AMD Radeon Pro 570 vs NVIDIA GeForce RTX 4070 Comparison

The NVIDIA GeForce RTX 4070 and the AMD Radeon Pro 570 occupy very different positions in the database, and the recorded data makes the gap unmistakable. In the two head-to-head benchmarks both cards share, the RTX 4070 wins by margins of 459 percent in Geekbench OpenCL and 444.7 percent in Geekbench Vulkan. This is not a close comparison; it is a generational mismatch between a 2023 Ada Lovelace desktop card and a 2017 Mac-focused GCN 4.0 part. Anyone choosing strictly on benchmark results should pick the RTX 4070 in every workload the database measures.

The Verdict

The data points one way. The RTX 4070 sits at the 81st percentile versus all GPUs in the database, with an average benchmark score of 37648, while the Radeon Pro 570 sits at the 78th percentile with an average score of 33207. Those percentile positions look deceptively close, but the shared benchmark results reveal the real distance: 154858 versus 27702 in Geekbench OpenCL, and 174152 versus 31974 in Geekbench Vulkan. The RTX 4070 wins both head-to-head tests, and the head-to-head record stands at 2 wins for the RTX 4070 and 0 for the Radeon Pro 570.

Choose the RTX 4070 for any performance-driven use case: gaming, ray tracing, compute, GPU-accelerated rendering. It has 46 RT cores and 184 tensor cores; the Radeon Pro 570 has neither. Choose the Radeon Pro 570 only in the narrow context its data implies: it is an integrated Mac part (the generation field lists "Radeon Pro Mac (500 Series)"), drawn from an older platform, with no discrete power connector and a 150 W TDP. The database contains no scenario in which it outperforms the RTX 4070.

Architecture Differences

The architectural distance between these two GPUs is the widest kind the database records. The RTX 4070 is built on TSMC's 5 nm process, packs 35,800 million transistors onto a 294 mm² die, and reaches a transistor density of 121.8M / mm². The Radeon Pro 570 is built on GlobalFoundries' 14 nm process with 5,700 million transistors on a 232 mm² die, giving a density of 24.6M / mm². That is roughly a six-fold transistor count advantage and roughly a five-fold density advantage for the Ada Lovelace card, spread across a die only modestly larger in area.

The core counts tell the same story. The RTX 4070 carries 5888 shading units, 184 TMUs, 64 ROPs, 46 RT cores, and 184 tensor cores. The Radeon Pro 570 carries 1792 shading units, 112 TMUs, and 32 ROPs, with no RT or tensor cores recorded at all. Fixed-function throughput follows: the RTX 4070 delivers 158.4 GPixel/s of pixel fill and 455.4 GTexel/s of texture fill, against 35.36 GPixel/s and 123.8 GTexel/s for the Pro 570. Raw compute sits at 29.15 TFLOPS FP32 for the RTX 4070 versus 3.960 TFLOPS for the Pro 570.

The feature sets diverge just as sharply. The RTX 4070 supports DirectX 12 Ultimate (feature level 12_2) and Vulkan 1.4; the Pro 570 supports DirectX 12 (12_0) and Vulkan 1.3. Both list OpenGL 4.6. The RTX 4070 connects over PCIe 4.0 x16, while the Pro 570 uses PCIe 3.0 x16. On memory, the RTX 4070 pairs 12 GB of GDDR6X on a 192-bit bus for 504.2 GB/s of bandwidth; the Pro 570 has 4 GB of GDDR5 on a 256-bit bus for 217.0 GB/s. Note the contrast: the older card has the wider bus but less than half the bandwidth and a third of the capacity.

Head-to-Head Benchmarks

Only two tests appear in both record sets, and both are landslide victories for the RTX 4070.

Geekbench OpenCL: the RTX 4070 scores 154858, the Radeon Pro 570 scores 27702. The delta is 459 percent in NVIDIA's favor. This measures general GPU compute, and the result mirrors the raw FP32 gap: 29.15 TFLOPS against 3.960 TFLOPS. The Pro 570's OpenCL score is not merely lower, it is an order of magnitude away from modern desktop compute performance.

Geekbench Vulkan: the RTX 4070 scores 174152, the Pro 570 scores 31974, a 444.7 percent margin. Interestingly, both cards score higher in Vulkan than in OpenCL, so the API shift helps both, but it does nothing to close the gap. The Pro 570's Vulkan result of 31974 is its strongest shared-test score, and it is still less than a fifth of the RTX 4070's figure.

Context from each card's own record set sharpens the picture. The RTX 4070 also posts 26927 in PassMark G3D, 14720 in PassMark GPU Compute, and 3854 in 3DMark Steel Nomad (DX12), plus PassMark DirectX results of 320 (DX9), 244 (DX11), 139 (DX10), and 103 (DX12), and a 2D score of 1164. The Pro 570's only additional recorded score is 39945 in Geekbench Metal, an Apple-platform API result that reflects its Mac heritage. Comparing across APIs is unreliable, but the Metal score is the one place the Pro 570 records a number above its OpenCL and Vulkan results.

