AMD Radeon RX 570 vs AMD Radeon RX 9070 Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive sweep for the AMD Radeon RX 9070 across all three shared benchmark tests. In 3DMark Steel Nomad DX12, the RX 9070 scores 6,290, while the RX 570 manages only 880. That is a delta of -86% relative to the newer card, meaning the RX 9070 delivers roughly 7.1 times the raw throughput in this DirectX 12 workload. This is the largest relative gap in the head-to-head set, and it reflects more than just clock speed differences; the architectural gulf between the two generations is fully exposed here.

Geekbench OpenCL shows a similar pattern, though slightly less extreme. The RX 9070 scores 131,539 against the RX 570’s 32,084, a delta of -75.6%. OpenCL is a compute-heavy benchmark, so the RX 9070’s 36.13 TFLOPS FP32 rating versus the RX 570’s 5.095 TFLOPS is the primary driver. The RX 9070 is 7.1 times faster in raw FP32, and the OpenCL score difference of 4.1 times suggests memory bandwidth and driver overhead also play roles, since the RX 9070’s 644.6 GB/s bandwidth is 2.9 times the RX 570’s 224.0 GB/s.

Geekbench Vulkan narrows the gap considerably. The RX 9070 scores 58,705, while the RX 570 posts 42,752, a delta of -27.2%. This is the closest contest in the dataset. Vulkan’s lower-level API reduces CPU overhead and allows the RX 570’s older GCN architecture to stretch its legs, but the RX 9070 still wins by a clear margin. The RX 570’s Vulkan score is actually higher than its OpenCL score, which is unusual and suggests that the RX 570’s GCN 4.0 design handles Vulkan’s explicit command queues better than OpenCL’s more abstract dispatch model.

The average benchmark score across all recorded tests tells a different story. The RX 570 has an average score of 28,766, while the RX 9070 averages 23,877. That means the RX 570’s average is 4,889 points higher, roughly 20.5% above the RX 9070’s average. This inversion is due to the RX 9070’s low scores in PassMark DirectX 9, 10, 11, and 12 tests, which range from 74 to 343, while the RX 570’s benchmark set includes no such legacy tests. The RX 9070’s PassMark G3D score of 25,381 and GPU compute score of 14,737 are substantial, but the legacy DX9 and DX10 scores drag its average down. The RX 570’s percentile rank of 74 versus the RX 9070’s 69 also reflects this average-score anomaly, though the head-to-head modern tests clearly favor the RX 9070.

Architecture Differences

The two GPUs come from entirely different design philosophies. The RX 570 uses the Polaris 20 chip on GCN 4.0 architecture, fabricated on a 14 nm process at GlobalFoundries. The RX 9070 uses the Navi 48 chip on RDNA 4.0, fabricated on a 4 nm process at TSMC. The process node shrink from 14 nm to 4 nm is the most fundamental difference, enabling the RX 9070 to pack 53,900 million transistors into a 357 mm² die, compared to the RX 570’s 5,700 million transistors on a 232 mm² die. Transistor density jumps from 24.6M per mm² to 151.0M per mm², a 6.1 times increase.

The shader configuration diverges sharply. The RX 570 has 2,048 shading units, 128 texture mapping units, and 32 raster operation units. The RX 9070 has 3,584 shading units, 224 TMUs, and 128 ROPs. The ROP count quadruples, which directly explains the pixel rate difference: 322.6 GPixel/s for the RX 9070 versus 39.81 GPixel/s for the RX 570. The texture rate also increases from 159.2 GTexel/s to 564.5 GTexel/s, a 3.5 times improvement that tracks with the TMU count increase.

The RX 9070 introduces hardware ray tracing cores, with 56 RT cores present, while the RX 570 has none. This is a generational feature addition, not just a performance improvement. The RX 9070 also supports DirectX 12 Ultimate (12_2), while the RX 570 is limited to DirectX 12 (12_0). Vulkan support advances from 1.3 to 1.4, and the display outputs shift from 1x DVI, 1x HDMI 2.0b, 3x DisplayPort 1.4a to 1x HDMI 2.1b, 3x DisplayPort 2.1a.

