AMD Radeon Pro Duo vs NVIDIA P104-100 Comparison

AMD Radeon Pro Duo and NVIDIA P104-100 represent two divergent approaches to GPU design, with the former built for professional compute workloads and the latter engineered for mining efficiency. The data shows that despite sharing the same 4 GB memory capacity, these cards occupy very different performance tiers, with the P104-100 delivering a decisive victory in the only directly comparable benchmark. The Radeon Pro Duo holds a higher overall percentile ranking at 80 versus the P104-100’s 77, but that aggregate position is misleading given the specific head-to-head results.

Head-to-Head Benchmarks

The only directly comparable benchmark between these two cards is Geekbench OpenCL, and the results are stark. The NVIDIA P104-100 scores 52,368, while the AMD Radeon Pro Duo manages 35,860. This represents a 31.5% deficit for the Radeon Pro Duo, meaning the P104-100 is roughly 46% faster in raw OpenCL compute throughput. The margin is substantial enough to define the entire competitive relationship between these two products.

Looking at the nearest rivals for each card contextualizes these scores further. The Radeon Pro Duo’s 35,860 OpenCL score places it between the NVIDIA T1000 (36,289, which is 1.2% higher) and the NVIDIA Quadro GV100 (35,520, which is 1% lower). This indicates the Radeon Pro Duo sits in a competitive band where small percentage differences separate it from other professional cards. The AMD Radeon RX 5300M is 1.8% ahead, while the NVIDIA Quadro GV100 is 1% behind — a tight cluster.

The P104-100’s 52,368 OpenCL score, by contrast, puts it in a different league relative to its nearest rivals. The NVIDIA T600 Mobile scores 32,849 (0.4% lower), the T550 Mobile scores 33,161 (0.5% lower), the GeForce RTX 3050 Mobile scores 33,170 (0.6% lower), and the AMD Radeon Pro 570 scores 33,207 (0.7% lower). The P104-100 outperforms all of these by roughly 58%, demonstrating that its raw compute capability far exceeds what its modest specification sheet might suggest.

The head-to-head delta of -31.5% for the Radeon Pro Duo is the single most decisive data point in this comparison. No other benchmark results exist to counterbalance this outcome, so the P104-100 wins the only direct comparison by a wide margin. The Radeon Pro Duo’s higher percentile rank (80 vs 77) stems from its single benchmark score being more consistent, whereas the P104-100’s average is pulled down by its additional 3DMark Steel Nomad DX12 score of 1,413, which is low relative to its OpenCL performance.

FAQ

Q: Which card has the higher Geekbench OpenCL score?

A: The NVIDIA P104-100 scores 52,368, which is 31.5% higher than the AMD Radeon Pro Duo’s 35,860. This is the only head-to-head benchmark available, and the P104-100 wins it decisively.

Q: How do the two cards compare in terms of overall GPU percentile rankings?

A: The AMD Radeon Pro Duo ranks in the 80th percentile among all GPUs, while the NVIDIA P104-100 ranks in the 77th percentile. Despite losing the head-to-head benchmark, the Radeon Pro Duo has a higher overall percentile position.

Q: What is the average benchmark score for each card?

A: The AMD Radeon Pro Duo has an average benchmark score of 35,860, which is identical to its single OpenCL score. The NVIDIA P104-100 has an average score of 32,982, which is lower than its OpenCL score due to the inclusion of its 3DMark Steel Nomad DX12 result of 1,413.

Q: Which card has better API support for Vulkan?

A: The NVIDIA P104-100 supports Vulkan 1.4, while the AMD Radeon Pro Duo supports Vulkan 1.2.170. The P104-100 also supports DirectX 12 (12_1), whereas the Radeon Pro Duo supports DirectX 12 (12_0).

Q: Do both cards have the same memory capacity?

A: Yes, both cards have 4 GB of memory. However, the Radeon Pro Duo uses HBM with a 4096-bit bus and 512.0 GB/s bandwidth, while the P104-100 uses GDDR5X with a 256-bit bus and 320.3 GB/s bandwidth.

Q: Which card has more shading units?

A: The AMD Radeon Pro Duo has 4,096 shading units, which is more than double the P104-100’s 1,920. The Radeon Pro Duo also has 256 texture mapping units versus 120, though both have 64 ROPs.

The Verdict

The data points to a clear winner for raw compute performance: the NVIDIA P104-100. Its 52,368 OpenCL score demolishes the Radeon Pro Duo’s 35,860, and the 31.5% delta is far too large to ignore. Anyone prioritizing OpenCL throughput should choose the P104-100 without hesitation.

However, the Radeon Pro Duo is not without merit. Its 80th percentile ranking versus the P104-100’s 77th suggests that in the broader GPU landscape, it holds a slightly stronger position. Its average benchmark score of 35,860 is higher than the P104-100’s 32,982, though this is driven entirely by the P104-100’s weak 3DMark Steel Nomad DX12 result.

The P104-100’s higher average score in the only direct comparison makes it the better choice for compute-centric workloads. The Radeon Pro Duo’s higher percentile rank and more consistent benchmark profile may appeal to those who value stability across a wider range of tasks, but the head-to-head data cannot be ignored.

For professional users who need OpenCL performance, the P104-100 is the superior option. For those who value the Radeon Pro Duo’s higher overall percentile and its display outputs — which the P104-100 lacks entirely — the AMD card remains viable, but it is not the compute champion here.

