News from the AI & ML world
DeeperML - #aichips
|
@www.marktechpost.com
//
Nvidia is reportedly developing a new AI chip, the B30, specifically tailored for the Chinese market to comply with U.S. export controls. This Blackwell-based alternative aims to offer multi-GPU scaling capabilities, potentially through NVLink or ConnectX-8 SuperNICs. While earlier reports suggested different names like RTX Pro 6000D or B40, B30 could be one variant within the BXX family. The design incorporates GB20X silicon, which also powers consumer-grade RTX 50 GPUs, but may exclude NVLink support seen in prior generations due to its absence in consumer-grade GPU dies.
Nvidia has also introduced Fast-dLLM, a training-free framework designed to enhance the inference speed of diffusion large language models (LLMs). Diffusion models, explored as an alternative to autoregressive models, promise faster decoding through simultaneous multi-token generation, enabled by bidirectional attention mechanisms. However, their practical application is limited by inefficient inference, largely due to the lack of key-value (KV) caching, which accelerates performance by reusing previously computed attention states. Fast-dLLM aims to address this by bringing KV caching and parallel decoding capabilities to diffusion LLMs, potentially surpassing autoregressive systems. During his keynote speech at GTC 2025, Nvidia CEO Jensen Huang emphasized the accelerating pace of artificial intelligence development and the critical need for optimized AI infrastructure. He stated Nvidia would shift to the Blackwell architecture for future China-bound chips, discontinuing Hopper-based alternatives following the H20 ban. Huang's focus on AI infrastructure highlights the industry's recognition of the importance of robust and scalable systems to support the growing demands of AI applications.
Share:
References :
Classification:
@www.artificialintelligence-news.com
//
Nvidia is actively working to maintain its presence in the rapidly expanding Chinese AI chip market, even amidst tightening U.S. export controls and rising domestic competition from companies like Huawei. To navigate these challenges, Nvidia is developing a cheaper, stripped-down AI chip based on its Blackwell architecture, specifically tailored for the Chinese market. This new chip is designed to comply with U.S. regulations while offering competitive performance, aiming to regain some of the market share Nvidia has been losing.
This strategic gamble involves significant cost reductions and performance trade-offs. The new processor, expected to be priced between $6,500 and $8,000, will use the RTX Pro 6000D foundation with GDDR7 memory, forgoing the high-bandwidth memory found in premium variants and Taiwan Semiconductor's advanced CoWoS packaging technology. This decision is driven by the need to meet export restrictions and lower manufacturing complexity, following a $5.5 billion writedown Nvidia absorbed after the prohibition of its H20 sales in China. Huawei's growing influence in the Chinese AI chip landscape is a key factor driving Nvidia's strategy. Huawei's Ascend 910C and 910B processors are gaining adoption among major domestic tech firms, challenging Nvidia's dominance, particularly in inference applications. The competitive pressure extends to complete infrastructure solutions, with Huawei's CloudMatrix 384 rack system directly challenging Nvidia's Blackwell GB200 NVL72 configuration. The company hopes the new budget option will help them survive the growth of the internal competion,
Share:
References :
Classification: |
BenchmarksBlogsResearch Tools |