AMD Radeon Pro W6800X vs NVIDIA A10G Comparison

The AMD Radeon Pro W6800X and NVIDIA A10G are both end-of-life workstation accelerators that occupy the 97th percentile of all GPUs, yet they represent fundamentally different design philosophies. The W6800X is a Mac-centric, quad-slot behemoth built for Apple MPX chassis, while the A10G is a single-slot server card with no display outputs. Benchmark data shows a single head-to-head result, with the A10G taking a decisive victory in OpenCL, but the overall average scores and architectural profiles reveal a more nuanced picture for different workloads.

Where Each One Wins

The NVIDIA A10G is the clear winner in raw compute throughput as measured by the only direct head-to-head benchmark available. In Geekbench OpenCL, the A10G scores 158,063 against the W6800X’s 124,498, a 21.2% advantage. This aligns with the A10G’s massive FP32 throughput of 31.52 TFLOPS, which is nearly double the W6800X’s 16.03 TFLOPS. For workloads that rely heavily on general-purpose compute—such as AI inference, data analytics, or scientific simulation—the A10G is the stronger candidate. Its 288 tensor cores further solidify its position for machine learning tasks, a feature the W6800X lacks entirely.

The AMD Radeon Pro W6800X, however, counters with advantages in memory capacity and bandwidth-per-dollar of silicon. It offers 32 GB of GDDR6 memory versus the A10G’s 24 GB, a 33% capacity advantage that matters for large dataset processing or high-resolution rendering scenes that exceed 24 GB. Its memory bandwidth is 512.0 GB/s, which is lower than the A10G’s 600.2 GB/s, but the extra 8 GB can be decisive for workloads that spill over the A10G’s capacity. The W6800X also has a higher boost clock (2087 MHz vs 1710 MHz) and a smaller die (520 mm² vs 628 mm²), which contributes to its higher transistor density of 51.5M/mm² versus 45.1M/mm². For Mac-based workflows—particularly those using Metal API—the W6800X’s Geekbench Metal score of 196,844 shows its strength, though no direct A10G Metal score exists for comparison.

Architecture Differences

The two GPUs are built on different process nodes and foundries. The W6800X uses TSMC’s 7 nm process with 26,800 million transistors on a 520 mm² die, while the A10G uses Samsung’s 8 nm process with 28,300 million transistors on a 628 mm² die. The AMD chip is based on RDNA 2.0 architecture (Navi 21), while the NVIDIA chip uses Ampere (GA102). This architectural split explains the core count disparity: the A10G packs 9,216 shading units, 288 TMUs, and 96 ROPs, versus the W6800X’s 3,840 shading units, 240 TMUs, and 96 ROPs. The A10G also has 72 RT cores and 288 tensor cores, while the W6800X has 60 RT cores and no tensor cores.

Memory subsystems differ significantly. The W6800X uses a 256-bit bus with 32 GB GDDR6 at 16 Gbps effective, yielding 512.0 GB/s. The A10G uses a wider 384-bit bus with 24 GB GDDR6 at 12.5 Gbps effective, yielding 600.2 GB/s. Clock speeds favor AMD: the W6800X runs at 1800 MHz base and 2087 MHz boost, while the A10G runs at 1320 MHz base and 1710 MHz boost. This explains why the W6800X achieves a higher pixel rate (200.4 GPixel/s vs 164.2 GPixel/s) despite having fewer cores. The texture rates are closer: 500.9 GTexel/s for AMD versus 492.5 GTexel/s for NVIDIA.

Form factor and power delivery are polar opposites. The W6800X is a quad-slot card using the Apple MPX power connector and bus interface, with a 200 W TDP and 550 W suggested PSU. The A10G is a single-slot card using an 8-pin EPS connector and PCIe 4.0 x16 interface, with a 150 W TDP and 450 W suggested PSU. The W6800X provides display outputs (1x HDMI 2.1 and 4x Thunderbolt), while the A10G has no outputs at all, making it a pure compute accelerator. The W6800X is slightly taller (120 mm vs 112 mm) but shares the same 267 mm length.

FAQ

Q: Which GPU has higher raw FP32 compute performance?

A: The NVIDIA A10G delivers 31.52 TFLOPS FP32, which is 96.6% higher than the AMD Radeon Pro W6800X’s 16.03 TFLOPS. This is reflected in the A10G’s 21.2% OpenCL benchmark victory.

Q: Can the AMD Radeon Pro W6800X be used in a standard PC server?

A: No. The W6800X uses the Apple MPX bus interface and power connector, and it is quad-slot wide. It is designed exclusively for Apple Mac Pro systems, whereas the A10G uses standard PCIe 4.0 x16 and an 8-pin EPS connector.

Q: Which card is better for large machine learning models?

