NVIDIA Expands NVLink Fusion with New NVHBM Memory
9 Articles
9 Articles
Nvidia no longer just draws the most coveted AI chips on the planet. The group now wants to make their memory, and it has a name for it: the NVHBM.
With NVHBM, AI specialist Nvidia has its own variant of High Bandwidth Memory. The highlight: The memory controller moves into the storage stack. And Nvidia thus binds even more customers to its ecosystem.read more on t3n.de
Samsung has an advantage over its rivals to supply NVHBM chips to Nvidia
This year, Samsung started mass production of HBM4 chips for NVIDIA's flagship Vera Rubin AI platform. This deal will generate billions of dollars in revenue. Now, the South Korean firm appears to have an advantage over its rivals in supplying custom HBM chips for NVIDIA's next flagship AI accelerator architecture. NVIDIA recently announced NVHBM, a custom high-bandwidth memory (HBM) chip designed for its next-generation AI platforms, Rubin Ult…
Samsung would be developing for NVIDIA an 8-layer HBM4E memory, with a target speed of up to 18 Gbps and possible debut in the Rubin Ultra Artificial Intelligence GPU. *** Samsung works on an 8-layer HBM4E requested by NVIDIA for its custom NVHBM memory. NVIDIA looks for speeds from 17 to 18 Gbps, above the 14 Gbps of stable speed of Samsung’s first HBM4E samples. Rubin Ultra could expand its interconnection from 72 configurations to a maximum o…
HBM is becoming faster, larger, and more complex – and in doing so, it occupies precisely the area that chip developers would prefer to use for compute units. NVIDIA aims to address this problem with NVHBM. Introduced on August 26, 2026, this technology combines a modified HBM base die with a special PHY and is expected to deliver up to 30 percent more memory bandwidth, 15 percent less HBM power consumption, and up to 25 percent more usable comp…
Coverage Details
Bias Distribution
- There is no tracked Bias information for the sources covering this story.
Factuality
To view factuality data please Upgrade to Premium












