RTX 5070 resembles an RTX 4070 Super with multi-frame generation capabilities.

RTX 5070 resembles an RTX 4070 Super with multi-frame generation capabilities.

      Nvidia’s RTX 50-series is now well established, offering high-end GPUs along with some more budget-friendly choices. If you're currently using a last-generation graphics card and contemplating whether to invest in a next-gen upgrade, it's important to determine if it’s worth the expense. With AMD putting pressure on the market, Nvidia needs to make its offerings compelling. Let's compare the RTX 5070 and the 4070 Super.

      **Pricing and Availability**

      The RTX 4070 Super was launched in January 2024 as a refresh of the RTX 4070, and it received positive reviews for delivering significant performance improvements while maintaining its $600 price tag. However, due to stock shortages in early 2025, the only available RTX 4070 Supers currently cost over $1,000, and even those are scarce.

      The RTX 5070 debuted in March 2025, but it quickly sold out like other RTX 50 graphics cards. Nvidia initially set its recommended retail price at $550, but very few cards have been sold at that price. When available, they typically range from $650 to $750, with some nearing $800. This is higher than the suggested retail price of the RTX 5070 Ti.

      **Specifications**

      | | Nvidia RTX 5070 | RTX 4070 Super |

      |-------------------|------------------|-----------------|

      | Graphics Cores | 6,144 | 7,168 |

      | RT Cores | 48, 4th generation| 56, 3rd generation|

      | Tensor Cores | 192, 5th generation| 224, 4th generation|

      | Boost clock | 2.51GHz | 2.46GHz |

      | Memory size | 12GB GDDR7 | 12GB GDDR6X |

      | Memory bus | 192-bit | 192-bit |

      | Memory speed | 28Gbps | 21Gbps |

      | Memory bandwidth | 672GBps | 504GBps |

      | TBP | 250W | 220W |

      The specifications of the RTX 5070 raised questions even before it was tested, as it appears to compromise in several areas compared to the previous generation. It features fewer CUDA, RT, and tensor cores, although it does utilize a new architecture with updated core designs. The clock speed is similar, with only a minor increase, and while the memory is quicker, it still has the same 12GB—considered limiting in 2025—with a constrained bus width. Memory bandwidth benefits from the enhanced GDDR7, but this comes with a 30W increase in power consumption despite the reduction in overall hardware.

      **Performance**

      During its reveal at CES 2025, Nvidia claimed the RTX 5070 would deliver "4090" performance at a price of $549. However, they cautioned that this was contingent on the use of AI, which may be more marketing exaggeration than reality.

      Nvidia's own benchmarks illustrated that the 5070 would double the performance of the 4070 in certain games when utilizing DLSS 4 and multi-frame generation, while the 4070 operated on DLSS 3.5 and single-frame generation. However, in scenarios where these enhancements are absent, the 5070 shows only about a 20% improvement over the regular 4070.

      Third-party evaluations found that the RTX 5070 often outperformed the 4070 Super by a mere few percentage points in games lacking multi-frame generation. Since the 4070 Super can also utilize DLSS 4, the real advantage for the 5070 arises mainly from the additional AI-generated frames.

      It’s clear why Nvidia only compared the 5070 with the standard 4070 rather than its faster counterpart. Ray tracing performance has marginally increased with this generation due to the new core design, but the lower core count is a disadvantage. Furthermore, multi-frame generation's effectiveness is limited as it thrives on high native frame rates, which the 5070 struggles to achieve compared to the higher-end 50-series models.

      While the RTX 5070 is technically the faster card, the difference is minimal, rendering its hefty price unreasonable. Paying over $700 for a card that does not substantially outperform a $600 model from over a year ago based on an older architecture certainly doesn’t qualify as a success.

      For context, the RTX 4090 significantly outperforms the RTX 5080 in all gaming tests. Nvidia's assertion that a 5070 can match the performance of a 4090 is simply not accurate; it barely manages to outperform the 4070 Super.

      **Recommendation**

      Nvidia has placed its customers in a difficult situation. With the 4070 Super no longer available and the RTX 5070 priced excessively, consumers wanting

RTX 5070 resembles an RTX 4070 Super with multi-frame generation capabilities. RTX 5070 resembles an RTX 4070 Super with multi-frame generation capabilities. RTX 5070 resembles an RTX 4070 Super with multi-frame generation capabilities. RTX 5070 resembles an RTX 4070 Super with multi-frame generation capabilities. RTX 5070 resembles an RTX 4070 Super with multi-frame generation capabilities.

Other articles

RTX 5070 resembles an RTX 4070 Super with multi-frame generation capabilities.

The Nvidia RTX 4070 Super is difficult to locate, and the 5070 isn’t particularly speedy. Are either of these GPUs worth considering in 2025?