Background
Visual Memory held U.S. Patent No. 5,953,740, covering a computer memory system that used programmable operational characteristics — specifically, a cache memory whose behavior could be configured based on the type of processor using it. The patent described a three-tier memory hierarchy and claimed a system in which certain operational attributes of the memory (such as whether it stored instructions, data, or both) could be programmed to match the specific performance characteristics and needs of different processor types. The goal was a more efficient memory system that could be adapted to work optimally with different processors without requiring separate hardware for each.
NVIDIA, whose graphics processors used complex memory subsystems, was sued for infringement. The district court dismissed the complaint at the pleading stage under Federal Rule 12(b)(6), holding that the patent claimed the abstract idea of categorical data storage — using memory in categories. Visual Memory appealed.
The Court’s Holding
Judge O’Malley, writing for the majority, reversed the dismissal and held the patent claims patent-eligible at Alice step one. Applying the Enfish framework — which asks whether claims are directed to an improvement in computer technology itself rather than to an abstract idea implemented on a computer — the court found that Visual Memory’s programmable memory system claims fell on the Enfish side of the line. The claims were directed to a specific, technical improvement in how computer memory operates: allowing a single memory system to be configured for different processor types without performance degradation. That was a concrete technological advance, not an abstract concept of categorizing information.
The court emphasized that the patent did not merely describe the result of faster or more efficient memory — it described the technical mechanism (programmable operational characteristics tied to processor type) by which that result was achieved. That specificity distinguished the claims from the type of purely functional, result-oriented claims that the court had found abstract in cases like Electric Power Group and Intellectual Ventures.
Judge Hughes dissented, arguing that the majority had read the claims too broadly and that the claimed invention was essentially the abstract idea of using different types of memory for different types of data — a concept as old as computer science itself, dressed up in processor-specific language.
Key Takeaways
- Patents on specific technical improvements to computer hardware architecture — not just using hardware to achieve an abstract result — can be patent-eligible at Alice step one under the Enfish framework.
- The key distinction is between claims that describe what the improved system does and claims that describe how it does it: specificity about the technical mechanism matters for § 101 eligibility.
- Visual Memory reinforced the growing body of cases — alongside Enfish, McRO, and BASCOM — that define viable pathways for software and hardware patents to survive Alice scrutiny.
- The dissent highlights that the majority/minority divide on § 101 for hardware patents tracks a fundamental disagreement about how specific a claim must be to avoid being characterized as abstract.
Why It Matters
Visual Memory v. NVIDIA is an important 2017 data point demonstrating that the Federal Circuit continued to find ways to uphold software and hardware patents under Alice — the pendulum had not swung entirely toward invalidity. For patent prosecutors in semiconductor and computer architecture fields, the case reinforces that claims tied to specific technical mechanisms for improving computer hardware performance are solid § 101 ground.
The case is also notable for what it did not require: the court did not demand that the claims be unconventional or require novelty-level analysis at § 101. It sufficed that the claims were directed to a technical improvement to computer technology, even if the individual components were each conventional. This reading continues to provide breathing room for patents on computer engineering innovations that operate at the system architecture level.