AMD Radeon Pro 570X vs AMD Radeon Pro Vega 20 Comparison

Head-to-Head Benchmarks

The recorded data points to a clear overall winner in direct comparison: the AMD Radeon Pro 570X takes all three benchmark tests against the AMD Radeon Pro Vega 20. The most decisive margin appears in the Geekbench Metal workload, where the 570X scores 40,066 against the Vega 20's 30,427, a 31.7% advantage. This is the largest performance gap in the entire head-to-head set, and it establishes the 570X as the stronger option for Metal-accelerated applications within this comparison.

The Geekbench Vulkan results show a narrower but still substantial lead for the 570X. It records 28,859, while the Vega 20 trails at 26,410, a 9.3% difference. This suggests that in Vulkan-based workloads, the architectural advantages of the 570X are still present, though less pronounced than in Metal. The smallest delta is in Geekbench OpenCL, where the 570X scores 27,604 versus 26,679 for the Vega 20, a 3.5% edge. While the 570X wins this test, the closeness of the scores indicates that OpenCL performance is nearly a wash between the two, with the Vega 20's higher boost clock partially compensating for its lower shading unit count.

Looking at the broader database context, the 570X's average benchmark score is 32,176, which places it in the 76th percentile of all GPUs. Its nearest rivals in the database include the AMD FirePro S10000 at 32,388 (0.7% higher), the AMD Radeon RX 7900 GRE at 32,456 (0.9% higher), and the AMD FirePro S9300 X2 at 32,540 (1.1% higher). The 570X sits just slightly below these peers, while edging out the Intel Arc Pro A30M, which scores 31,894 (0.9% lower). For the Vega 20, the average benchmark score is 27,839, placing it in the 73rd percentile. Its nearest rivals include the AMD Radeon RX 7800M at 27,883 (0.2% higher), the AMD Radeon Pro W5500X at 27,973 (0.5% higher), and the NVIDIA GeForce GTX 980 Ti at 28,020 (0.6% higher). Interestingly, the NVIDIA GeForce RTX 3090 sits 1% lower at 27,565, which indicates the Vega 20 holds its own against much larger GPUs in this specific scoring system.

The head-to-head data does not show a single test where the Vega 20 emerges victorious. All three deltas favor the 570X, with the win count at 3 for the 570X and 0 for the Vega 20. This consistency across different API workloads is notable, as it suggests the 570X's higher raw throughput metrics translate into real-world benchmark advantages regardless of the underlying software layer.

FAQ

Q: Which GPU wins the Geekbench Metal benchmark, and by how much?

A: The AMD Radeon Pro 570X wins the Geekbench Metal test with a score of 40,066, while the AMD Radeon Pro Vega 20 scores 30,427. This represents a 31.7% advantage for the 570X.

Q: Are there any benchmark tests where the AMD Radeon Pro Vega 20 outperforms the Radeon Pro 570X?

A: No. In the recorded head-to-head benchmarks for Geekbench Metal, OpenCL, and Vulkan, the 570X wins all three tests. The Vega 20's closest result is in OpenCL, where it trails by only 3.5%.

Q: How does the average benchmark score of each GPU compare to its nearest rivals?

A: The 570X has an average score of 32,176, which is 0.7% lower than the AMD FirePro S10000 and 0.9% lower than the AMD Radeon RX 7900 GRE. The Vega 20 averages 27,839, which is 0.2% lower than the AMD Radeon RX 7800M and 0.5% lower than the AMD Radeon Pro W5500X.

Q: What is the difference in shading unit count between the two GPUs?

A: The AMD Radeon Pro 570X features 1,792 shading units, while the AMD Radeon Pro Vega 20 has 1,280 shading units. The 570X also has more TMUs, with 112 versus 80 for the Vega 20, while both have 32 ROPs.

Q: Which GPU has a higher boost clock?

