NVIDIA has officially confirmed that its highly anticipated GeForce RTX 4070 SUPER GPU will come equipped with a full 48 MB L2 cache on the AD104 chip. This confirmation rectifies an earlier error in the specifications sheet, which initially stated the L2 cache as 36 MB. The correction aligns the RTX 4070 SUPER’s cache specifications with the RTX 4070 Ti GPU.
The GeForce RTX 4070 SUPER, unveiled just last week by NVIDIA, has been generating buzz for its promising features aimed at elevating gaming experiences to new heights. The revelation about the 48 MB L2 cache underscores the company’s commitment to delivering cutting-edge technology and performance in the graphics card market.
The significance of this update lies in the crucial role played by the L2 cache in enhancing the GPU’s capabilities. Specifically designed for high-refresh-rate 1080p and 1440p gaming PCs, the RTX 4070 SUPER’s larger L2 cache becomes a key asset. With a 192-bit memory bus interface and advanced G6X memory, the GPU is poised to handle the demands of modern gaming, including higher-resolution textures and increased overall resolution.
The substantial L2 cache near the GPU serves as a high-speed interconnect across all GPCs, mitigating potential bottlenecks within the memory subsystem. This optimization is particularly crucial in scenarios where memory strain could lead to slow-downs during gaming sessions. By incorporating a 48 MB L2 cache, NVIDIA ensures that the GeForce RTX 4070 SUPER is well-equipped to deliver a seamless and immersive gaming experience.
The decision to provide 48 MB of L2 cache in the RTX 4070 SUPER is being hailed as a commendable move, marking a significant improvement over its Non-SUPER variant. This enhancement reinforces NVIDIA’s commitment to pushing the boundaries of GPU performance and delivering top-notch products to its user base.
Delving into the specifications of the RTX 4070 SUPER, it features an upgraded AD104 GPU with a 20% increase in core count, boasting configurations with either AD104-350 or AD103-175 GPU. Both variants maintain 7,168 cores, ensuring a robust foundation for high-performance gaming. The GPU retains a 12 GB GDDR6X memory with a 192-bit bus interface. Notably, the TBP for the RTX 4070 SUPER is set at 220W, indicating a 20W increase over the existing 4070 model.
In terms of performance, the RTX 4070 SUPER stands out, surpassing the RTX 4070 Non-SUPER by approximately 15%. This improvement marks the most substantial uplift in the SUPER lineup. Impressively, the GPU also outperforms an RTX 3090 in gaming performance, with only slight variations depending on the specific game. NVIDIA has confirmed that the RTX 4070 SUPER is just 5% slower than its predecessor, the RTX 4070 Ti, which originally retailed for $799 US.
AI capabilities are a significant talking point for the RTX 4070 SUPER, promising up to 2x faster performance with DLSS 3.5. Additionally, the GPU demonstrates a 70% faster image generation and 50% faster video generation speeds, thanks to ML optimizations through the Tensor-RT and Tensor-RT LLM stack, available on Windows and continuously updated.
Another noteworthy aspect is the outstanding power efficiency of the RTX 4070 SUPER. Despite being rated at 220W, the actual gaming wattage averages around 200W, demonstrating remarkable performance per watt. This efficiency surpasses competitors such as the RTX 4090 and the Radeon RX 7900 series.
The much-anticipated NVIDIA GeForce RTX 4070 SUPER GPU is scheduled for release on January 17th in both Founders Edition and custom variants, with an enticing starting price of $599 US. As the gaming community eagerly awaits its arrival, the confirmed 48 MB L2 cache adds another layer of anticipation and excitement to what promises to be a game-changing graphics card in the world of PC gaming.