Blackwell

Headlines · 4

1. 3d
   ASUS ProArt P16 review: first NVIDIA RTX Spark laptop arrives

   The ASUS ProArt P16 is among the first laptops to ship with NVIDIA's RTX Spark platform, built around the N1X SoC — a 20-core Arm CPU plus a 6,144-CUDA-core Blackwell GPU, both on TSMC 3nm — supporting up to 128GB of LPDDR5X unified memory, roughly 300 GB/s of bandwidth and 1,000 TOPS of NVFP4 AI compute. Wccftech's early benchmarks show class-leading memory bandwidth but multi-core and single-core 3DMark CPU Profile scores behind x86 rivals such as the Core Ultra X9 388H and Ryzen AI MAX+ 395. NVIDIA claims the laptop can sustain full performance on battery, pricing has not been announced, and further benchmarks are still pending.

   Wccftech · 🔥 0
2. 4d
   vLLM adds day-0 support for Reflection's 501B Beam model with Blackwell optimizations

   vLLM published day-0 serving support for Reflection's Beam, a 501B-parameter mixture-of-experts model with 23B active parameters per token, optimized for NVIDIA Blackwell. The engine handles reasoning, tool calling and structured outputs out of the box and can run Beam with tensor, data or expert parallelism, with a Docker image and launch command provided. The post says coding agents autonomously tuned the implementation so batch-1 decode is 2.71× faster and prefill 1.87× faster than the first version, and that Beam's DFlash2 drafter reaches 708 tokens/s for a single request on TP8 with an acceptance length of 3.7. It also describes day-0 reinforcement-learning support built with the SkyRL team, reporting AIME25 average accuracy rising from about 61% to about 79% after 50 training steps.

   vLLM Blog · 🔥 0
3. 5d
   KYEC signs NVIDIA prepayment deal to build dedicated US chip testing capacity

   Taiwanese semiconductor testing firm King Yuan Electronics (KYEC) said its board approved a customer prepayment agreement with NVIDIA International, Inc., under which NVIDIA will advance funds for KYEC and its wholly owned US subsidiary AeternZ Semiconductor Testing Corporation to build testing capacity in the United States, reserving dedicated capacity and testing services for NVIDIA, with the prepayment offset against future service payments until fully deducted. KYEC said the deal deepens its long-term cooperation with NVIDIA and expands US testing capacity, and that the revised contract structure means the parent company no longer needs to provide a US$300 million endorsement guarantee, which it has cancelled. The report notes KYEC's board approved a US plant investment in July 2026 with a cap of US$1.4 billion, and that its 2026 capital expenditure was raised from NT$39.372 billion to NT$50 billion, with AI orders spanning NVIDIA's Blackwell and Rubin series plus ASIC and CPU testing; NVIDIA did not disclose the prepayment amount or full contract details, after signing a US$1.5 billion multi-year deal with Amkor in July 2026.

   科技新报 · 🔥 0
4. Oct 9
   Citi: Meta Muse agents could shift CPU-GPU ratios toward 1:1

   In an Oct. 6 research note, Citi argues that always-on personal AI agents such as Meta Muse consume compute at orders of magnitude above chatbots, pushing the CPU-to-GPU ratio from 1:8 in training and 1:4 in inference toward 1:1 or higher for agentic workloads. The bank estimates 100 million Muse daily active users would require roughly 200,000–390,000 Blackwell-class GPUs, worth $7 billion–$19 billion in one-time revenue for Nvidia, and projects the global CPU total addressable market growing from $29 billion in 2025 to $300 billion by 2030. It also flags knock-on demand for DRAM, SSDs and data-centre network traffic.

   36氪 人工智能 · 🔥 0

Experience and discussion from the community

Share my Blackwell experienceAsk about Blackwell