A: The AMD Radeon Pro Vega 20 has a higher boost clock at 1283 MHz, compared to 1105 MHz for the AMD Radeon Pro 570X. However, the 570X has a higher base clock at 1000 MHz versus 815 MHz for the Vega 20.

Q: How do the FP32 and FP16 performance figures differ between the two?

A: The 570X delivers 3.960 TFLOPS for FP32 and the same 3.960 TFLOPS for FP16 (at a 1:1 ratio). The Vega 20 delivers 3.284 TFLOPS for FP32 but 6.569 TFLOPS for FP16 (at a 2:1 ratio), making it substantially faster in half-precision work despite being slower in full-precision.

Architecture Differences

The two GPUs come from different generations of AMD's Graphics Core Next architecture. The AMD Radeon Pro 570X is built on GCN 4.0, using the Ellesmere chip, while the AMD Radeon Pro Vega 20 uses GCN 5.0 with the Vega 12 chip. Both are manufactured on a 14 nm process by GlobalFoundries, so the process node itself is identical, but the microarchitectural changes between GCN 4.0 and GCN 5.0 introduce significant differences in feature support and compute capabilities.

The most prominent architectural divergence is in the memory subsystem. The 570X uses a 256-bit GDDR5 interface with 4 GB of memory, delivering a bandwidth of 217.0 GB/s. The Vega 20, by contrast, employs a 1024-bit HBM2 interface, also with 4 GB, but its bandwidth is rated at 189.4 GB/s. While the Vega 20's HBM2 bus is four times wider, its effective memory clock of 1480 Mbps is substantially lower than the 570X's 6.8 Gbps effective transfer rate, resulting in lower overall bandwidth. This is a notable inversion, as HBM2 typically offers higher bandwidth, but in this specific configuration, the GDDR5 solution on the 570X wins.

Compute throughput also differs. The 570X's FP32 rate is 3.960 TFLOPS, while the Vega 20 manages 3.284 TFLOPS. However, the Vega 20's FP16 rate is 6.569 TFLOPS, achieved through a 2:1 ratio, meaning it can perform twice as many half-precision operations as full-precision. The 570X offers FP16 at a 1:1 ratio, so its FP16 throughput equals its FP32 throughput at 3.960 TFLOPS. For workloads that leverage FP16 arithmetic, the Vega 20 is clearly the superior architecture, with a 65.8% higher FP16 rate.

DirectX feature support also differs. The 570X supports DirectX 12 (12_0), while the Vega 20 supports DirectX 12 (12_1). This indicates the Vega 20 includes additional feature levels, such as conservative rasterization and rasterizer-ordered views, which are not available on the 570X. Both GPUs support OpenGL 4.6 and Vulkan 1.3, so the API compatibility is otherwise equal.

The pixel and texture rates reflect the clock and shader configuration. The 570X has a pixel rate of 35.36 GPixel/s and a texture rate of 123.8 GTexel/s. The Vega 20, despite its lower FP32 throughput, has a higher pixel rate at 41.06 GPixel/s, thanks to its higher boost clock, but a lower texture rate at 102.6 GTexel/s due to its fewer TMUs. This means the Vega 20 can fill pixels slightly faster, while the 570X excels at texture-heavy workloads.

Specification Differences

The following specifications differ between the AMD Radeon Pro 570X and the AMD Radeon Pro Vega 20:

