AMD Radeon AI PRO R9700 vs AMD Radeon RX 590 Comparison

The AMD Radeon RX 590 and the AMD Radeon AI PRO R9700 represent two distinct eras of graphics hardware, separated by nearly seven years of architectural evolution. The data shows a generational chasm: the R9700 delivers roughly 6.5 times the raw 3D performance of the RX 590 in the sole shared benchmark, while also introducing features and memory capacities that the older card cannot approach. The RX 590, however, remains a relevant baseline for legacy workloads, and its benchmark percentiles are surprisingly close to its modern counterpart, though for very different reasons.

Head-to-Head Benchmarks

The only directly comparable data point is the 3DMark Steel Nomad DX12 test. In this demanding, modern API workload, the AMD Radeon AI PRO R9700 scores 7006, while the AMD Radeon RX 590 manages 1081. This represents a -84.6% delta from the R9700’s perspective, meaning the R9700 is approximately 6.5 times faster. This is not an incremental improvement; it is a complete overhaul in rasterization throughput, fill rates, and raw compute power.

Looking at the broader benchmark averages, the two cards are nearly tied. The RX 590 has an average benchmark score of 24744, while the R9700 sits at 23315. This near-parity is deceptive. The RX 590’s score is buoyed by its two legacy tests (Geekbench Metal at 48407 and the Steel Nomad run), whereas the R9700’s average is dragged down by several Passmark DirectX 9/10/11 tests where it scores in the low hundreds (363, 155, 272 respectively). The R9700’s strength lies in modern compute and high-fidelity workloads: it posts a Geekbench OpenCL score of 118202 and a Vulkan score of 63550, figures that dwarf anything the RX 590 can produce. The data suggests the R9700 is optimized for contemporary APIs, while the RX 590 holds its own in older, less demanding test suites.

In terms of percentile ranking against all GPUs, the RX 590 sits at the 70th percentile, slightly ahead of the R9700’s 69th percentile. This is a statistical artifact; the RX 590’s nearest rivals include the NVIDIA RTX A5000 Mobile (24763, delta -0.1%) and Intel Arc A350M (24647, delta 0.4%), indicating it is a solid mid-range performer. The R9700’s nearest rivals are the AMD Radeon R9 M290X (23276, delta 0.2%) and RX 6600M (23273, delta 0.2%), which places it in a similar average bracket, but its peak performance in modern tests is dramatically higher than those older competitors.

Architecture Differences

The foundational difference is the architecture generation. The RX 590 is built on GCN 4.0 (chip Polaris 30), a design from the mid-2010s. The R9700 uses RDNA 4.0 (chip Navi 48), the latest generation. This is reflected in their process nodes: the RX 590 uses a 12 nm process at GlobalFoundries, while the R9700 uses a 4 nm process at TSMC. This node shrink allows the R9700 to pack 53,900 million transistors into a 357 mm² die, achieving a density of 151.0M / mm². The RX 590 has only 5,700 million transistors on a 232 mm² die, with a density of 24.6M / mm². The R9700 has nearly 10 times the transistors.

Compute resources are similarly lopsided. The R9700 features 4096 shading units, 256 TMUs, and 128 ROPs. The RX 590 has 2304 shading units, 144 TMUs, and only 32 ROPs. Critically, the R9700 is the first card here to include 64 dedicated ray tracing cores, an entire feature class absent from the RX 590. The R9700 also supports DirectX 12 Ultimate (12_2), while the RX 590 is limited to DirectX 12 (12_0). The R9700’s pixel rate is 373.8 GPixel/s versus 49.44 GPixel/s for the RX 590, and its texture rate is 747.5 GTexel/s versus 222.5 GTexel/s. FP32 compute is 47.84 TFLOPS for the R9700, a massive leap over the RX 590’s 7.119 TFLOPS.

Memory architecture is another clear divider. The R9700 uses 32 GB of GDDR6 across a 256-bit bus, delivering 644.6 GB/s of bandwidth. The RX 590 uses 8 GB of GDDR5 on the same bus width, yielding only 256.0 GB/s. The R9700’s memory clock is 2518 MHz (20.1 Gbps effective), compared to 2000 MHz (8 Gbps effective) for the RX 590. The R9700 also supports PCIe 5.0 x16, while the RX 590 is limited to PCIe 3.0 x16. Display outputs differ as well: the R9700 offers HDMI 2.1b and DisplayPort 2.1a, while the RX 590 has HDMI 2.0b and DisplayPort 1.4a.

FAQ

Q: Which card is faster in modern DirectX 12 workloads?

A: The AMD Radeon AI PRO R9700 is overwhelmingly faster. In the 3DMark Steel Nomad DX12 test, it scores 7006 against the RX 590’s 1081, representing a -84.6% delta in favor of the R9700.

Q: Do both cards support the same feature set for ray tracing?

A: No. The R9700 has 64 dedicated ray tracing cores and supports DirectX 12 Ultimate (12_2). The RX 590 has no ray tracing cores and is limited to DirectX 12 (12_0).

