B300
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 5
- DoNews reports Nvidia GPU rental prices rising, with B300 up over 60%
DoNews reported that Nvidia GPU rental prices broadly increased, including a rise of more than 60% for the B300. The report also said China’s access to computing capacity was under pressure.
DoNews · 🔥 0 - NVIDIA Reports Rising GPU Rental Rates and Larger Supply Commitments
NVIDIA interim CEO and CFO Collette Kress said multi-year Hopper GPU rental agreements were recontracted at up to twice their original prices, while B300 rates rose more than 60% since introduction. Wccftech reported that supply commitments increased from $279 billion to $309 billion as of NVIDIA’s fiscal third quarter, primarily reflecting multi-year memory procurement. Kress also cited Meta Muse’s adoption and said agent-harness advances were accelerating agentic AI use on NVIDIA infrastructure.
Wccftech · 🔥 0 - BIS Tightens Advanced AI Chip Controls on China Diversion Routes
Wccftech reports that the U.S. Bureau of Industry and Security issued an interim final rule restricting remote access to advanced AI chips by China- and Macau-based customers. The rule adds licensing and anti-diversion requirements for Singapore, Indonesia, Malaysia, Thailand and Brazil, expands customer due diligence and restricts the commingling of controlled chips with foreign-built products.
Wccftech · 🔥 0 - Epoch AI estimates Horinger has 7% of Colossus 2’s compute
Epoch AI used satellite imagery, public sources, Huawei documentation, and a cooling-equipment model to estimate that Huawei’s Horinger data center in Inner Mongolia has up to 286 MW of IT power capacity for AI. It estimates the operational campus houses about 105,000 Ascend 910C and 8,000 Ascend 910B chips, equivalent to roughly 81,000 H100 chips, or 7% of Colossus 2’s capacity. Epoch AI projects that Horinger will remain substantially behind Colossus 2 through at least 2028, even after planned expansion.
Epoch AI · Gradient Updates · 🔥 0 - Inside Cerebras WSE: a wafer-scale chip built to break the GPU memory wall
A long technical analysis walks through Cerebras's Wafer Scale Engine: instead of dicing a wafer, it turns a full 300mm wafer into a 21.5cm x 21.5cm chip with 84 dies, roughly 900,000 active cores, 4 trillion transistors and about 44GB of on-chip SRAM, with claimed on-chip bandwidth of 21PB/s. The piece argues that keeping weights in SRAM next to the compute units sidesteps the HBM memory wall that leaves GPUs idle while shuffling weights for every token. It also details reticle stitching across scribe lines, whole-wafer testing and fault-tolerant routing, and a four-layer “engine block” assembly of cold plate, wafer, flexible connector and PCB (23kW peak power, vertical power delivery, microchannel liquid cooling).
创业邦 科技 · 🔥 0
Experience and discussion from the community
Share my B300 experienceAsk about B300
Nobody has shared their experience with B300 yet.