Memory architecture changes are equally stark. The RX 570 uses 4 GB of GDDR5 on a 256-bit bus, yielding 224.0 GB/s bandwidth. The RX 9070 uses 16 GB of GDDR6 on the same 256-bit bus, but with memory clocks of 2518 MHz and 20.1 Gbps effective, bandwidth reaches 644.6 GB/s. The fourfold capacity increase is matched by a 2.9 times bandwidth increase. The RX 9070’s base clock is 1330 MHz with a boost of 2520 MHz and a game clock of 2070 MHz, while the RX 570’s base is 1168 MHz and boost is 1244 MHz. The boost clock difference of 1,276 MHz is substantial, but the RX 9070’s higher clock rates are only possible due to the 4 nm process’s efficiency.

Power requirements scale accordingly. The RX 570 has a TDP of 150 W with a 1x 6-pin connector and a 450 W suggested PSU. The RX 9070 has a TDP of 220 W, requires 2x 8-pin connectors, and suggests a 550 W PSU. The RX 570 uses PCIe 3.0 x16, while the RX 9070 uses PCIe 5.0 x16. The bus interface upgrade doubles the available bandwidth per lane, though the practical impact depends on workload.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The RX 570 has an average benchmark score of 28,766, which is 4,889 points higher than the RX 9070’s 23,877. However, this average includes legacy DirectX 9 and 10 tests where the RX 9070 scores very low, so the modern benchmark results tell a different story.

Q: How much faster is the RX 9070 in 3DMark Steel Nomad DX12?

A: The RX 9070 scores 6,290 versus the RX 570’s 880, a delta of -86%. This means the RX 9070 is approximately 7.1 times faster in this specific DirectX 12 test.

Q: Does the RX 9070 support ray tracing?

A: Yes, the RX 9070 has 56 dedicated ray tracing cores. The RX 570 has no ray tracing cores at all, as its GCN 4.0 architecture predates hardware ray tracing support.

Q: What is the memory capacity difference?

A: The RX 570 has 4 GB of GDDR5 memory, while the RX 9070 has 16 GB of GDDR6 memory. Both use a 256-bit memory bus, but the RX 9070’s memory clock of 2518 MHz provides 644.6 GB/s bandwidth versus 224.0 GB/s for the RX 570.

Q: Why does the RX 9070 have a lower percentile rank?

A: The RX 9070 ranks in the 69th percentile of all GPUs, while the RX 570 ranks in the 74th. This is based on average benchmark scores, which are dragged down by the RX 9070’s very low PassMark DirectX 9 score of 343 and DirectX 10 score of 141. The RX 570’s benchmark set does not include these legacy tests.

Q: What is the transistor density difference?

A: The RX 9070 has a transistor density of 151.0M per mm², compared to 24.6M per mm² for the RX 570. This 6.1 times density increase comes from the shift from 14 nm GlobalFoundries production to 4 nm TSMC production.

Specification Differences

The two GPUs differ in nearly every measurable specification. The process node changes from 14 nm to 4 nm, and the foundry shifts from GlobalFoundries to TSMC. Transistor count rises from 5,700 million to 53,900 million, while die size increases from 232 mm² to 357 mm². The architecture changes from GCN 4.0 to RDNA 4.0, and the chip from Polaris 20 to Navi 48.

Clock speeds differ across the board. Base clock rises from 1168 MHz to 1330 MHz, boost clock from 1244 MHz to 2520 MHz. The RX 9070 also has a game clock of 2070 MHz, which the RX 570 lacks. Memory clock jumps from 1750 MHz (7 Gbps effective) to 2518 MHz (20.1 Gbps effective).

Memory configuration changes from 4 GB GDDR5 to 16 GB GDDR6. Bandwidth increases from 224.0 GB/s to 644.6 GB/s. Shading units increase from 2,048 to 3,584, TMUs from 128 to 224, and ROPs from 32 to 128. The RX 9070 adds 56 RT cores, which the RX 570 does not have.

Pixel rate increases from 39.81 GPixel/s to 322.6 GPixel/s. Texture rate increases from 159.2 GTexel/s to 564.5 GTexel/s. FP32 compute rises from 5.095 TFLOPS to 36.13 TFLOPS. FP16 performance is listed as 1:1 with FP32 for both, but the absolute numbers scale identically.