Q: How does memory capacity and bandwidth compare?

A: The R9700 has 32 GB of GDDR6 memory with 644.6 GB/s bandwidth. The RX 590 has 8 GB of GDDR5 with 256.0 GB/s bandwidth. The R9700 also uses a faster 20.1 Gbps effective memory clock versus 8 Gbps on the RX 590.

Q: Are the average benchmark scores similar?

A: Yes, the overall averages are close: the RX 590 averages 24744 and the R9700 averages 23315. However, this is misleading. The R9700’s average is suppressed by low scores in legacy Passmark DirectX 9/10 tests, while its modern compute scores (Geekbench OpenCL 118202) are far higher than anything the RX 590 achieves.

Q: What is the transistor count difference?

A: The R9700 contains 53,900 million transistors on a 357 mm² die, while the RX 590 contains 5,700 million on a 232 mm² die. The R9700’s density is 151.0M / mm² versus 24.6M / mm².

Q: Which card is more recent and what is its production status?

A: The R9700 was released on 2025-07-22 and is Active in production. The RX 590 was released on 2018-11-14 and is End-of-life.

Specification Differences

| Specification | AMD Radeon RX 590 | AMD Radeon AI PRO R9700 |

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

| Chip | Polaris 30 | Navi 48 |

| Architecture | GCN 4.0 | RDNA 4.0 |

| Process Node | 12 nm | 4 nm |

| Foundry | GlobalFoundries | TSMC |

| Transistors | 5,700 million | 53,900 million |

| Die Size | 232 mm² | 357 mm² |

| Transistor Density | 24.6M / mm² | 151.0M / mm² |

| Base Clock | 1469 MHz | 1660 MHz |

| Boost Clock | 1545 MHz | 2920 MHz |

| Game Clock | None | 2350 MHz |

| Memory Clock | 2000 MHz (8 Gbps) | 2518 MHz (20.1 Gbps) |

| Memory Size | 8 GB | 32 GB |

| Memory Type | GDDR5 | GDDR6 |

| Memory Bus | 256 bit | 256 bit |

| Bandwidth | 256.0 GB/s | 644.6 GB/s |

| Shading Units | 2304 | 4096 |

| TMUs | 144 | 256 |

| ROPs | 32 | 128 |

| RT Cores | None | 64 |

| Pixel Rate | 49.44 GPixel/s | 373.8 GPixel/s |

| Texture Rate | 222.5 GTexel/s | 747.5 GTexel/s |

| FP32 | 7.119 TFLOPS | 47.84 TFLOPS |

| TDP | 175 W | 300 W |

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

| Suggested PSU | 450 W | 700 W |

| Bus Interface | PCIe 3.0 x16 | PCIe 5.0 x16 |

| Display Outputs | 1x HDMI 2.0b, 3x DP 1.4a | 1x HDMI 2.1b, 3x DP 2.1a |

| DirectX | 12 (12_0) | 12 Ultimate (12_2) |

| Vulkan | 1.3 | 1.4 |

| Length | 241 mm (9.5 inches) | 267 mm (10.5 inches) |

| Height | None | 111 mm (4.4 inches) |

| Width | None | 40 mm (1.6 inches) |

| Release Date | 2018-11-14 | 2025-07-22 |

| Production Status | End-of-life | Active |

Where Each One Wins

The AMD Radeon AI PRO R9700 wins decisively in every modern, compute-intensive scenario. Its 47.84 TFLOPS FP32 performance and 64 ray tracing cores make it the clear choice for real-time ray tracing, AI inference, and professional 3D rendering workloads that leverage DirectX 12 Ultimate. The 32 GB memory buffer is essential for large datasets, high-resolution textures, and massive compute workloads, far exceeding the RX 590’s 8 GB capacity. The data confirms this with the R9700’s dominant 7006 score in Steel Nomad DX12 and its 118202 OpenCL score. This card is built for the future, with PCIe 5.0 and DisplayPort 2.1a ensuring it is not a bottleneck for next-generation systems.

The AMD Radeon RX 590 wins in legacy compatibility and power efficiency. Its 175 W TDP is significantly lower than the R9700’s 300 W, and it requires only a 450 W PSU versus 700 W. For older games and applications using DirectX 9/10/11, the RX 590’s performance is more balanced; the R9700’s Passmark scores in those tests are anomalously low (363 for DX9, 155 for DX10, 272 for DX11), suggesting driver optimization is focused on modern APIs. The RX 590 also has a smaller physical footprint at 241 mm length versus 267 mm for the R9700. Its 70th percentile ranking, slightly higher than the R9700’s 69th, indicates it remains a surprisingly capable card for its age within the broader database, particularly in Metal-based workloads where it scores 48407. The RX 590 is the pragmatic choice for a system that prioritizes lower power draw and compatibility with a wide range of older software, while the R9700 is the uncompromising performer for current-generation tasks.