Specification Differences

The two cards differ in nearly every major specification category. The AMD Radeon Pro Duo uses a 28 nm process node from TSMC, while the NVIDIA P104-100 uses a 16 nm node, also from TSMC. The Radeon Pro Duo packs 8,900 million transistors on a 596 mm² die, while the P104-100 has 7,200 million transistors on a much smaller 314 mm² die. This yields a transistor density of 14.9M per mm² for the AMD card and 22.9M per mm² for the NVIDIA card.

Clock speeds differ drastically: the P104-100 has a base clock of 1607 MHz and a boost clock of 1733 MHz, while the Radeon Pro Duo lists no base or boost clocks — only a memory clock of 500 MHz (1000 Mbps effective). The P104-100’s memory runs at 1251 MHz (10 Gbps effective).

Memory configurations are similar in capacity (4 GB) but different in type and bandwidth. The Radeon Pro Duo uses HBM on a 4096-bit bus with 512.0 GB/s bandwidth, while the P104-100 uses GDDR5X on a 256-bit bus with 320.3 GB/s bandwidth.

Compute resources differ significantly: the Radeon Pro Duo has 4,096 shading units and 256 TMUs, versus the P104-100’s 1,920 shading units and 120 TMUs. Both have 64 ROPs. Pixel rates favor the P104-100 at 110.9 GPixel/s versus 64.00 GPixel/s, but texture rates favor the Radeon Pro Duo at 256.0 GTexel/s versus 208.0 GTexel/s.

FP32 performance is higher on the Radeon Pro Duo at 8.192 TFLOPS versus 6.655 TFLOPS. FP16 performance shows the biggest gap: the Radeon Pro Duo delivers 8.192 TFLOPS (1:1 ratio), while the P104-100 manages only 104.0 GFLOPS (1:64 ratio).

Power requirements differ: the Radeon Pro Duo has a 350 W TDP with 3x 8-pin connectors and a 750 W suggested PSU, while the P104-100 has no listed TDP, a single 8-pin connector, and a 200 W suggested PSU. The P104-100 also uses a PCIe 1.0 x4 interface versus the Radeon Pro Duo’s PCIe 3.0 x16.

Architecture Differences

The architectural divide is fundamental. The AMD Radeon Pro Duo is built on GCN 3.0 architecture (chip codename "Capsaicin") from the Radeon Pro GCN generation, released on 2016-04-25. The NVIDIA P104-100 uses Pascal architecture (chip GP104) from the Mining GPUs generation, released on 2017-12-11.

The process node difference is significant: 28 nm for AMD versus 16 nm for NVIDIA, both fabricated by TSMC. This explains the P104-100’s superior transistor density despite fewer total transistors.

Memory architecture differs completely. The Radeon Pro Duo uses HBM with a 4096-bit memory bus, which provides 512.0 GB/s bandwidth — a 60% advantage over the P104-100’s 320.3 GB/s. However, the P104-100 compensates with much higher clock speeds.

Compute feature sets diverge on FP16 support. The Radeon Pro Duo supports FP16 at a 1:1 ratio with FP32, delivering 8.192 TFLOPS. The P104-100 has FP16 at a 1:64 ratio, yielding only 104.0 GFLOPS — a massive 78x disadvantage in half-precision throughput.

API support differs: the Radeon Pro Duo supports DirectX 12 (12_0) and Vulkan 1.2.170, while the P104-100 supports DirectX 12 (12_1) and Vulkan 1.4. Both support OpenGL 4.6.

Physical and connectivity differences are pronounced. The Radeon Pro Duo measures 277 mm (10.9 inches) in length and 111 mm (4.4 inches) in height, with 1x HDMI 1.4a and 3x DisplayPort 1.2 outputs. The P104-100 is 267 mm (10.5 inches) long with no display outputs at all — it is a compute-only card. Both are dual-slot designs.

Where Each One Wins

The NVIDIA P104-100 wins decisively in OpenCL compute performance, as demonstrated by its 52,368 score versus 35,860. This makes it the clear choice for applications that rely heavily on OpenCL acceleration. Its higher pixel rate (110.9 GPixel/s vs 64.00 GPixel/s) also suggests better fill-rate-bound performance, despite fewer shading units.

The AMD Radeon Pro Duo wins in several specification-driven categories that don’t appear in the head-to-head benchmark. Its 8.192 TFLOPS FP32 performance exceeds the P104-100’s 6.655 TFLOPS. Its FP16 capability of 8.192 TFLOPS (1:1) is vastly superior to the P104-100’s 104.0 GFLOPS (1:64), making it the better choice for workloads that leverage half-precision arithmetic. Its 512.0 GB/s memory bandwidth is 60% higher than the P104-100’s 320.3 GB/s, which benefits memory-intensive tasks.

The Radeon Pro Duo also offers display outputs (1x HDMI 1.4a, 3x DisplayPort 1.2), while the P104-100 has none. This makes the AMD card the only option for any workload requiring visual output. Its higher percentile rank (80 vs 77) and higher average benchmark score (35,860 vs 32,982) further support its broader applicability.

The P104-100’s lower power requirements — no listed TDP, a single 8-pin connector, and a 200 W suggested PSU versus the Radeon Pro Duo’s 350 W TDP, 3x 8-pin connectors, and 750 W suggested PSU — make it the more practical choice for power-constrained environments. Its PCIe 1.0 x4 interface is a limiting factor, however, compared to the Radeon Pro Duo’s PCIe 3.0 x16.

In summary, the P104-100 wins for raw OpenCL throughput and efficiency, while the Radeon Pro Duo wins for FP16 compute, memory bandwidth, display capability, and overall percentile ranking. The choice depends entirely on whether the workload prioritizes the P104-100’s dominant OpenCL score or the Radeon Pro Duo’s broader feature set.