Relative to their respective rivals, both cards sit near the top of their immediate clusters. The RTX 4070's average score of 37648 is within a fraction of the NVIDIA Tesla P4 (37628, 0.1 percent delta), the AMD Radeon RX Vega 56 (37507, 0.4 percent delta), the NVIDIA GeForce RTX 4080 Mobile (38135, a 1.3 percent deficit), and the AMD Radeon PRO W6400 (37157, 1.3 percent ahead). The Pro 570's 33207 sits alongside the NVIDIA GeForce RTX 3050 Mobile (33170, 0.1 percent), the NVIDIA T550 Mobile (33161, 0.1 percent), the NVIDIA P104-100 (32982, 0.7 percent), and the NVIDIA T600 Mobile (32849, 1.1 percent). Each is competitive in its own tier; the tiers themselves are far apart.

Specification Differences

The fields where the two cards differ, per the database:

• Chip and architecture: AD104 on Ada Lovelace (GeForce 40-series) versus Ellesmere on GCN 4.0.
• Process: 5 nm at TSMC versus 14 nm at GlobalFoundries.
• Transistors, die, density: 35,800 million, 294 mm², 121.8M / mm² versus 5,700 million, 232 mm², 24.6M / mm².
• Clocks: 1920 MHz base and 2475 MHz boost for the RTX 4070, with memory at 21 Gbps effective; 1000 MHz base and 1105 MHz boost for the Pro 570, with memory at 6.8 Gbps effective.
• Memory: 12 GB GDDR6X, 192-bit, 504.2 GB/s versus 4 GB GDDR5, 256-bit, 217.0 GB/s.
• Compute units and throughput: 5888 shading units, 184 TMUs, 64 ROPs, 46 RT cores, 184 tensor cores, 29.15 TFLOPS FP32 versus 1792 shading units, 112 TMUs, 32 ROPs, no RT or tensor cores, 3.960 TFLOPS FP32.
• Fill rates: 158.4 GPixel/s and 455.4 GTexel/s versus 35.36 GPixel/s and 123.8 GTexel/s.
• Power and physical: the RTX 4070 has a 200 W TDP, a dual-slot width, one 16-pin power connector, a suggested 550 W PSU, and dimensions of 240 mm length, 110 mm height, and 40 mm width; the Pro 570 has a 150 W TDP, is listed as IGP width, needs no power connectors, and has no recorded dimensions or suggested PSU.
• Interface and outputs: PCIe 4.0 x16 with one HDMI 2.1 and three DisplayPort 1.4a outputs versus PCIe 3.0 x16 with portable-device-dependent outputs.
• APIs: DirectX 12 Ultimate (12_2) and Vulkan 1.4 versus DirectX 12 (12_0) and Vulkan 1.3.
• Lifecycle: the RTX 4070 launched on April 11, 2023 with a launch MSRP of 599 USD, sits between the GeForce 30 predecessor and GeForce 50 successor lines, and is now end-of-life; the Pro 570 launched on June 4, 2017, has no recorded predecessor, successor, or MSRP, and is also end-of-life.

FAQ

Q: Which GPU wins the shared benchmarks?

A: The RTX 4070 wins both. It leads Geekbench OpenCL 154858 to 27702 (459 percent) and Geekbench Vulkan 174152 to 31974 (444.7 percent).

Q: Can the Radeon Pro 570 handle ray tracing?

A: No. The database records 46 RT cores on the RTX 4070 and none on the Pro 570, so hardware ray tracing is exclusive to the NVIDIA card in this pairing.

Q: How do their memory configurations compare?

A: The RTX 4070 has 12 GB of GDDR6X with 504.2 GB/s bandwidth on a 192-bit bus; the Pro 570 has 4 GB of GDDR5 with 217.0 GB/s on a wider 256-bit bus. Capacity and bandwidth both favor the RTX 4070 heavily.

Q: Are both cards still in production?

A: No. Both are marked end-of-life in the database. The RTX 4070 launched on April 11, 2023; the Radeon Pro 570 launched on June 4, 2017.

Q: What kind of product is the Radeon Pro 570?

A: The database identifies it as part of the "Radeon Pro Mac (500 Series)" generation, an integrated Mac-focused part on PCIe 3.0 x16 with no discrete power connectors.

Q: Which cards are their closest rivals?

A: The RTX 4070's nearest rivals are the Tesla P4, RX Vega 56, RTX 4080 Mobile, and Radeon PRO W6400, all within about 1.3 percent of its average score. The Pro 570's nearest rivals are the RTX 3050 Mobile, T550 Mobile, P104-100, and T600 Mobile, all within about 1.1 percent.

Where Each One Wins

The RTX 4070 wins everywhere the database measures. It wins OpenCL compute by 459 percent and Vulkan by 444.7 percent. It wins in raw capability across every recorded axis: seven times the FP32 compute (29.15 versus 3.960 TFLOPS), roughly four and a half times the pixel fill rate, close to four times the texture fill rate, more than double the memory bandwidth, three times the memory capacity, and exclusive access to RT cores, tensor cores, DirectX 12 Ultimate, and Vulkan 1.4. Its own benchmark record adds PassMark G3D, GPU Compute, DirectX, and 3DMark Steel Nomad results that the Pro 570 simply does not have entries for.

The Radeon Pro 570's wins are contextual rather than performance-based. It draws less power at 150 W versus 200 W, requires no power connector at all, and lives inside the Mac platform its generation field describes. Its Geekbench Metal score of 39945 shows it functioning within Apple's graphics stack. As a practical matter, the database positions it against mobile-tier rivals like the RTX 3050 Mobile and T600 Mobile, not against desktop flagships. For anyone evaluating these two by results, the RTX 4070 is the only data-supported choice.