AMD FirePro D700 vs NVIDIA P106-100 Comparison

The AMD FirePro D700 and the NVIDIA P106-100 are two end-of-life GPUs built for entirely different jobs, and the recorded data shows the mining-focused NVIDIA card edging out AMD's workstation part in every shared compute benchmark. The D700, a 28 nm Tahiti-based FirePro from the GCN 1.0 era, posts an average benchmark score of 25842 and sits in the 71st percentile of all GPUs in the database. The P106-100, a Pascal mining card with no display outputs, averages 23249 and ranks in the 68th percentile. Yet in the two direct head-to-head tests the database holds for both cards, the P106-100 wins twice, taking Geekbench OpenCL by 34 percent and Geekbench Vulkan by 15 percent. The result is a comparison where the aggregate ranking and the direct matchup point in different directions, and where the right pick depends on whether raw compute throughput or the broader feature set matters more.

The Verdict

The data favors the P106-100 for anyone whose workload is purely compute. It wins both shared benchmarks, and it does so at a TDP of 120 W against the D700's 274 W, which means substantially more performance per watt of heat and power draw. It also carries a suggested PSU of just 300 W versus the D700's 600 W, and it draws power through a single 6-pin connector. For sustained OpenCL or Vulkan compute workloads, the recorded scores make the case clearly: 35951 versus 23716 in OpenCL, 32897 versus 27968 in Vulkan.

The D700 still earns its place in the database's 71st percentile, and its nearest rivals explain the context. It performs within a rounding error of the AMD FirePro W7100 (average score 25856, a gap of just 0.1 percent), the AMD Radeon R9 M395X (25891, 0.2 percent ahead), the NVIDIA GeForce RTX 3080 Ti Mobile (25740, 0.4 percent behind the D700), and the AMD Radeon Pro W5700 (25726, 0.5 percent behind). That is tightly packed company, and it includes cards from far newer generations, which is a respectable showing for a 28 nm part.

Where the D700 pulls ahead of the P106-100 is capability rather than benchmark throughput. It has six mini-DisplayPort outputs plus an SDI output, against zero outputs on the P106-100, so only the AMD card can actually drive displays. It offers a 384-bit memory bus with 263.0 GB/s of bandwidth against the P106-100's 192-bit bus and 192.2 GB/s, a 32 ROP advantage in memory-heavy or high-resolution output scenarios offset by its lower pixel rate, and a PCIe 3.0 x16 interface versus the NVIDIA card's PCIe 1.0 x16. If the workload involves display output, wide memory access, or modern host interconnect, the D700 is the only viable option of the two. If it is headless compute, the P106-100's numbers win.

Architecture Differences

These cards come from different design philosophies separated by a process-node jump. The FirePro D700 is built on TSMC's 28 nm process using the Tahiti chip, AMD's GCN 1.0 architecture, and belongs to the FirePro Data Center (Dx00) generation. It packs 4,313 million transistors into a 352 mm² die, a density of 12.3M per mm². The P106-100 uses the GP106 chip on Pascal, classified in the database under the Mining GPUs generation, fabricated on TSMC's 16 nm node. Its 4,400 million transistors fit into just 200 mm², giving a density of 22.0M per mm², nearly double the AMD card's transistor density on a die less than 60 percent of the size.

The core counts tell the rest of the story. The D700 has 2048 shading units and 128 TMUs against the P106-100's 1280 shading units and 80 TMUs, but the NVIDIA card runs much faster: 1506 MHz base and 1709 MHz boost, while the D700's core clocks are not specified in the database. The frequency advantage more than compensates for the smaller core count in raw compute, which is why the P106-100 posts 4.375 TFLOPS FP32 against the D700's 3.482 TFLOPS. The Pascal card also has more ROPs, 48 versus 32, and its texture rate of 136.7 GTexel/s beats the D700's 108.8 GTexel/s. Pixel output is more lopsided still: 82.03 GPixel/s for the P106-100 versus 27.20 GPixel/s for the D700, though with no display outputs the NVIDIA card can never use that advantage for graphics.

Neither card has RT cores or tensor cores. The D700 records no FP16 figure, while the P106-100 lists FP16 at 68.36 GFLOPS with a 1:64 ratio, indicating its half-precision throughput is severely throttled relative to FP32, a limitation worth noting for mixed-precision compute.

Feature support splits along generational lines. The P106-100 supports DirectX 12 at feature level 12_1 and Vulkan 1.4, while the D700 supports DirectX 12 at feature level 11_1 and Vulkan 1.2.170. Both support OpenGL 4.6. Both are end-of-life, both are dual-slot cards, and both carry 6 GB of GDDR5. The lineage differs: the D700 succeeds the FirePro Terascale line and was itself succeeded by Radeon Instinct, while the P106-100 has no listed predecessor or successor, fitting for a product built for a single purpose.