• Architecture: GCN 4.0 (570X) versus GCN 5.0 (Vega 20)
• Chip: Ellesmere (570X) versus Vega 12 (Vega 20)
• Transistor count: 5,700 million (570X) versus not recorded (Vega 20)
• Die size: 232 mm² (570X) versus not recorded (Vega 20)
• Transistor density: 24.6M / mm² (570X) versus not recorded (Vega 20)
• Base clock: 1000 MHz (570X) versus 815 MHz (Vega 20)
• Boost clock: 1105 MHz (570X) versus 1283 MHz (Vega 20)
• Memory clock: 1695 MHz (570X) versus 740 MHz (Vega 20)
• Memory type: GDDR5 (570X) versus HBM2 (Vega 20)
• Memory bus width: 256 bit (570X) versus 1024 bit (Vega 20)
• Memory bandwidth: 217.0 GB/s (570X) versus 189.4 GB/s (Vega 20)
• Shading units: 1792 (570X) versus 1280 (Vega 20)
• TMUs: 112 (570X) versus 80 (Vega 20)
• Pixel rate: 35.36 GPixel/s (570X) versus 41.06 GPixel/s (Vega 20)
• Texture rate: 123.8 GTexel/s (570X) versus 102.6 GTexel/s (Vega 20)
• FP32 performance: 3.960 TFLOPS (570X) versus 3.284 TFLOPS (Vega 20)
• FP16 performance: 3.960 TFLOPS (570X) versus 6.569 TFLOPS (Vega 20)
• TDP: 150 W (570X) versus 100 W (Vega 20)
• DirectX support: 12 (12_0) (570X) versus 12 (12_1) (Vega 20)
• Power connectors: None (570X) versus not recorded (Vega 20)
• Release date: 2019-03-17 (570X) versus 2018-11-13 (Vega 20)

Both share the same 4 GB memory capacity, 32 ROPs, PCIe 3.0 x16 bus interface, IGP slot width, OpenGL 4.6 and Vulkan 1.3 support, and end-of-life production status.

The Verdict

The benchmark data is unambiguous: the AMD Radeon Pro 570X is the faster GPU in this pairing across all measured workloads. Its 31.7% lead in Geekbench Metal is the headline result, and it also wins OpenCL by 3.5% and Vulkan by 9.3%. For users of Metal-accelerated applications, particularly those in a macOS environment given the "Radeon Pro Mac" generation label, the 570X is the clear choice if raw performance is the sole criterion.

The 570X achieves this through a combination of more shading units (1,792 versus 1,280), higher FP32 throughput (3.960 TFLOPS versus 3.284 TFLOPS), and greater memory bandwidth (217.0 GB/s versus 189.4 GB/s). Its higher TDP of 150 W versus 100 W for the Vega 20 suggests it draws more power, but the database does not list power consumption measurements, so the efficiency comparison cannot be quantified here.

However, the Vega 20 is not without its own strengths. Its FP16 performance of 6.569 TFLOPS is more than double its FP32 rate, and it is 65.8% higher than the 570X's FP16 figure. For compute workloads that utilize half-precision arithmetic, the Vega 20 is the better processor. Additionally, its DirectX 12_1 support adds feature levels that the 570X lacks, which may matter for specific graphics applications that use conservative rasterization or rasterizer-ordered views.

The Vega 20 also has a higher pixel rate (41.06 GPixel/s versus 35.36 GPixel/s), meaning it can fill more pixels per second. This could provide an edge in certain rendering scenarios, though the 570X compensates with a higher texture rate. The lower TDP of the Vega 20 might also be relevant for thermal-constrained systems, though the database does not provide thermal performance data.

The percentile ranking places the 570X at the 76th percentile and the Vega 20 at the 73rd percentile, confirming that the 570X holds a slightly higher overall position in the database's GPU hierarchy. The average benchmark scores of 32,176 versus 27,839 reflect a 15.6% difference in aggregate performance, aligning with the head-to-head results.

For a buyer choosing between these two end-of-life mobile GPUs, the decision hinges on workload characteristics. For general-purpose graphics, Metal-based rendering, and full-precision compute, the 570X is superior. For FP16-heavy machine learning or compute tasks, the Vega 20's 2:1 FP16 ratio gives it a distinct advantage. The data does not support a recommendation for the Vega 20 in any of the three tested Geekbench workloads, but its architectural features suggest it was designed for a different set of priorities. The verdict from the recorded numbers is straightforward: the 570X wins the benchmark comparison, while the Vega 20 offers specific compute capabilities that the 570X cannot match.