By Manny Kressel

CEO and Founder of BitMindz

Why the NVIDIA Blackwell Chipset Is Perfect for Password Cracking

The NVIDIA Blackwell chipset represents a monumental leap in GPU architecture, delivering unprecedented computational power that makes it a game-changer for tasks such as password cracking, which we prefer to call decryption engines. Designed to fuel the next era of AI, high-performance computing, and generative workloads, Blackwell’s cutting-edge features also position it as an ideal platform for cracking password hashes at blistering speeds. In this article, we’ll explore why the Blackwell chipset is perfectly suited for decryption, delving into its technical prowess, efficiency, and real-world implications, such as the BitMindz Decryption Engine.

Unmatched Computational Power with CUDA Cores

Password cracking relies heavily on brute-forcing or intelligently guessing password combinations, a process that demands massive parallel computation. The Blackwell chipset excels here due to its vast number of CUDA cores, NVIDIA’s proprietary parallel processing units. While exact specifications for Blackwell’s CUDA core count vary by model, NVIDIA has emphasized that Blackwell GPUs offer significant improvements over previous architectures, such as Hopper and Ampere, which already boasted tens of thousands of cores. Our last benchmark testing with these tens of thousands of cores proved to be a significant advancement for our decryption engines compared to the previous architecture generations.

Each CUDA core can process multiple hashing operations simultaneously, making tasks like cracking MD5, SHA-1, or NTLM hashes exponentially faster. For example, a single NVIDIA RTX 4090, based on the older Ada Lovelace architecture, can achieve theoretical hash rates of 255 GH/s (giga-hashes per second) for NTLM hashes, cracking an 8-character password in just 7.8 hours. With Blackwell’s enhanced core count and optimized architecture, we can expect even higher hash rates, potentially reducing cracking times for complex passwords to mere hours or minutes when scaled across multiple GPUs.

Advanced FP8 Precision for Hashing Efficiency

Blackwell introduces transformative support for FP8 (8-bit floating-point) precision, enabling high-throughput computations with minimal loss of accuracy. Password cracking doesn’t require the numerical precision of scientific simulations but benefits immensely from the raw throughput FP8 provides. By processing more hash calculations per second, Blackwell GPUs can iterate through password combinations faster than ever, making them ideal for both brute-force and dictionary-based attacks.

This efficiency is particularly valuable when cracking modern hashing algorithms, which are designed to be computationally intensive. For instance, while an RTX 4090 takes 99 years to crack an 8-character bcrypt hash, Blackwell’s improved FP8 performance and higher core count could significantly reduce this time, making even “secure” algorithms more vulnerable to well-resourced attackers.

Multi-Instance GPU (MIG) for Scalable Decryption Engine Rigs

One of Blackwell’s standout features is its Multi-Instance GPU (MIG) technology, which allows a single GPU to be partitioned into multiple isolated instances. This is a boon for password cracking, as it enables cybersecurity professionals to allocate specific GPU resources to different cracking tasks or algorithms simultaneously. For example, one MIG instance could focus on cracking NTLM hashes using a brute-force attack, while another tackles bcrypt with a dictionary attack on the same GPU.  This is a massive advancement for decryption engines if the run-of-the-mill password cracking software can leverage the technology associated with Multi-Instance GPU.

This flexibility maximizes resource utilization, reducing the need for multiple physical GPUs in a cracking rig. For organizations building dedicated password-auditing systems, MIG lowers costs and simplifies scaling, allowing a single Blackwell-powered server to handle diverse workloads efficiently.

Energy Efficiency: Cracking More with Less

Password cracking is notoriously power-hungry, especially when running multiple GPUs for extended periods. Blackwell addresses this with up to 25x lower energy consumption compared to its predecessors for certain workloads. While password cracking differs from generative AI tasks, the chipset’s energy-efficient design still translates to lower operational costs for cracking rigs. This is critical for red teams or penetration testers who need to run cracking operations for days or weeks without breaking the bank on electricity bills.

For comparison, a rig with eight RTX 4090s, drawing 700–900W per power supply, can, in theory, crack an 8-character password in 48 minutes but requires dedicated circuits to avoid tripping breakers. Yes, we’ve done that before!!! Blackwell’s efficiency could allow similar performance with less power draw, making it feasible to build compact, high-performance cracking stations that don’t require industrial-grade power infrastructure.

Support for Hashcat and CUDA-Accelerated Tools

The Blackwell chipset is fully compatible with NVIDIA’s CUDA platform, which powers popular password-cracking decryption tools like Hashcat. Hashcat uses CUDA to distribute hashing tasks across thousands of GPU cores, achieving unparalleled performance. Blackwell’s enhanced CUDA core architecture and higher memory bandwidth ensure that Hashcat can push the boundaries of cracking speed, especially for complex algorithms.

Additionally, Blackwell’s support for NVIDIA’s latest software development kits (SDKs) ensures seamless integration with cracking tools. Examiners can use optimized libraries to fine-tune their cracking workflows, whether they’re targeting stolen password databases or auditing corporate Active Directory environments.

Real-World Implications

The Blackwell chipset’s password-cracking prowess underscores the importance of robust cybersecurity practices. While it empowers examiners to identify weak passwords and strengthen systems, it also highlights the risks posed by outdated hashing algorithms like MD5, which modern GPUs can crack in seconds. Even bcrypt hashes are at risk if passwords lack sufficient length and complexity.

Cybersecurity professionals can use Blackwell-powered rigs to perform password audits, simulating real-world attacks to uncover vulnerabilities. For instance, a single Blackwell GPU could crack an 8-digit numeric PIN in hours, while a multi-GPU setup could reduce this to minutes. By demonstrating how quickly weak passwords fall, organizations can justify adopting stronger policies, such as 12-character minimums, multi-factor authentication (MFA), and advanced hashing algorithms.

Conclusion

The NVIDIA Blackwell chipset is a powerhouse for password cracking, combining unmatched CUDA core density, FP8 precision, MIG flexibility, and energy efficiency to redefine what’s possible in computational brute-forcing. Its ability to accelerate tools like Hashcat makes it a dream for examiners.  It’s something we are already starting to implement in the BitMindz like of Decrytion Engines.

 

For questions or comments, please reach out to Manny Kressel at manny@bitmindz.com