Nvidia’s next-gen RTX 3090, RTX 3080 and RTX 3070 GPUs revealed

Nvidia have pulled back the curtain to unveil their next generation of high-end GPUs, promising increased performance at the company’s price levels improved ability to process taxing ray-tracing effect, and improved power efficiency. However, it will set you back a pretty penny.

Across the board, they’re leaning on established Nvidia technologies like Deep Learning Super Sampling for resolution upscaling and enhancement, and the Tensor cores from the 20 series cards to assist in processing the complex ray traced lighting effects. They’re combining these with new innovations, such as faster GDDR6X RAM and PCIe 4.0 connectivity, as well as having on-GPU decompression that collaborates with Microsoft on DirectStorage. As in PlayStation 5 and Xbox Series X, this will allow the GPU to directly communicate with ultra-fast SSDs to lessen the CPU load and speed up loading.

Nvidia GeForce RTX 3080

The RTX 3080 is the company’s new flagship GPU. Priced at $699, it’s a replacement for the current RTX 2080 Super, sporting double the performance of the older RTX 2080 with ray tracing, and easily surpassing the RTX 2080 Ti, the current ultra high-end halo product from the current era of Nvidia cards, and boasting single precision performance over two times the Xbox Series X – this isn’t actually a good measure of GPU power, it should be said. It’s expected to handle 4K at 60FPS with ray tracing.

• CUDA Cores – 8704
• Boost Clock – 1.71Ghz
• VRAM – 10GB
• Single Precision Performance – 29.8 TFLOPs
• Tensor Performance FP16 – 238 TFLOPs
• Ray Tracing Performance – 58 TFLOPs
• TFP – 320W
• Release date – 17th September 2020
• Price – $699

Nvidia GeForce RTX 3090

Speaking of ultra high-end, that crown now goes to the RTX 3090. This could have fallen into the company’s Titan range, which pushes their GPU chipsets to the limit, but is instead part of the GeForce line up. It features a huge 24GB of VRAM, and goes close to three times the TFLOPs of the Xbox Series X. It’s got an eye-watering price point to match: $1499… more than double the RTX 3080.

• CUDA Cores – 10496
• Boost Clock – 1.7Ghz
• VRAM – 24GB
• Single Precision Performance – 35.7 TFLOPs
• Tensor Performance FP16 – 285 TFLOPs
• Ray Tracing Performance – 69 TFLOPs
• TFP – 350W
• Release date – 24th September 2020
• Price – $1499

Nvidia GeForce RTX 3070

And finally, we come to what will actually be most likely to end up in a gaming PC. The RTX 3070 will launch in October with a $499 price, and it’s impressively well specced for the price (in the context of recent Nvidia GPU prices, at least). It cuts things back significantly from the 3080, but is still pitched as being slightly more powerful than the RTX 2080 Ti, and having the greatest generational leap in performance at this price point.

• CUDA Cores – 5888
• Boost Clock – 1.73Ghz
• VRAM – 8GB
• Single Precision Performance – 20.4 TFLOPs
• Tensor Performance FP16 – 163 TFLOPs
• Ray Tracing Performance – 40 TFLOPs
• TFP – 220W
• Release date – October 2020
• Price – $499

It’s all impressive sounding, though we’ll have to wait until reviews drop to see how they actually perform and how they can handle the ever-increasing demands that ray tracing will put on gaming machines. However, when the cheapest of the cards announced today is also at the price point many pundits expect Sony and Microsoft to pick for their new consoles, these are very much high-end luxury gaming items.

Nvidia have had great success at these price points over the last few years though, and these are huge steps forward in performance. The real question is if AMD can respond with their own next-generation of GPU. The company have struggled to compete with Nvidia’s halo products for years, and it will take a lot for them to stack up against the 30 series in general with their upcoming RDNA 2 products.

via Anandtech