A: The A10G is better suited for ML due to its 288 tensor cores and higher memory bandwidth (600.2 GB/s). However, the W6800X has 32 GB of memory versus 24 GB, so it can hold larger single models if capacity is the bottleneck.

Q: Are both GPUs still in production?

A: No. Both the AMD Radeon Pro W6800X and the NVIDIA A10G are marked as end-of-life products. The W6800X was released on August 2, 2021, while the A10G was released earlier on April 11, 2021.

Q: How do their average benchmark scores compare?

A: The W6800X has a higher average benchmark score of 160,671 versus the A10G’s 151,963, a 5.4% difference. This is curious given the A10G’s OpenCL win, suggesting the W6800X’s Metal performance (196,844) boosts its average.

Q: Which card has better ray tracing hardware?

A: The A10G has 72 RT cores versus the W6800X’s 60, giving NVIDIA a 20% advantage in ray tracing core count. Both support DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4.

Specification Differences

| Specification | AMD Radeon Pro W6800X | NVIDIA A10G |

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

| Architecture | RDNA 2.0 | Ampere |

| Process Node | 7 nm (TSMC) | 8 nm (Samsung) |

| Transistors | 26,800 million | 28,300 million |

| Die Size | 520 mm² | 628 mm² |

| Transistor Density | 51.5M / mm² | 45.1M / mm² |

| Base Clock | 1800 MHz | 1320 MHz |

| Boost Clock | 2087 MHz | 1710 MHz |

| Memory Size | 32 GB GDDR6 | 24 GB GDDR6 |

| Memory Bus Width | 256 bit | 384 bit |

| Memory Bandwidth | 512.0 GB/s | 600.2 GB/s |

| Effective Memory Speed | 16 Gbps | 12.5 Gbps |

| Shading Units | 3840 | 9216 |

| TMUs | 240 | 288 |

| ROPs | 96 | 96 |

| RT Cores | 60 | 72 |

| Tensor Cores | None | 288 |

| Pixel Rate | 200.4 GPixel/s | 164.2 GPixel/s |

| Texture Rate | 500.9 GTexel/s | 492.5 GTexel/s |

| FP32 Performance | 16.03 TFLOPS | 31.52 TFLOPS |

| FP16 Performance | 32.06 TFLOPS (2:1) | 31.52 TFLOPS (1:1) |

| TDP | 200 W | 150 W |

| Slot Width | Quad-slot | Single-slot |

| Power Connector | Apple MPX | 8-pin EPS |

| Suggested PSU | 550 W | 450 W |

| Bus Interface | Apple MPX | PCIe 4.0 x16 |

| Display Outputs | 1x HDMI 2.1, 4x Thunderbolt | No outputs |

| Dimensions (H) | 120 mm | 112 mm |

Head-to-Head Benchmarks

The sole direct benchmark comparison is Geekbench OpenCL, where the NVIDIA A10G decisively outperforms the AMD Radeon Pro W6800X. The A10G scores 158,063 against the W6800X’s 124,498, representing a 21.2% margin. This result is consistent with the A10G’s 31.52 TFLOPS FP32 throughput, which is nearly double the W6800X’s 16.03 TFLOPS. The A10G’s 9,216 shading units and 288 tensor cores provide a massive compute advantage that the W6800X’s higher clocks (2087 MHz boost vs 1710 MHz) cannot overcome.

However, the average benchmark scores tell a different story. The W6800X averages 160,671 across all benchmarks, which is 5.4% higher than the A10G’s 151,963. This gap is explained by the W6800X’s Geekbench Metal score of 196,844, a test where the A10G has no comparable result. The W6800X’s Metal performance is 58.1% higher than its own OpenCL score, indicating that AMD’s architecture is far more efficient under Apple’s API. In contrast, the A10G’s Vulkan score of 145,863 is only 7.7% lower than its OpenCL score, suggesting more consistent performance across APIs.

Looking at nearest rival comparisons, the W6800X sits just 1.1% below the NVIDIA A100 PCIe 40 GB (162,504 average) and 2.6% below the AMD Radeon PRO W7800 (164,894). The A10G, meanwhile, is 1.1% above the NVIDIA Tesla V100 PCIe 32 GB (150,305) and 9.3% above the AMD Instinct MI100 (139,035). This positions the W6800X as a stronger all-rounder in the average benchmark metric, despite losing the only head-to-head test. The A10G’s 24 GB memory capacity and 600.2 GB/s bandwidth may be more useful in compute-heavy server workloads, but the W6800X’s 32 GB capacity and superior Metal integration make it the better choice for Mac-based creative professionals. Ultimately, the A10G wins on raw compute, while the W6800X wins on platform-specific performance and memory capacity.