Nvidia tests at least three Rubin Ultra GPU variants amid HBM shortage
The chipmaker is weighing a design change that would use less high-bandwidth memory in Rubin Ultra to cope with tight supplies of advanced HBM chips, The Information reported.
Fact Check
The core claim—NVIDIA weighing a design change to use less HBM in Rubin Ultra amid HBM shortage and testing multiple variants—is corroborated by independent sources. The TrendForce report (via I-Connect007) explicitly lists three reduced HBM configurations (8-Hi HBM4e, 12-Hi HBM4, 8-Hi HBM4) beyond the original 12-Hi HBM4e baseline, and Tech Times independently reports NVIDIA evaluating multiple 8-Hi/12-Hi HBM4/HBM4E options driven by supply constraints. The Odaily newsflash restates the claim exactly, attributing it to The Information. The 'at least three variants' wording aligns with TrendForce's enumeration.
Summary
Nvidia is considering a significant design adjustment for its Rubin Ultra GPU lineup as shortages of advanced HBM (high-bandwidth memory) chips pressure supply. The company has tested at least three Rubin Ultra GPU variants in recent weeks, with some carrying less memory than Nvidia had previously outlined, according to The Information. The move would help Nvidia manage a market where enough advanced HBM chips may not be available, underscoring how memory supply can shape high-end AI chip configurations as demand for AI computing hardware remains strong.
Terms & Concepts
- HBM: High-bandwidth memory used in AI accelerators to move data faster than conventional memory.
- GPU: Graphics processing unit, a chip architecture widely used for parallel workloads such as AI training and inference.
- Rubin Ultra GPU: A next-generation Nvidia GPU platform referenced in the report.