandwidth across a full rack. AMD claims the entire Helios platform can deliver nearly three exaflops of AI compute.
The Instinct MI455X GPU is a massive chip with two main compute dies and sixteen HBM4 memory stacks. They are positioning it against Nvidia's upcoming rival hardware, claiming...
