AMD Radeon Pro W6900X vs NVIDIA A10G Comparison

The NVIDIA A10G and AMD Radeon Pro W6900X are both end-of-life workstation-class accelerators, yet they represent fundamentally different design philosophies. The A10G is a server-oriented Ampere card with no display outputs, while the W6900X is a Mac-focused RDNA 2 part with a full video output suite. Benchmark data shows a close overall contest, with the A10G holding a narrow edge in two of three tested workloads, but the W6900X countering with a decisive win in Apple’s Metal API. This analysis walks through the head-to-head results, architectural divergence, and where each card’s strengths lie.

Head-to-Head Benchmarks

The Geekbench OpenCL test delivers the largest margin in this comparison. The NVIDIA A10G scores 158,063 against the AMD Radeon Pro W6900X’s 131,392, a 20.3% advantage. This is a substantial gap that reflects the A10G’s raw FP32 throughput, which is listed at 31.52 TFLOPS versus the W6900X’s 22.23 TFLOPS. OpenCL workloads tend to favor raw compute density, and the NVIDIA card’s higher shading unit count (9,216 vs 5,120) appears to translate directly into this performance lead. The A10G also wins in Vulkan, but by a much slimmer margin: 145,863 versus 143,199, a 1.9% difference. That near-tie suggests Vulkan performance is more balanced, with the W6900X’s higher clock speeds and pixel rate partially offsetting NVIDIA’s compute advantage.

The W6900X’s only benchmark victory comes in Geekbench Metal, a workload not available on the A10G. The AMD card scores 205,557 in Metal, which is notably higher than its own OpenCL result of 131,392 — a 56.5% improvement. This indicates that the W6900X is heavily optimized for Apple’s graphics API, which is expected given its Radeon Pro Mac positioning. For users in a macOS ecosystem, that Metal score is likely the most relevant number, as it shows the card’s true potential in native applications. However, in cross-platform APIs like OpenCL and Vulkan, the A10G holds the advantage in both tests. The average benchmark score across all available tests is 151,963 for the A10G and 160,049 for the W6900X, meaning the AMD card’s Metal result pulls its overall average 5.1% ahead — yet that average includes the Metal test the NVIDIA card cannot run. When comparing only the two shared tests, the A10G wins both, with deltas of 20.3% and 1.9%.

Architecture Differences

The architectural chasm between these two cards is stark. The NVIDIA A10G uses the GA102 chip on an 8 nm Samsung process, while the AMD Radeon Pro W6900X uses the Navi 21 chip on a 7 nm TSMC process. The smaller process node allows AMD to pack 51.5 million transistors per square millimeter, versus NVIDIA’s 45.1M per mm², even though the A10G has a larger absolute transistor count (28,300 million vs 26,800 million) and a larger die size (628 mm² vs 520 mm²). Clock speeds favor AMD dramatically: the W6900X runs at a base of 1825 MHz and boosts to 2171 MHz, while the A10G sits at 1320 MHz base and 1710 MHz boost. Despite lower clocks, the A10G achieves higher FP32 performance (31.52 TFLOPS vs 22.23 TFLOPS) thanks to its 9,216 shading units. In FP16, the situation reverses: the W6900X delivers 44.46 TFLOPS via a 2:1 rate, while the A10G matches its FP32 at 31.52 TFLOPS (1:1).

Ray tracing hardware differs in count but not in kind. The A10G has 72 RT cores, while the W6900X has 80 RT cores. The NVIDIA card also includes 288 tensor cores, which the AMD card lacks entirely — a notable absence for any AI or machine learning acceleration. Memory architecture diverges in capacity and bandwidth. The A10G offers 24 GB of GDDR6 on a 384-bit bus, yielding 600.2 GB/s of bandwidth. The W6900X doubles capacity to 32 GB but uses a 256-bit bus, capping bandwidth at 512.0 GB/s. This means the A10G has 17.2% more memory bandwidth, while the W6900X has 33.3% more memory capacity. Texture and pixel throughput favor AMD: the W6900X achieves 694.7 GTexel/s and 277.9 GPixel/s, versus 492.5 GTexel/s and 164.2 GPixel/s for the A10G. The power envelope is also lopsided — the A10G draws 150 W TDP, while the W6900X is rated at 300 W, double the power for lower compute throughput.

Where Each One Wins

The NVIDIA A10G wins in raw compute-bound workloads, particularly those using OpenCL. Its 20.3% OpenCL lead over the W6900X is the single largest gap in this comparison, and it stems from the sheer number of shading units and tensor cores. The A10G’s 31.52 TFLOPS FP32 rating is 41.8% higher than the W6900X’s 22.23 TFLOPS, making it the superior choice for general-purpose GPU compute, scientific simulations, or any task that leverages CUDA or OpenCL pipelines. The Vulkan win, though small at 1.9%, still demonstrates that the A10G holds a slight edge in cross-platform graphics workloads. Additionally, the A10G’s 600.2 GB/s memory bandwidth provides a 17.2% advantage over the W6900X, which benefits memory-intensive rendering or data processing tasks.