TDP rises from 150 W to 220 W. Power connectors change from 1x 6-pin to 2x 8-pin, and suggested PSU from 450 W to 550 W. Bus interface upgrades from PCIe 3.0 x16 to PCIe 5.0 x16. Display outputs change from 1x DVI, 1x HDMI 2.0b, 3x DisplayPort 1.4a to 1x HDMI 2.1b, 3x DisplayPort 2.1a.

DirectX support moves from 12 (12_0) to 12 Ultimate (12_2). Vulkan support moves from 1.3 to 1.4. OpenGL remains at 4.6 for both. Production status changes from end-of-life to active. Release dates are 2017-04-17 for the RX 570 and 2025-03-05 for the RX 9070. The RX 570’s successor is listed as Vega, while the RX 9070 has no successor. The RX 570’s predecessor is Arctic Islands, and the RX 9070’s is Navi III.

The Verdict

The data presents a clear generational split. For any modern workload, the RX 9070 is categorically superior. In the three head-to-head benchmarks, it wins all three, with the smallest margin being 27.2% in Geekbench Vulkan. The RX 9070’s 36.13 TFLOPS FP32 performance, 644.6 GB/s bandwidth, and 56 RT cores make it the only viable choice for contemporary DirectX 12 Ultimate games or ray-traced content.

The RX 570’s higher average benchmark score and better percentile rank are artifacts of the benchmark set composition. The RX 9070’s PassMark DirectX 9 score of 343 and DirectX 10 score of 141 are anomalously low, likely reflecting driver prioritization of modern APIs. The RX 570’s benchmark suite includes Geekbench Metal, which the RX 9070 does not have, and the Metal score of 39,349 contributes to the RX 570’s higher average.

For legacy DirectX 9 and DirectX 10 applications, the RX 9070’s low scores (343 and 141 respectively) suggest it does not prioritize these antiquated APIs. The RX 570, with its GCN 4.0 architecture from 2017, was designed when these APIs were more relevant. However, the RX 9070’s PassMark DirectX 11 score of 281 and DirectX 12 score of 74 are also low, which is puzzling given its strong 3DMark Steel Nomad result. The PassMark suite may not be optimized for RDNA 4.0’s scheduling.

The production status difference is decisive for long-term use. The RX 570 is end-of-life, while the RX 9070 is active. The RX 9070’s launch MSRP is 549 USD, and the RX 570’s was 169 USD. The price difference is substantial, but the performance delta in modern workloads is even larger.

Where Each One Wins

The RX 9070 wins in every modern benchmark category. In 3DMark Steel Nomad DX12, it is 86% faster. In Geekbench OpenCL, it is 75.6% faster. In Geekbench Vulkan, it is 27.2% faster. Its 16 GB GDDR6 memory provides four times the capacity and 2.9 times the bandwidth, which matters for high-resolution textures and compute workloads. Its 128 ROPs enable 322.6 GPixel/s fill rates, suitable for high-resolution rendering with multiple displays.

The RX 570 wins in the average benchmark score comparison, with 28,766 versus 23,877. Its percentile rank of 74 exceeds the RX 9070’s 69. The RX 570’s Geekbench Metal score of 39,349 is a benchmark the RX 9070 does not participate in, and its Geekbench Vulkan score of 42,752 is respectable for its era. The RX 570 also has a smaller die (232 mm²) and lower TDP (150 W), which could matter for constrained chassis.

The RX 9070’s 56 RT cores provide hardware ray tracing capability that the RX 570 completely lacks. Its DirectX 12 Ultimate support enables mesh shaders and variable rate shading, features unavailable on the RX 570’s DirectX 12 (12_0) implementation. The RX 9070’s PCIe 5.0 x16 interface offers greater bandwidth headroom than the RX 570’s PCIe 3.0 x16.

The RX 570’s strengths are confined to legacy scenarios and its broader API compatibility. Its Vulkan 1.3 support is one version behind the RX 9070’s Vulkan 1.4, but its GCN architecture has proven longevity in older games. The RX 570’s 224.0 GB/s bandwidth is sufficient for 1080p gaming of its generation, and its 5.095 TFLOPS FP32 is adequate for older compute tasks. The RX 9070 wins every direct comparison in the recorded data, and the RX 570’s only advantages are in aggregate scores that include benchmarks the RX 9070 does not run competitively.