AMD Radeon RX 6900 XT vs NVIDIA GeForce RTX 5070 Comparison

Head-to-Head Benchmarks

The benchmark data shows a clear overall winner in this pairing: the NVIDIA GeForce RTX 5070 takes 7 of 10 head-to-head tests, while the AMD Radeon RX 6900 XT wins only 3. The largest single margin belongs to the RTX 5070 in 3DMark Steel Nomad DX12, where it scores 5077 against the RX 6900 XT's 4079, a 19.7% advantage. That is a decisive lead in a modern DirectX 12 workload, and it sets the tone for the rest of the comparison.

The RTX 5070 also wins decisively in Geekbench Vulkan, scoring 178923 versus 148525, a 17% gap. This indicates that NVIDIA's Blackwell architecture handles the Vulkan API with notably more efficiency than AMD's RDNA 2.0 part. The pattern continues in legacy DirectX 9, where the RTX 5070 scores 320 against 268, a 16.2% margin. Even in Passmark G2D, a 2D graphics test, the RTX 5070 leads 1305 to 1055, a 19.2% difference, which suggests faster display and compositing throughput.

The RTX 5070's other wins are smaller but consistent. In Passmark DirectX 10, it scores 180 versus 167, a 7.2% edge. In Passmark G3D, the overall 3D graphics score, it leads 29137 to 26732, an 8.3% margin. The compute test, Passmark GPU Compute, also goes to NVIDIA, 15787 against 14547, a 7.9% advantage. These results show that the RTX 5070 is not just faster in one isolated workload; it holds a broad advantage across different graphics APIs and compute tasks.

The AMD Radeon RX 6900 XT does have its own victories, and they are worth examining. Its best win is in Geekbench OpenCL, where it scores 187673 against 172660, an 8.7% margin. This is a significant result for OpenCL compute workloads. The RX 6900 XT also wins both Passmark DirectX 11 and DirectX 12, though by narrow margins. In DirectX 11, it scores 279 versus 277, a 0.7% edge. In DirectX 12, it scores 113 versus 108, a 4.6% advantage. These are close results, but they show that AMD's older architecture still holds its own in certain API-specific scenarios.

Looking at the average benchmark scores provides a different perspective. The RX 6900 XT has an average benchmark score of 50951, while the RTX 5070 has an average of 40377. This is a surprising reversal, given the head-to-head results. The reason is that the RX 6900 XT has a much higher Geekbench OpenCL score, which pulls up its average. The RTX 5070's average is dragged down by its lower OpenCL score and by the fact that it lacks a Geekbench Metal result in the database. The percentile ranks tell a similar story: the RX 6900 XT sits at the 86th percentile among all GPUs, while the RTX 5070 sits at the 82nd percentile. This means that, across the entire database of recorded GPU results, the older AMD card ranks slightly higher overall, despite losing most head-to-head tests.

The nearest rivals for each card further contextualize these numbers. The RX 6900 XT's closest competitor is the NVIDIA CMP 50HX, which scores 51790 on average, a 1.6% difference in favor of the CMP 50HX. The AMD Radeon RX Vega 64 is 1.9% behind the RX 6900 XT, and the NVIDIA GeForce RTX 5070 Ti is 2% behind, while the Intel Arc A550M trails by 2.4%. For the RTX 5070, the nearest rival is the AMD Radeon Pro 580, which is just 0.1% behind, followed by the AMD Radeon Pro WX 7100 at 0.8% behind. The AMD Radeon Pro 5300 is 1.2% ahead, and the NVIDIA RTX A500 Mobile is 2% behind. These rival comparisons show that the RX 6900 XT sits in a higher overall performance tier in the database, while the RTX 5070 is clustered with professional workstation cards.

FAQ

Q: Which card wins more head-to-head benchmark tests?

A: The NVIDIA GeForce RTX 5070 wins 7 of 10 head-to-head tests, while the AMD Radeon RX 6900 XT wins 3.