The AMD Radeon Pro W6900X wins in Apple-centric workflows. Its Geekbench Metal score of 205,557 is not just a win — it is a statement. That score is 56.5% higher than its own OpenCL result, indicating that the card is specifically tuned for Metal. For users running macOS applications, the W6900X’s 32 GB of VRAM offers twice the capacity of the A10G, which is critical for large scenes, high-resolution textures, or multi-GPU rendering setups. The W6900X also leads in pixel rate (277.9 GPixel/s vs 164.2 GPixel/s) and texture rate (694.7 GTexel/s vs 492.5 GTexel/s), which benefits rasterization-heavy workloads like 4K video editing or real-time 3D viewports. Its display outputs — 1x HDMI 2.1 and 4x Thunderbolt — make it a functional workstation card, whereas the A10G has no outputs at all, requiring a separate GPU for display.

Specification Differences

The two cards differ across nearly every major specification. The A10G uses an 8 nm process from Samsung, while the W6900X uses a 7 nm process from TSMC. Transistor counts are close (28,300 million vs 26,800 million), but die size differs (628 mm² vs 520 mm²), as does transistor density (45.1M / mm² vs 51.5M / mm²). Clock speeds are dramatically different: the A10G runs at 1320 MHz base and 1710 MHz boost, while the W6900X runs at 1825 MHz base and 2171 MHz boost. Memory configurations contrast sharply: the A10G has 24 GB GDDR6 on a 384-bit bus with 600.2 GB/s bandwidth, while the W6900X has 32 GB GDDR6 on a 256-bit bus with 512.0 GB/s bandwidth. Shading units favor NVIDIA (9,216 vs 5,120), while texture mapping units favor AMD (320 vs 288), and render output units favor AMD (128 vs 96). Ray tracing core counts are close (72 vs 80), but tensor cores exist only on the A10G (288). FP32 performance is higher on the A10G (31.52 TFLOPS vs 22.23 TFLOPS), while FP16 is higher on the W6900X (44.46 TFLOPS vs 31.52 TFLOPS). Power consumption is 150 W for the A10G and 300 W for the W6900X, with suggested PSU ratings of 450 W and 700 W respectively. The A10G uses PCIe 4.0 x16, while the W6900X uses Apple MPX. Display outputs are absent on the A10G but present on the W6900X (1x HDMI 2.1, 4x Thunderbolt). Physical dimensions are similar in length (both 267 mm) but differ in height (112 mm vs 120 mm). The W6900X has a launch MSRP of 5,999 USD.

FAQ

Q: Which card has better OpenCL performance?

A: The NVIDIA A10G scores 158,063 in Geekbench OpenCL, which is 20.3% higher than the AMD Radeon Pro W6900X’s 131,392.

Q: Does the AMD Radeon Pro W6900X win any benchmark?

A: Yes, the W6900X scores 205,557 in Geekbench Metal, a test the A10G does not run. This is 56.5% higher than the W6900X’s own OpenCL score.

Q: How do memory capacities compare?

A: The A10G has 24 GB of GDDR6 on a 384-bit bus, while the W6900X has 32 GB of GDDR6 on a 256-bit bus. The A10G offers higher bandwidth (600.2 GB/s vs 512.0 GB/s), but the W6900X has 33.3% more capacity.

Q: What is the FP32 performance difference?

A: The A10G delivers 31.52 TFLOPS, which is 41.8% higher than the W6900X’s 22.23 TFLOPS. However, in FP16, the W6900X leads with 44.46 TFLOPS versus 31.52 TFLOPS.

Q: Which card supports ray tracing and tensor cores?

A: Both cards support ray tracing, with the A10G having 72 RT cores and the W6900X having 80 RT cores. Only the A10G has tensor cores (288), which the W6900X lacks entirely.

Q: Are these cards still in production?

A: No, both the NVIDIA A10G and AMD Radeon Pro W6900X are marked as end-of-life products. The A10G was released on 2021-04-11, and the W6900X on 2021-08-02.

The Verdict

The data points to a clear split based on workload and platform. The NVIDIA A10G is the superior card for compute-heavy tasks in cross-platform environments. It wins both shared benchmarks — OpenCL by 20.3% and Vulkan by 1.9% — and offers higher FP32 performance (31.52 TFLOPS vs 22.23 TFLOPS), more memory bandwidth (600.2 GB/s vs 512.0 GB/s), and tensor cores for AI acceleration. Its 150 W TDP also makes it far more power-efficient than the W6900X’s 300 W TDP, which is a significant consideration for dense server deployments. The A10G’s lack of display outputs is irrelevant in a server context, where it is designed to sit alongside a separate GPU.

The AMD Radeon Pro W6900X is the better choice for macOS users who need a functional workstation card with video output. Its Geekbench Metal score of 205,557 is the highest single benchmark result in this comparison, and its 32 GB VRAM doubles the A10G’s capacity. The W6900X also leads in pixel rate (277.9 GPixel/s vs 164.2 GPixel/s) and texture rate (694.7 GTexel/s vs 492.5 GTexel/s), making it more responsive in rasterization-heavy workflows. Its display outputs (1x HDMI 2.1, 4x Thunderbolt) and Apple MPX interface are designed for Mac Pro systems. The launch MSRP of 5,999 USD reflects its premium workstation positioning, but for users already invested in Apple’s ecosystem, the W6900X’s Metal optimization and memory headroom justify its selection. In short, choose the A10G for raw compute and efficiency, or the W6900X for macOS integration and memory capacity.