Hardware
GPUs, workstations, AI accelerators and chip industry news that matter to 3D and AI work.
AI imageTSMC reportedly weighing CoWoS packaging price hikes of up to 20%
According to a 28 October report by TechPowerUp, TSMC may be considering raising CoWoS packaging prices by up to 20%. The company has not confirmed a price increase, while rising demand and growing order backlogs have been reported across the industry.
AI imageMicrosoft Maia 300 shown at OCP 2026: 288 GB of HBM4 memory
Microsoft showed its Maia 300 AI accelerator at OCP 2026 in a four-accelerator system node. The chip has eight HBM4 stacks and 288 GB of memory in total.
AI imageApple Announces 14-inch MacBook Pro with M6: 12-Core CPU and GPU
Apple has announced a new 14-inch MacBook Pro powered by the M6 chip. The company claims up to 30% faster graphics and up to 25% faster AI workload performance compared with M5.
AI imageIntel Crescent Island D270 die shot: 32 Xe3P cores
The D270, part of Intel’s Crescent Island D-series cards announced for AI inference, stands out in a die shot reported by TechPowerUp, featuring 32 Xe3P cores and 256 XMX engines.
AI imageTSMC A14: 19% Smaller 6T SRAM Cell Expected
According to a TechPowerUp report dated October 26, TSMC’s A14 process, expected to enter production in 2028, could reduce the 6T SRAM bit cell from 0.021 µm² to 0.017 µm². The source does not include a direct statement from the company.
AI imageASRock Radeon AI Pro R9700 Brings 32 GB of Memory to Workstations
ServeTheHome reviewed the ASRock Creator Radeon AI Pro R9700 workstation card on 26 October 2026. Based on RDNA 4, the model focuses on AI inference with 32 GB of GDDR6 memory, FP8 support and 300 W total board power.
AI imageAivres showcases Vera Rubin NVL72 and HGX Rubin NVL8 systems
At OCP Global Summit 2026 in San Jose, Aivres showcased its Vera Rubin NVL72 rack and KL2288 compute tray based on HGX Rubin NVL8. The company shared details on the systems’ cooling setups and technical specifications.
AI imageGeForce RTX 5090 Cards in China Are Being Converted Into AI Servers
According to an article published by Tom’s Hardware on 23 October, leaked images show new GeForce RTX 5090 cards in China being stripped for parts and converted into AI server cards with dual-slot blower cooling.
AI imageMSI XpertStation WS300 reviewed: GB300 brings AI power to the desktop
ServeTheHome reviewed the NVIDIA GB300-based MSI XpertStation WS300. The tower offers 496 GB of system memory, 252 GB of HBM3E and dual 400 Gbps network connections, with substantial power and cooling requirements.
AI imageMicron shelves 4F² DRAM and turns to 3D memory
Micron said at its October 21 technology webinar that it will not commercialize 4F² DRAM. The company will continue developing its 6F² design while turning to 3D DRAM, arguing that capacitor placement—not just cell area—limits density.
Google Cloud A5X General Availability Planned for Next Month
NVIDIA announced that Google Cloud plans to make Vera Rubin-based A5X instances generally available next month. The infrastructure targets AI training, inference and large-scale reinforcement learning workloads.
AI imageNVIDIA BlueField-4 STX with CMX Server Card Shown at OCP 2026
NVIDIA’s BlueField-4 STX server card was shown at OCP 2026 for Context Memory (CMX) systems. The design uses a single Vera CPU and SOCAMM memory, and is described as a lower-cost option than the earlier dual-Vera setup.
AI imageNVIDIA VBIOS Vix Editor: Change RTX settings without flashing
Panchovix’s tools let users apply modified VBIOS settings to GeForce RTX 30, 40 and 50 series cards without writing to the flash chip. While this approach reduces flash-related risks, it does not eliminate the possibility that extreme settings could damage hardware.
