NVIDIA Blackwell GB202 GPU and GDDR7 Memory Revealed in Stunning Die Shots, ASUS GeForce RTX 5090 Astral Overclocked to 3.4 GHz with LN2, Consuming Nearly 1000W

Recently, ASUS showcased high-resolution die shots of its GeForce RTX 5090’s Blackwell GB202 GPU, alongside impressive overclocking results achieved with the RTX 5090 Astral model, which exceeded 3 GHz.

ASUS GeForce RTX 5090 Astral: Overclocked Performance Metrics

Tony Yu, ASUS’s manager in China, recently ran tests on the ROG GeForce RTX 5090 Astral GPU using liquid nitrogen (LN2) cooling, leading to remarkable findings. Under these extreme cooling conditions, the GPU reached a peak frequency of 3390 MHz and a memory frequency of 34 Gbps. This performance marks a 21% boost over the standard memory clock of 28 Gbps, showcasing the card’s potential for high-performance gaming and applications.

NVIDIA Blackwell GB202 GPU Overclocked Results — Image Source: ASUS’s Tony Yu

With this advanced cooling method, we can anticipate even greater records as professional testers continue to push boundaries in GPU performance.

Benchmarking Performance

The overclocked ASUS GeForce RTX 5090 was assessed using various benchmarks, including 3DMark Time Spy Extreme, Fire Strike Ultra, and Port Royal. The results underscore a significant performance leap compared to previous models:

The overclocking of the ASUS RTX 5090 ROG Astral led to a notable performance increase, showing a 25% gain in 3DMark Port Royal and an 18% gain in Time Spy Extreme. However, this impressive performance is coupled with significant power requirements—approximately 1000W for the GPU itself, with total system consumption nearing 1800W.

NVIDIA Blackwell GB202 GPU Close-Up — Image Source: ASUS’s Tony Yu

Innovations in GPU Technology: The Blackwell GB202

The NVIDIA RTX Blackwell GPUs are produced using TSMC’s 4nm process node, encapsulating a staggering 92 billion transistors. They provide an impressive 125 TFLOPS of FP32 compute power and boast GDDR7 memory, which offers a bandwidth of up to 1.8 TB/s. This architecture represents a substantial advancement in GPU technology, setting a high bar for gaming performance.

The launch of GPUs based on GDDR7 signifies a much-needed shift from GDDR6/X memory, enhancing both bandwidth and data rates considerably while also reducing power consumption through sophisticated engineering techniques.

This cutting-edge technology, combined with impressive die size increases, provides insight into the premium pricing strategy around the RTX 5090. While NVIDIA has yet to confirm a Titan or Ti variant, the ongoing advancements indicate room for future models, possibly designed for data-intensive workstation applications.