Latest

Every published item, in order · 6240 items
Oct 13
Tomorrow 08:00, the daily lands in your inboxTelegram channel or email. Same screening as this page, no ads.
ProductsBuildersMulti-sourceSelected10:14
NVIDIA Opens Vera Rubin Rack Designs Through OCP

Open-sourcing rack-level hardware design shifts AI data-centre building from proprietary reference designs toward a vendor-neutral standard, letting system integrators assemble Vera Rubin-class infrastructure without NVIDIA-branded enclosures. Cloud providers and hardware teams in the region evaluating GPU cluster builds can adopt the MGX 1.2 specs and DSX components to cut deployment time, while the Open Data Center Specification v0.7 gives them a cross-vendor power and cooling blueprint alongside AMD and Intel-platform alternatives.

Event
IT之家Source
ProductsBuildersSelected09:45
SK hynix reveals H-TPU processor architecture for AI data centre SSDs

The bottleneck in AI data centres is shifting from GPUs to storage and memory, and fixing the SSD controller's power draw targets exactly that layer. Engineers designing storage or AI infrastructure can track how H-TPU pairs with SK hynix's upcoming HBF flash to cut controller energy costs. The 2.5% power figure gives a concrete benchmark against conventional CPU-based SSD controllers when planning capacity.

IT之家Source
Oct 12
ProductsUS vs ChinaMulti-sourceSelected23:56
Microsoft Maia 300 AI Chip Uses Chiplet Architecture for 7x Tokens

The 7x token gain over Maia 200 shows Microsoft is closing the efficiency gap with Nvidia GPUs on its own cloud stack, which could eventually lower inference costs for Azure-hosted models used across Southeast Asia. Developers and enterprises running long-context or reasoning workloads on Azure should watch which regions and services get Maia 300 capacity first.

Event
Blognone@mkSource
ProductsBuildersMulti-sourceSelected19:23
Alibaba Qwen releases QwenGyre RL framework for long-horizon agents

Training agents on multi-hour tasks usually wastes GPU cycles waiting for the slowest rollout, and QwenGyre's elastic scheduling attacks exactly that bottleneck. Teams building agent RL pipelines get an open reference for turning partial, timed-out trajectories into training signal instead of discarding them. The 1.78× speedup over Async makes it a concrete alternative for labs weighing Qwen's stack against existing open-source RL training frameworks.

Event
IT之家Source

Only the latest 2000 items are listed; for older ones see the daily archive