Q: What is the biggest performance gap in the head-to-head results?

A: The largest margin is in 3DMark Steel Nomad DX12, where the RTX 5070 leads by 19.7% (5077 vs 4079).

Q: Does the AMD card win any compute benchmarks?

A: Yes, the RX 6900 XT wins Geekbench OpenCL by 8.7% (187673 vs 172660), and it also wins Passmark GPU Compute? No, actually the RTX 5070 wins Passmark GPU Compute by 7.9% (15787 vs 14547). The RX 6900 XT's only compute win is in OpenCL.

Q: How do the average benchmark scores compare?

A: The RX 6900 XT has a higher average benchmark score of 50951, compared to 40377 for the RTX 5070. This is influenced by the RX 6900 XT's strong OpenCL result and the RTX 5070's lack of a Metal benchmark entry.

Q: Which card ranks higher as a percentage of all GPUs?

A: The RX 6900 XT is at the 86th percentile, while the RTX 5070 is at the 82nd percentile.

Q: In which API does the RTX 5070 have its largest lead?

A: The RTX 5070's largest lead is in 3DMark Steel Nomad DX12 at 19.7%, followed closely by its 17% lead in Geekbench Vulkan and its 16.2% lead in Passmark DirectX 9.

The Verdict

The data points to the NVIDIA GeForce RTX 5070 as the stronger card for most modern workloads. It wins the 3DMark Steel Nomad DX12 test by a 19.7% margin, which is the most representative metric for current DirectX 12 gaming in this dataset. It also shows a 17% lead in Vulkan, another widely used API in modern games. For anyone building a system for contemporary titles, the benchmark results clearly favor the RTX 5070.

The AMD Radeon RX 6900 XT, on the other hand, has a specific strength in OpenCL compute, where it leads by 8.7%. It also edges out the RTX 5070 in DirectX 11 and DirectX 12 Passmark tests, though by small margins of 0.7% and 4.6% respectively. However, these wins are narrow and in less demanding or legacy scenarios. The RX 6900 XT's higher average benchmark score of 50951 versus 40377 is notable, but this is heavily skewed by its OpenCL result and the absence of a Metal score for the RTX 5070.

For gaming, the RTX 5070 is the clear choice based on its dominant wins in modern API benchmarks. For compute workloads that rely on OpenCL, the RX 6900 XT deserves consideration. The RX 6900 XT also has a higher percentile rank, at 86 versus 82, which reflects its standing across the entire database. But in a direct head-to-head comparison, the RTX 5070 wins more tests and wins them by larger margins in the most relevant categories.

Specification Differences

| Specification | AMD Radeon RX 6900 XT | NVIDIA GeForce RTX 5070 |

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

| Memory Size | 16 GB | 12 GB |

| Memory Type | GDDR6 | GDDR7 |

| Memory Bus Width | 256 bit | 192 bit |

| Memory Bandwidth | 512.0 GB/s | 672.0 GB/s |

| Shading Units | 5120 | 6144 |

| TMUs | 320 | 192 |

| ROPs | 128 | 80 |

| RT Cores | 80 | 48 |

| Tensor Cores | None | 192 |

| Base Clock | 1825 MHz | 2325 MHz |

| Boost Clock | 2250 MHz | 2512 MHz |

| Game Clock | 2015 MHz | Not specified |

| Memory Clock | 16 Gbps effective | 28 Gbps effective |

| FP32 Performance | 23.04 TFLOPS | 30.87 TFLOPS |

| FP16 Performance | 46.08 TFLOPS (2:1) | 30.87 TFLOPS (1:1) |

| Pixel Rate | 288.0 GPixel/s | 201.0 GPixel/s |

| Texture Rate | 720.0 GTexel/s | 482.3 GTexel/s |

| TDP | 300 W | 250 W |

| Slot Width | Triple-slot | Dual-slot |

| Power Connectors | 2x 8-pin | 1x 16-pin |

| Suggested PSU | 700 W | 600 W |

| Bus Interface | PCIe 4.0 x16 | PCIe 5.0 x16 |

| Display Outputs | 1x HDMI 2.1, 2x DisplayPort 1.4a, 1x USB Type-C | 1x HDMI 2.1b, 3x DisplayPort 2.1b |