Jetson Orin Nano 2 announced: A new robotics computer for edge AI
NVIDIA has introduced Jetson Orin Nano 2 for robotics and on-device image processing applications. The company says the new model increases inference capacity while reducing power consumption.
NVIDIA RTX Spark Goes on Sale, Targeting Local AI and Content Creation
NVIDIA has put slim Windows PCs under the RTX Spark name on sale, targeting local AI, content creation, software development, and gaming. The official page does not disclose pricing or technical specifications.
AI imageSurface Laptop Ultra review: Arm laptop hardware with 128 GB of memory
ServeTheHome reviewed a configuration of the Microsoft Surface Laptop Ultra with 128 GB of memory. The Arm laptop with an NVIDIA RTX Spark N1X processor stands out for its display, ports and considerations for architectural workflows.
AI imageNVIDIA RTX Spark laptops go on sale: built for local AI and creative work
NVIDIA announced that Windows laptops powered by RTX Spark chips are now available. The company says they offer up to 128 GB of unified memory, while the ASUS ProArt P16 and P14 have reportedly gone on sale in Canada; the cited sources contain no information about Turkey.
AI imageTOPC TXA2 Mini Workstation Goes on Sale with 192 GB of Unified Memory
TOPC has launched the TXA2 mini workstation, powered by a Ryzen AI Max+ Pro 495 processor, with 2 TB and 4 TB storage options. Up to 160 GB of its 192 GB of unified memory can be shared with the iGPU.
AI imageIntel GPU D-Series 512 GB Memory Claim Remains Unverified
Some partner cards for Intel’s GPU D-Series accelerator may have 512 GB of memory, according to a claim. Intel has not confirmed it; the company has stated a partner limit of 480 GB.
AI imageGeForce RTX 5090 Listed as “Not Carried” at Micro Center
According to a 15 October report by Tom’s Hardware, Micro Center has marked the GeForce RTX 5090 as “Not Carried.” This does not confirm that production has stopped; NVIDIA has not made an official statement on the matter.
AI imageWindows is being tuned for local AI on NVIDIA RTX Spark laptops
The Verge reported on 15 October that Microsoft and NVIDIA are preparing Windows for RTX Spark laptops. The optimizations focus on local AI agents, Arm app compatibility, and memory management.
AI imageTuxedo Stellaris 16 Gen 8 arrives with GPU options up to RTX 5090
Tuxedo Computers has launched the Stellaris 16 Gen 8 laptop. It features a Core Ultra 9 290HX Plus, GPU options ranging from the RTX 5070 to the RTX 5090, and optional external liquid cooling.
AI imageNVIDIA RTX 50: Some Cards Cost More in Japan Than in the US
In Tom’s Hardware’s comparison of retail prices in Japan with Amazon US listings, some RTX 50 models were more expensive in Japan. The price difference for the RTX 5060 Ti 8 GB was calculated at 41.10%.
AI imageDLSS 5 October update: NVIDIA announces a twofold speed increase
According to NVIDIA, its October update for DLSS 5 doubles model performance and reduces VRAM use by 20% at 4K. The update is scheduled to arrive in NBA 2K27 on 16 October.
About this category
The Hardware category covers hardware news that directly affects 3D, architectural visualization, and AI work: graphics cards from NVIDIA, AMD, and Intel; professional workstation cards; processors; render workstations; AI accelerators; and AI PCs. We share new products’ VRAM, performance, and pricing details, and explain what they mean for rendering and running AI models locally.
Industry developments such as chip production, supply shortages, export restrictions, and price changes also feature here. A factory shutdown or a new restriction, for example, could affect the price and availability of the graphics card you plan to buy a few months from now. We compile our coverage from official manufacturer announcements and reliable hardware publications.
Rather than focusing on gaming benchmarks, we prioritize real workloads: render times in V-Ray, Blender, D5, or Unreal Engine; memory capacity for large scenes; and the hardware required to run AI models locally. This helps you make upgrade decisions based on your own workflow.