GoKawiil
Tech News
Benchmark comparisons cited by MacRumors show the Apple M6 scores 93,217 on the Metal graphics test, far below the M5 Ultra's 366,744, which is roughly on par with NVIDIA's 2022-era RTX 4080. In practical use, the M6 Mac Mini still runs Cyberpunk 2077 at an average 55 frames per second with default Mac settings.
Block, Jack Dorsey's company, released a new voice model called Dessa from its research lab, trained entirely on human speech rather than text and runs on a single NVIDIA GPU. The model is full-duplex, meaning it can speak and listen simultaneously, and reportedly produces natural-sounding speech with laughs, pauses, and emotional cadence built directly into the audio rather than driven by text or voice tags.
Nvidia released GPU driver 621.01, enabling Microsoft's Advanced Shader Delivery, a DirectX 12/Windows 11 technology that precompiles shaders in the cloud instead of on a player's PC. Gears of War: E-Day and Call of Duty: Modern Warfare 4 support the feature when downloaded through the Xbox PC app, even if purchased on Steam, and it's also supported on AMD and Qualcomm Snapdragon X2 drivers.
Nvidia's flagship RTX 5090 graphics card remains technically in production, but retail availability has dried up, with resale listings reportedly reaching as high as $10,000. Nvidia has pointed to continued manufacturing as evidence the card is still active in its lineup, even as consumers struggle to find it near its original price.
Reports and rumors indicate Nvidia's RTX 5090 flagship GPU is being phased out, with no replacement expected until at least 2028. Retail prices have climbed from a $1,999 MSRP to over $7,500, and Micro Center is said to be ending its stock of the standalone card, leaving prebuilt PCs as one of the remaining purchase routes.
rGPU is a newly shown project that lets a PyTorch program send tensor operations over TCP to execute on a remote NVIDIA GPU while the application itself keeps running locally. It offers two integration routes: a simple PyTorch 'rgpu' device for programs that opt in, and a CUDA shim that intercepts existing Linux CUDA binaries using libcuda, CUDA Runtime, cuBLAS, cuBLASLt, and cuDNN. Setup involves installing via pip and using an SSH tunnel through the rgpu-run command to connect to the GPU host.
Former EVGA product manager Brendon Ray Hedrick published an account describing how Nvidia's 2016 Founders Edition cards put it in direct competition with board partners like EVGA. He also says Nvidia required partners to sell at least one model at the announced MSRP even at a loss, threatening reduced chip allocations for those who didn't comply, forcing EVGA to offer loss-leading cards alongside pricier premium versions.
An open-source project called virtio-nvgpu forwards Nvidia kernel driver ioctls directly between a Linux KVM guest and host, letting the guest run Nvidia's unmodified user-mode drivers for Vulkan and NVENC without API-level translation. Benchmarks on an RTX 3060 show guest frame times within about 2% of bare metal for frames above roughly 2ms, with CPU usage in the guest matching or slightly beating the host over a 12-second, ~100fps test.
A new Ornn Data paper argues that NVIDIA's older Ampere-generation GPUs, particularly the A100, remain economically useful far longer than assumed because open-weight models can be run cheaply on them. The analysis found that self-hosted open-weight models like gpt-oss-120b produce output more cheaply on A100 chips than on newer H100 hardware, and that A100 rental prices have stayed unusually stable over a five-year contract period compared to Hopper and Blackwell chips.
Asus is selling its ROG Edition 20 anniversary bundle—including the ROG Astral RTX 5090, a Crosshair X870E motherboard, a 3000W Thor Titanium PSU, and an open-frame case—for $10,849.96 on Newegg. With standalone RTX 5090 stock nearly gone in the US and third-party resellers charging close to $10,000 just for the card, this fully-loaded bundle now works out to be one of the least expensive ways to obtain the GPU.
Dutch startup Euclyd raised a 200-million-euro ($231 million) Series A round, with Samsung, Somerset Capital Partners, EQT's Scaleup Europe Fund and Innovation Industries co-leading the investment. Founded in 2024, Euclyd is building an AI chip system with a different architecture from Nvidia's GPUs, aimed specifically at inference workloads, covering both processor and memory design.
Jon Peddie Research says standalone graphics processor shipments for consumer PCs rose 12.2% quarter-over-quarter and 14.1% year-over-year in Q2 2026, even as component shortages pushed prices higher and overall CPU shipments fell. Notebook GPU shipments jumped 16.8% sequentially while desktop GPU shipments slipped 4%, with AMD picking up market share in the process.