| Dimensions (LxHxW) | 267 x 120 x 50 mm | 245 x 115 x 40 mm |

Architecture Differences

The architectural divide between these two cards is substantial. The RX 6900 XT is built on RDNA 2.0, uses a 7 nm process from TSMC, and has a die size of 520 mm² with 26,800 million transistors. The RTX 5070 uses the newer Blackwell 2.0 architecture on a 5 nm process, also from TSMC, with a much smaller die of 263 mm² that packs 31,100 million transistors. This gives the RTX 5070 a transistor density of 118.3M per mm², more than double the RX 6900 XT's 51.5M per mm².

The RTX 5070 has more shading units at 6144 versus 5120, and it includes 192 tensor cores, which the RX 6900 XT lacks entirely. The RTX 5070's RT core count is lower at 48 versus 80, but its ray tracing implementation is newer. The FP32 throughput favors the RTX 5070 at 30.87 TFLOPS versus 23.04 TFLOPS. However, the RX 6900 XT has a higher FP16 rate of 46.08 TFLOPS (2:1 ratio), while the RTX 5070 runs FP16 at the same rate as FP32, 30.87 TFLOPS (1:1 ratio).

Memory architecture differs significantly. The RX 6900 XT uses 16 GB of GDDR6 on a 256-bit bus with 512.0 GB/s bandwidth. The RTX 5070 uses 12 GB of GDDR7 on a 192-bit bus but achieves 672.0 GB/s bandwidth due to the higher 28 Gbps effective memory clock. The RTX 5070 also uses a PCIe 5.0 x16 interface, while the RX 6900 XT is limited to PCIe 4.0 x16. Display outputs are a notable difference: the RX 6900 XT provides HDMI 2.1 and DisplayPort 1.4a, plus a USB Type-C port, while the RTX 5070 offers HDMI 2.1b and three DisplayPort 2.1b outputs.

Where Each One Wins

The RTX 5070 wins in scenarios that stress modern graphics APIs and high-resolution rendering. Its 19.7% lead in 3DMark Steel Nomad DX12 is the strongest indicator of this, as that test is designed for current-generation hardware. It also wins in Vulkan by 17%, which covers a broad range of games and applications using that API. The RTX 5070's higher boost clock of 2512 MHz and higher FP32 throughput of 30.87 TFLOPS support its dominance in these compute-heavy tasks. Its 192 tensor cores also make it the better choice for any AI-accelerated workloads, though the dataset does not include specific tensor core benchmarks.

The RX 6900 XT wins in OpenCL compute, with an 8.7% margin. This is relevant for applications that rely on OpenCL for general-purpose GPU computing. It also wins in DirectX 11 and DirectX 12 Passmark tests, though by slim margins. The RX 6900 XT's larger 16 GB memory buffer and wider 256-bit bus may benefit workloads that need more capacity at the expense of raw bandwidth. Its higher pixel rate of 288.0 GPixel/s versus 201.0 GPixel/s suggests it may perform better in fill-rate-bound scenarios, though the RTX 5070's higher texture rate of 482.3 GTexel/s versus 720.0 GTexel/s is actually lower, so this is a mixed picture.

For users prioritizing modern gaming, the RTX 5070 is the winner. For users with specific OpenCL compute needs or those who value the higher percentile ranking and larger memory capacity of the RX 6900 XT, the AMD card retains relevance. The RTX 5070 is also more power-efficient with a 250 W TDP versus 300 W, and it is a dual-slot card versus the RX 6900 XT's triple-slot design, which may matter for smaller cases. The data overall favors the RTX 5070 for most buyers, but the RX 6900 XT has clear niches where it remains competitive.