AMD FirePro S9300 X2 vs NVIDIA P104-100 Comparison

Head-to-Head Benchmarks

The benchmark data presents a clear hierarchy between these two compute-oriented cards. In the two shared tests, the NVIDIA P104-100 wins decisively across the board. The largest margin appears in Geekbench OpenCL, where the P104-100 scores 51,663 against the FirePro S9300 X2's 27,971, a delta of 84.7%. That is not a narrow edge; it is a near-doubling of the AMD card's result. The Vulkan test narrows the gap somewhat, but the P104-100 still leads with 45,165 versus 37,109, a 21.7% advantage.

Looking at the broader context from the nearestRivals data, the P104-100's average benchmark score sits at 32,747, marginally above the FirePro S9300 X2's 32,540 — a 0.6% difference in favor of the NVIDIA part. This is a curious situation: the head-to-head deltas are large, yet the overall averages are nearly identical. The explanation lies in the benchmark sets. The P104-100 has three recorded benchmarks (including a 3DMark Steel Nomad DX12 score of 1,413), while the FirePro has only two. The 3DMark result, being a low absolute number, drags down the P104-100's average relative to its Geekbench dominance. Consequently, the average-score comparison flatters the AMD card, but the direct head-to-head tests tell a different story.

When placing these results against the wider field, both cards occupy the 76th percentile among all GPUs. The nearest rivals list shows the NVIDIA T600 Mobile at 32,849 (0.3% above the P104-100) and the AMD Radeon RX 7800 XT at 32,619 (0.4% below). The FirePro S9300 X2 sits 0.2% below the RX 7800 XT and 0.5% above the FirePro S10000. In practical terms, these two cards are statistical peers in aggregate performance, despite the lopsided head-to-head outcomes. The data suggests that the P104-100's advantage is workload-specific — strong in OpenCL and Vulkan compute — rather than universal.

FAQ

Q: Which card wins the Geekbench OpenCL test?

A: The NVIDIA P104-100 wins by a wide margin, scoring 51,663 versus the AMD FirePro S9300 X2's 27,971, a 84.7% advantage.

Q: How do the two cards compare in average benchmark score?

A: The P104-100 averages 32,747, while the FirePro S9300 X2 averages 32,540. The NVIDIA card leads by 0.6%, though both sit at the 76th percentile among all GPUs.

Q: Does the FirePro S9300 X2 win any shared benchmark?

A: No. Across the two head-to-head tests (Geekbench OpenCL and Geekbench Vulkan), the FirePro loses both. The P104-100 records 2 wins, the FirePro 0.

Q: What is the difference in memory bandwidth between the two?

A: The FirePro S9300 X2 has 512.0 GB/s bandwidth from its 4096-bit HBM interface, while the P104-100 has 320.3 GB/s from a 256-bit GDDR5X bus. Despite the FirePro's higher bandwidth, it trails in the compute benchmarks recorded.

Q: How does the Vulkan result compare to the OpenCL result?

A: The P104-100's Vulkan lead is 21.7% (45,165 vs 37,109), which is far smaller than its 84.7% OpenCL lead. This indicates the NVIDIA card's advantage shrinks under Vulkan but remains substantial.

Q: What is the launch MSRP of the FirePro S9300 X2?

A: The launch MSRP is 5,999 USD. The P104-100 has no listed launch MSRP in the data.

Architecture Differences

The two cards represent fundamentally different design philosophies and eras. The NVIDIA P104-100 uses the GP104 chip built on TSMC's 16 nm process, part of the Pascal architecture. It packs 7,200 million transistors into a 314 mm² die, yielding a transistor density of 22.9M per mm². The AMD FirePro S9300 X2, in contrast, uses the Capsaicin chip on a much older 28 nm node, with 8,900 million transistors spread across a 596 mm² die — a density of just 14.9M per mm². The process advantage gives NVIDIA a clear efficiency edge in transistor packing, though AMD compensates with raw scale.

The FirePro S9300 X2 is based on GCN 3.0 architecture, belonging to the FirePro Server (Sx300) generation. It has 4,096 shading units and 256 texture mapping units, exactly double the P104-100's 1,920 shading units and 120 TMUs. Both cards have 64 ROPs. This doubling explains the FirePro's higher theoretical texture rate (249.6 GTexel/s vs 208.0 GTexel/s) and FP32 throughput (7.987 TFLOPS vs 6.655 TFLOPS). However, the P104-100's pixel rate is higher at 110.9 GPixel/s versus 62.40 GPixel/s — a counterintuitive result given equal ROP counts, likely stemming from clock speed differences.

