New Research Unveils Vulnerability in Nvidia GPUs with GDDR6 Memory Exploitable Despite IOMMU Protection

• April 5, 2026

Security researchers have identified a critical vulnerability affecting Nvidia GPUs that utilize GDDR6 memory, exposing systems to the GeForge attack. This exploit enables attackers to manipulate GPU memory contents, potentially leading to broad system control.

Prior to this discovery, similar attacks such as GDDRHammer and GeForge were known to target the contents of GDDR6 memory on Nvidia graphics cards. These attacks specifically compromised GPU page tables, redirected central processing unit (CPU) memory accesses, and threatened the overall security of the host system.

Breach of Assumed IOMMU Protection

Historically, it was believed that activating the Input-Output Memory Management Unit (IOMMU) provided a robust defense against this class of attacks. The IOMMU mechanism is designed to manage and restrict device access to memory, thereby isolating critical system resources from unauthorized manipulation.

However, recent findings demonstrate that the GeForge attack remains effective even when IOMMU protections are enabled. This indicates a more profound vulnerability affecting the hardware and firmware layers of Nvidia GPUs using GDDR6 memory than previously understood.

The implications are significant because the ability to compromise GPU memory and bypass memory management safeguards means attackers could escalate privileges to superuser levels. Achieving such control presents a risk not only to graphics processing operations but also to the integrity of the entire computing system.

Details about the exact method by which GeForge circumvents IOMMU defenses have yet to be publicly disclosed. Likewise, Nvidia and security experts have not released an official response or mitigation strategies addressing this vulnerability.

This discovery underscores the need for ongoing scrutiny of hardware security, especially as advanced GPUs with high-speed GDDR6 memory are increasingly integrated into critical computing environments, including artificial intelligence workloads and data centers.

Users and organizations deploying Nvidia GPUs are advised to monitor updates from hardware manufacturers and cybersecurity authorities. Due to the nature of the exploit, traditional software patches or driver updates may be insufficient without corresponding firmware or architectural improvements.

As hardware-based attacks continue to evolve, comprehensive collaboration between chip makers, security researchers, and industry specialists will be essential to develop effective defenses against emerging threats that challenge the foundational security assumptions of modern computing platforms.

Nvidia GPUs with GDDR6 memory are susceptible to the GeForge attack, bypassing previous IOMMU safeguards and risking full system compromise.