After a long but worth-it wait, Nvidia has announced the chipmaker's spanking new GeForce RTX 30-series (codename Ampere) graphics cards. The GeForce RTX 3090 and GeForce RTX 3080 will be available on September 24 and 17, respectively, with the GeForce RTX 3070 coming at a later date in October. These three cards boast impressive specs that will vie for our Best Graphics Cards for Gaming list.
Built on a custom Samsung 8nm process, Ampere comes equipped with Nvidia's second-generation Ray Tracing cores and third-generation Tensor cores. The GeForce RTX 3090, RTX 3080 and RTX 3070 are also the first Nvidia consumer graphics cards to come with PCIe 4.0 support. We have the architectural deep dive details here.
The GeForce RTX 3090 is the behemoth of the Ampere lineup. The triple-slot graphics card measures 12.3 x 5.4 inches (313 x 138mm) and flaunts specifications would impress even the most demanding enthusiasts. In terms of performance, Nvidia claims that the GeForce RTX 3090 is up to 50% faster than the Titan RTX.
The GeForce RTX 3090 is aimed at 8K gaming at 60 frames per second, so it comes equipped with 10,496 CUDA cores that feature a boost clock up to 1.7 GHz. There's also 24GB of 19.5 Gbps GDDR6X memory across a 384-bit memory interface. The graphics card has a 350W TDP (thermal design power) and requires two 8-pin PCIe power connectors.
Professional and hardcore enthusiasts will be delighted to know that the GeForce RTX 3090 is the only Ampere-based graphics card to support SLI through Nvidia's NVLink connector. This opens the door to pairing up two of these beasts together for an awesome compute machine. The GeForce RTX NVLink Bridge costs $79.99 and will be available on the same day as the GeForce RTX 3090.
Nvidia Ampere / RTX 30-Series Specifications
| Header Cell - Column 0 | GeForce RTX 3090 | GeForce RTX 3080 | GeForce RTX 3070 | Titan RTX | GeForce RTX 2080 Ti |
|---|---|---|---|---|---|
| Architecture (GPU) | Ampere (GA102) | Ampere (GA102) | Ampere (GA104)* | Turing (TU102) | Turing (TU102) |
| CUDA Cores | 10,496 | 8,704 | 5,888 | 4,608 | 4,352 |
| RT Cores | 82 | 68 | 46 | 72 | 68 |
| Tensor Cores | 328 | 272 | 184 | 576 | 544 |
| Texture Units | 328 | 272 | 184 | 288 | 272 |
| Base Clock Rate | 1,400 MHz | 1,440 MHz | 1,500 MHz | 1,350 MHz | 1,350 MHz |
| Boost Clock Rate | 1,700 MHz | 1,710 MHz | 1,730 MHz | 1,770 MHz | 1,545 MHz |
| Memory Capacity | 24GB GDDR6X | 10GB GDDR6X | 8GB GDDR6 | 24GB GDDR6 | 11GB GDDR6 |
| Memory Speed | 19.5 Gbps | 19 Gbps | 14 Gbps | 14 Gbps | 14 Gbps |
| Memory Bus | 384-bit | 320-bit | 256-bit | 384-bit | 352-bit |
| Memory Bandwidth | 935.8 GBps | 760 GBps | 448 GBps | 672 GBps | 616 GBps |
| ROPs | 96 | 88 | 64 | 96 | 88 |
| L2 Cache | 6MB | 5MB | 4MB | 6MB | 5.5MB |
| TDP | 350W | 320W | 220W | 280W | 250W |
| Transistor Count | ? | ? | ? | 18.6 billion | 18.6 billion |
| Die Size | ? | ? | ? | 754 mm² | 754 mm² |
| MSRP | $1,499 | $699 | $499 | $2,499 | $999 |
*Some specifications are unconfirmed.
Nvidia touts the GeForce RTX 3080 as the flagship Ampere SKU. The chipmaker is promising up to double the performance over the previous GeForce RTX 2080. According to Nvidia, the GeForce RTX 3080 is capable of providing a perfect 60 frames per second gaming experience at 4K even with ray tracing enabled.
The GeForce RTX 3080 comes in at 11.2 x 4.4 inches (285 x 112mm), sporting up to 8,704 CUDA cores that top out at 1.71GHz. The 10GB of GDDR6X memory communicates via a 320-bit memory bus. Nvidia rates the GeForce RTX 3080 for 320W, so like the GeForce RTX 3090, the graphics card still depends on a pair of 8-pin PCIe power connectors. Head to our Nvidia GeForce RTX 3080 Everything We Know post for more details.
Ultimately, the GeForce RTX 3070 will continue to be the sweet spot for gamers. The graphics card starts at $499 and delivers higher performance than last generation's flagship GeForce RTX 2080 Ti at half the price. We have the deep-dive details in our Nvidia GeForce RTX 3070: RTX 2080 Ti Performance at $499 article.
The GeForce RTX 3070 arrives with dimensions of 9.5 x 4.4 inches (242 x 112mm) and packing 5,888 CUDA cores and 8GB of GDDR6 memory. The maximum boost clock on this model is 1.73 GHz, while the memory works with a 256-bit memory interface. The GeForce RTX 3070 has a much more reasonable TDP (220W) so it only requires one 8-pin PCIe power connector to operate.
Regardless of the model, the Ampere-based graphics cards offer three DisplayPort 1.4a outputs and one HDMI 2.1 port. Gone is the VirtualLink port that debuted with Turing. It doesn't come as a huge surprise since the standard never really caught on.
Given the power requirements, Nvidia recommends a 750W power supply for the GeForce RTX 3090 and GeForce RTX 3080, while the GeForce RTX 3070 can get by with a 650W unit. Nvidia's recommendations are based around a high-end system with the Intel Core i9-10900K so you could get away with a power supply that has a lower capacity than the suggested.