Head-to-Head Benchmarks

The database holds two direct comparisons, and the NVIDIA card takes both.

Geekbench OpenCL is the biggest gap. The P106-100 scores 35951 against the D700's 23716, a 34 percent win for NVIDIA. That margin aligns with the FP32 capability difference on paper: 4.375 TFLOPS versus 3.482 TFLOPS suggests roughly a 25 percent theoretical edge, and the measured result exceeds it, meaning the Pascal card extracts more real OpenCL throughput from its hardware than the raw shader math alone predicts.

Geekbench Vulkan is closer but still decisive. Here the P106-100 scores 32897 and the D700 scores 27968, a 15 percent NVIDIA win. The narrower gap suggests the D700's GCN architecture holds up better under Vulkan than under OpenCL, or that its wider memory subsystem, with 263.0 GB/s of bandwidth against 192.2 GB/s, helps in workloads where the 192-bit Pascal bus becomes a constraint.

The average benchmark scores complicate the picture slightly. The D700's average of 25842 exceeds the P106-100's 23249, and its 71st percentile ranking beats the NVIDIA card's 68th. The reason is that the D700's average draws on its full recorded test set, while the P106-100's average includes a 3DMark Steel Nomad DX12 score of 899 that drags its mean down despite its two strong Geekbench results. Since only the two Geekbench tests are shared, the head-to-head record of two wins for the P106-100 and zero for the D700 is the cleanest measure of relative performance, and it favors NVIDIA.

Context from rivals sharpens the read. The P106-100's average of 23249 lands it essentially level with the AMD Radeon Pro Vega 16 (23250, a 0 percent gap), the Radeon RX 6600M (23273, 0.1 percent ahead), the Radeon R9 M290X (23276, 0.1 percent ahead), and the Radeon AI PRO R9700 (23315, 0.3 percent ahead). The D700's average sits about 11 percent above its rival cluster's scores while the P106-100 sits a fraction below its own, but in the tests where they actually meet, NVIDIA wins.

Specification Differences

The cards share a foundry (TSMC), 6 GB of GDDR5 memory, a dual-slot footprint, OpenGL 4.6 support, end-of-life status, and the absence of RT and tensor cores. Everything else differs:

• Process node: 28 nm (D700) versus 16 nm (P106-100)
• Die size and density: 352 mm² at 12.3M per mm² versus 200 mm² at 22.0M per mm²
• Transistors: 4,313 million versus 4,400 million
• Shader, TMU, and ROP counts: 2048 / 128 / 32 versus 1280 / 80 / 48
• Core clocks: unspecified versus 1506 MHz base, 1709 MHz boost
• Memory clocks: 1370 MHz (5.5 Gbps effective) versus 2002 MHz (8 Gbps effective)
• Bus width and bandwidth: 384 bit, 263.0 GB/s versus 192 bit, 192.2 GB/s
• FP32 compute: 3.482 TFLOPS versus 4.375 TFLOPS; FP16 listed only for the P106-100
• Pixel and texture rates: 27.20 GPixel/s and 108.8 GTexel/s versus 82.03 GPixel/s and 136.7 GTexel/s
• TDP: 274 W versus 120 W; suggested PSU 600 W versus 300 W
• Power connectors: unspecified versus 1x 6-pin
• Bus interface: PCIe 3.0 x16 versus PCIe 1.0 x16
• Display outputs: 6x mini-DisplayPort 1.2 plus 1x SDI versus none
• DirectX and Vulkan: 12 (11_1) and 1.2.170 versus 12 (12_1) and 1.4
• Length: 279 mm (11 inches) versus 250 mm (9.8 inches)
• Release dates: January 2014 for the D700, June 2017 for the P106-100

FAQ

Q: Which card is faster in the shared benchmarks? A: The NVIDIA P106-100 wins both, taking Geekbench OpenCL 35951 to 23716 (34 percent) and Geekbench Vulkan 32897 to 27968 (15 percent).

Q: Which card has the higher average benchmark score and percentile? A: The FirePro D700, with an average of 25842 and 71st percentile ranking, versus 23249 and 68th percentile for the P106-100.

Q: Can the P106-100 drive a monitor? A: No. It has no display outputs at all. The D700 offers six mini-DisplayPort 1.2 outputs and one SDI output.

Q: Which card uses less power? A: The P106-100, at 120 W TDP with a single 6-pin connector and a 300 W suggested PSU. The D700 is rated at 274 W with a 600 W suggested PSU.

Q: Which has more memory bandwidth? A: The D700, at 263.0 GB/s on a 384-bit bus, against 192.2 GB/s on the P106-100's 192-bit bus. Both carry 6 GB of GDDR5.

Q: Which supports newer graphics APIs? A: The P106-100 supports DirectX 12 feature level 12_1 and Vulkan 1.4, while the D700 supports DirectX 12 feature level 11_1 and Vulkan 1.2.170. Both support OpenGL 4.6.