In the ever-evolving world of PC hardware, the pursuit of ultimate performance often leads to extreme measures. Recently, Tony Yu, General Manager of Asus China, took the Nvidia GeForce RTX 5090D, a China-exclusive graphics card, to new heights by overclocking it to an astounding 3.4 GHz using liquid nitrogen cooling. This feat of engineering resulted in the card consuming a staggering 1,000W of power but also delivered benchmark scores that surpassed even dual RTX 3090 Ti and quad GTX 1080 Ti configurations.
This groundbreaking overclocking experiment occurred in late January 2025, shortly after the release of the RTX 5090 series. Yu, a renowned figure in the overclocking community, conducted the tests at Asus’s lab in China, utilizing a high-end test bench featuring an Asus ROG Astral RTX 5090D, Ryzen 7 9800X3D processor, Asus ROG X870E Hero motherboard, and G.Skill DDR5-6000 CL30 memory. The purpose of this endeavor? To showcase the raw potential of the RTX 5090D and its Blackwell architecture when pushed to the absolute limit.
Unveiling the Beast: RTX 5090D and its Overclocking Potential
The RTX 5090D, while sharing similar specifications with its global counterpart, faces artificial limitations imposed by Nvidia to comply with US export restrictions. These limitations primarily affect its AI and cryptocurrency mining performance. However, Yu’s experiment demonstrates that the card’s gaming capabilities remain unhindered and ripe for extreme overclocking.
To achieve the 3.4 GHz clock speed, Yu employed liquid nitrogen cooling and performed physical modifications to the graphics card, likely involving shunt mods. These modifications, coupled with the extreme cooling, allowed the RTX 5090D to maintain stability at a core clock speed that is significantly higher than its stock boost clock. Additionally, the memory was also overclocked to an impressive 34 Gbps (17 GHz effective).
Benchmark Domination: Crushing the Competition
The results of this overclocking endeavor were nothing short of extraordinary. The RTX 5090D, running at 3.4 GHz, shattered previous records in several 3DMark benchmarks, including Port Royal, Fire Strike Ultra, and Wild Life Extreme. In Port Royal, a benchmark heavily focused on ray tracing performance, the overclocked RTX 5090D outperformed a dual RTX 3090 Ti setup. Furthermore, it decimated a quad GTX 1080 Ti configuration in Fire Strike, showcasing its immense rasterization prowess.
These benchmark results highlight the incredible performance gains achievable through extreme overclocking. While such overclocks are not practical for everyday use due to the requirement of specialized cooling solutions and potential risks, they serve as a testament to the underlying architecture’s capabilities.
A Glimpse into Blackwell: Die-Shots and Architectural Insights
Yu’s experiment also provided valuable insights into the Blackwell architecture powering the RTX 5090D. High-resolution die-shots of the GB202 chip, the heart of the RTX 5090 series, were captured, revealing intricate details of its design. The GB202 die measures around 744mm², making it the largest consumer GPU since the TU102 (Turing) and significantly larger than the AD102 found in the RTX 4090.
The die-shots showcase the complex arrangement of various functional blocks within the GB202, including the streaming multiprocessors, memory controllers, and other essential components. These images offer a fascinating look into the advancements made in GPU technology and provide valuable information for enthusiasts and researchers alike.
GDDR7 Memory: A New Era of Bandwidth
The RTX 5090D is the first graphics card to feature GDDR7 memory, a significant leap forward in memory technology. This new memory standard offers substantially higher bandwidth compared to its predecessor, GDDR6, contributing to the card’s impressive performance. Yu’s experiment also included die-shots of a Samsung GDDR7 module, highlighting its compact size and advanced design.
The adoption of GDDR7 memory marks a crucial step in the evolution of graphics cards, paving the way for even more powerful and capable GPUs in the future. As games and other applications become increasingly demanding, the increased bandwidth offered by GDDR7 will be essential for maintaining smooth and responsive performance.
The Future of Overclocking and GPU Technology
Yu’s overclocking experiment with the RTX 5090D not only demonstrates the card’s incredible potential but also provides a glimpse into the future of GPU technology. As manufacturing processes continue to advance and new architectures emerge, we can expect even greater performance gains and overclocking headroom in the years to come.
The pursuit of extreme performance will undoubtedly continue, driving innovation and pushing the boundaries of what’s possible with graphics cards. Enthusiasts and overclockers will continue to explore new ways to extract every ounce of performance from these powerful devices, leading to new records and groundbreaking achievements.
My Thoughts and Experiences
Personally, I’ve always been fascinated by the world of overclocking. Pushing hardware to its limits and witnessing the performance gains firsthand is an incredibly rewarding experience. While I haven’t had the opportunity to experiment with liquid nitrogen cooling, I’ve dabbled in traditional overclocking with air and liquid coolers, achieving respectable results.
Seeing Yu’s experiment with the RTX 5090D truly inspires me to further explore the world of extreme overclocking. The level of dedication and expertise required to achieve such results is remarkable, and it motivates me to continue learning and experimenting with new techniques.
The RTX 5090D’s record-breaking performance under liquid nitrogen cooling highlights the incredible advancements in GPU technology. It’s exciting to imagine what the future holds for graphics cards and the experiences they will enable. As an avid gamer and PC enthusiast, I eagerly await the next generation of GPUs and the new possibilities they will unlock.
Add Comment