NVIDIA Blackwell GB202 GPU & GDDR7 Memory Pictured In Beautiful Die Shots, ASUS GeForce RTX 5090 Astral Overclocked To 3.4 GHz With LN2 & Consumes Almost 1000W

ASUS has posted high-res die shots of the GeForce RTX 5090’s Blackwell GB202 GPU, the GDDR7 memory & also some fantastic overclocked results by pushing its Astral flagship over 3 GHz.

ASUS GeForce RTX 5090 Astral Overclocked With LN2 Over 3 GHz, Results In Almost 1000W Power Consumption

ASUS’s China manager, Tony Yu, managed to take the flagship ROG GeForce RTX 5090 Astral GPU for a spin with LN2 cooling and achieved some surprisingly good figures.

For overclocking, Tony is utilizing the best kind of cooling for extreme overclocks, which is LN2, and with this, he was able to achieve a maximum GPU frequency of 3390 MHz and a memory frequency of 34 Gbps, which is 21 percent higher than the stock frequency rated at 28 Gbps. These are both impressive figures, and we can expect even higher overclock once professionals start rolling out their records.

As for the performance, the graphics card was tested in 3DMark Time Spy Extreme, Fire Strike Ultra, Port Royal, and Wild Life Extreme benchmarks. The following is how these results compare to the numbers from our review of the RTX 5090:

In terms of performance improvement, the ASUS GeForce RTX 5090 ROG Astral overclocked GPU achieved a 25% uplift in 3DMark Port Royal, an 18% uplift in Time Spy Extreme, and a 12% uplift in FireStrike Ultra. This is a nice improvement, but we should remember that this is from a 41% uplift in its boost clock to 3.4 GHz (almost 1 GHz higher than the stock NVIDIA GeForce RTX 5090) and the card itself was consuming around 1000W of power with the whole system’s power reaching almost 1800W.

Tony also shared some high-resolution and detailed die shots of the NVIDIA GB202 “Blackwell” GPU and the new GDDR7 memory featured on the GeForce RTX 5090 graphics card, which you can see in the gallery below:

NVIDIA RTX Blackwell GB202 GPU

What we have known so far about the NVIDIA RTX Blackwell Gaming GPUs is that they are based on TSMC’s 4nm process node, feature up to 92 billion transistors with 4000 AI TOPS, 380 RT TFLOPs, a 125 TFLOPS of FP32 compute, the fastest GDDR7 memory interface with up to 1.8 TB/s bandwidth and come with a brand new Founders Edition design.

GDDR7 also brings a much-needed upgrade over GDDR6/X memory, offering twice the bandwidth and data rate of G6 memory with higher frequency and lower wattage. GDDR7 also supports PAM4 signaling and the PCB materials used on RTX 50 GPUs are top-of-the-line from an engineering point of view.

The NVIDIA Blackwell GB202 GPU measures 922mm2, a 21% increase in die size on the same node. This is a substantially bigger die and explains why NVIDIA is charging a $2000 US premium for RTX 5090 graphics cards. There is still more headroom available for NVIDIA to do a Titan or Ti variant, & while we would definitely like to see one, we can’t say for sure if there will be a higher-end variant, as that didn’t happen with the RTX 4090, though the chances of a workstation variant with a bigger core config and more memory remains on the table.