NVIDIA NVHBM Memory Promises 30% Higher Bandwidth Than HBM4E
NVHBM moves the memory controller into the 3D HBM stack instead of the compute die, reducing power by 15% and cutting I/O area by up to 67% versus JEDEC HBM4E. NVIDIA plans to use it in future GPUs and offer it to third-party XPU customers via NVLink Fusion.
Related Hardware News
AI CuratedNVIDIA GPU Frame Interpolation Support Lands in FFmpeg
TechPowerUp reports that FFmpeg now supports NVIDIA GPU frame interpolation via Vulkan, using the Optical Flow Accelerator to generate intermediate frames. This allows for efficient video frame rate up-conversion without high CPU load, and avoids proprietary dependencies by using Vulkan.
GeForce RTX 5090 Retail Prices Reach $5,099 as Standard Air-Cooled Cards Approach Halo-Model Pricing
Guru3D reports that standard air-cooled RTX 5090 cards are now selling for roughly $5,000, citing PCPartPicker listings, with the ASUS TUF Gaming OC at $5,099.
ASUS ProArt P14 Combines NVIDIA RTX Spark With 128GB Unified Memory
ASUS's ProArt P14 combines an NVIDIA Grace CPU with a Blackwell RTX GPU, offering 6,144 CUDA cores and 128GB unified memory, according to Guru3D.
[TechPowerUp] NVIDIA DLSS 4.5 Ray Reconstruction review - Transformer Improved
TechPowerUp's review of NVIDIA DLSS 4.5 Ray Reconstruction details the improvements from the new Transformer model, analyzing image quality and performance across supported games.