Memory architecture differs sharply. The P104-100 uses 4 GB of GDDR5X on a 256-bit bus with 320.3 GB/s bandwidth, while the FirePro uses 4 GB of HBM on a 4096-bit bus with 512.0 GB/s. The FirePro's HBM is a server-oriented choice, but the NVIDIA card's GDDR5X runs at 1251 MHz (10 Gbps effective) versus the FirePro's modest 500 MHz (1000 Mbps effective). API support also diverges: the P104-100 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, while the FirePro supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.2.170. Notably, the P104-100 lists FP16 at 104.0 GFLOPS (1:64 ratio), while the FirePro has no FP16 figure recorded.

Specification Differences

The cards differ in nearly every measurable specification except memory size (both 4 GB), ROP count (both 64), slot width (both dual-slot), length (both 267 mm), and display outputs (both none). The P104-100 has a base clock of 1607 MHz and boost of 1733 MHz, while the FirePro has no base or boost clock listed. Transistor counts differ: 7,200 million for NVIDIA versus 8,900 million for AMD. Die size is nearly double on the AMD side (596 mm² vs 314 mm²). The P104-100 has 1,920 shading units and 120 TMUs; the FirePro has 4,096 and 256 respectively.

Power requirements diverge significantly. The P104-100 has no TDP listed but suggests a 200 W PSU with a single 8-pin connector. The FirePro S9300 X2 has a TDP of 300 W, requires a 700 W PSU, and uses two 8-pin connectors. Bus interface also differs: the P104-100 runs on PCIe 1.0 x4, while the FirePro uses PCIe 3.0 x16 — a major interface gap. The FirePro has physical dimensions of 267 mm length and 111 mm height; the P104-100 lists only length. Release dates are separated by roughly 20 months: the FirePro launched in March 2016, the P104-100 in December 2017. The FirePro's predecessor is FirePro Terascale and successor is Radeon Pro GCN; the P104-100 has no listed predecessor or successor.

The Verdict

The data points to a straightforward conclusion for compute workloads: the NVIDIA P104-100 is the stronger performer in the recorded benchmarks. It wins both head-to-head tests, with an 84.7% lead in OpenCL and a 21.7% lead in Vulkan. Even though the average benchmark scores are nearly identical (32,747 vs 32,540), that parity is an artifact of the P104-100 including a low 3DMark result that the FirePro lacks. In the shared tests, the NVIDIA card is categorically faster.

However, the FirePro S9300 X2 is not without merits. It offers higher raw compute specifications: more shading units (4,096 vs 1,920), higher FP32 (7.987 vs 6.655 TFLOPS), and more bandwidth (512.0 vs 320.3 GB/s). Its 300 W TDP and 700 W PSU requirement suggest a card designed for sustained server workloads, though the P104-100's lower power footprint (200 W suggested PSU) may be preferable in constrained environments. The FirePro's PCIe 3.0 x16 interface is vastly superior to the P104-100's PCIe 1.0 x4, which could bottleneck data transfer in some scenarios.

For buyers prioritizing OpenCL or Vulkan compute performance, the P104-100 is the clear choice based on benchmark evidence. For those needing higher theoretical throughput or relying on HBM's bandwidth characteristics, the FirePro presents a valid alternative, albeit one that loses every direct comparison in this dataset. The FirePro's launch MSRP of 5,999 USD also positions it as a premium server part, while the P104-100's lack of a listed MSRP suggests a different market channel.

Where Each One Wins

NVIDIA P104-100: This card wins every benchmark where both are tested. Its Geekbench OpenCL score of 51,663 is 84.7% higher than the FirePro's, making it the dominant choice for OpenCL-based compute tasks. Its Vulkan score of 45,165 is 21.7% higher, so it also leads in Vulkan compute. The P104-100's higher pixel rate (110.9 GPixel/s) and smaller die (314 mm²) suggest efficiency advantages. Its 16 nm process and 22.9M/mm² density indicate a more modern design. For applications that rely on OpenCL or Vulkan, the P104-100 is the data-backed winner.

AMD FirePro S9300 X2: The FirePro does not win any recorded benchmark, but it has theoretical advantages that could matter in unmeasured workloads. Its 512.0 GB/s memory bandwidth is 60% higher than the P104-100's, which could benefit memory-bound tasks. Its 7.987 TFLOPS FP32 output exceeds the NVIDIA card's 6.655 TFLOPS by 20%. The 4096-bit HBM bus and 256 TMUs give it a higher texture rate (249.6 GTexel/s). The PCIe 3.0 x16 interface is a significant upgrade over the P104-100's PCIe 1.0 x4. For workloads that scale with shading units or memory bandwidth — and that do not appear in the Geekbench tests — the FirePro could potentially outperform. The data does not confirm this, but the specification sheet leaves that door open.