NVIDIA may be getting ready to expand its Blackwell GPU lineup again, after five new PCI device IDs tied to Blackwell graphics processors were spotted in an online PCI ID database. While PCI IDs don’t reveal full specifications, new entries like these often point to fresh silicon variants in development, upcoming accelerators, or internal revisions that haven’t been announced yet.
What’s been spotted includes five new GPU identifiers: one labeled as GB110, three listed as GB112, and one noted as GB120. The GB110 naming is especially interesting because it aligns with NVIDIA’s Blackwell Ultra direction for data center hardware, commonly associated with the next step in its server and AI accelerator roadmap. The GB112 and GB120 entries, meanwhile, look like additions that haven’t been clearly mapped to publicly known products so far.
Blackwell is already a broad umbrella for NVIDIA. It spans multiple segments, including high-end data center platforms and consumer graphics cards. On the consumer side, NVIDIA’s current GeForce RTX 50 series is tied to the GB200 consumer family, which includes GPUs such as GB202, GB203, GB205, GB206, and GB207—each already represented by an RTX 50-series product. Because these newly listed IDs are GB110/GB112/GB120 rather than the GB20x naming typically seen in GeForce-class GPUs, the evidence points more toward data center or enterprise hardware than a gaming GPU refresh.
At this stage, no specs, performance targets, or product names are attached to GB112 and GB120. However, there are a couple of plausible explanations. They could be refined Blackwell Ultra variants aimed at NVIDIA’s major server platforms, such as DGX, HGX, or MGX systems, where NVIDIA frequently introduces tuned configurations to better match power envelopes, memory setups, or deployment needs. The other possibility is that these represent entirely new Blackwell-based parts designed for specific system roles that haven’t been publicly discussed yet.
The timing also fits a data center-focused update. NVIDIA’s next-generation Rubin platform is expected to be revealed at major industry events, but broader production isn’t expected until the second half of the year. With Blackwell and Blackwell Ultra ramping in the meantime, NVIDIA has strong incentive to keep optimizing its current stack—either by adding new SKUs, refining yields, or introducing incremental variants that improve efficiency and supply flexibility for hyperscalers and enterprise customers.
Some readers may wonder whether the new PCI IDs hint at a GeForce RTX 50 SUPER lineup. Based on what’s known, that seems unlikely. A “SUPER” refresh would typically continue using existing consumer GB200-family GPUs, adjusting memory configurations and core counts rather than introducing brand-new GB11x/GB120 identifiers. Current expectations place any RTX 50 SUPER-series timing around mid-2026, although ongoing memory supply constraints could influence schedules.
For now, the appearance of GB110, GB112, and GB120 in PCI ID listings is a strong signal that more Blackwell-based GPUs are on the way—most likely aimed at data center platforms where NVIDIA is under constant pressure to deliver more performance per watt, better availability, and more tailored accelerator options for AI training and inference.
