AMD Radeon RX 6700 vs NVIDIA P104-100 Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive sweep in favor of the AMD Radeon RX 6700 across all three shared benchmark tests. The NVIDIA P104-100 does not register a single win in any head-to-head comparison, with the AMD card taking all three tests by substantial margins.

In 3DMark Steel Nomad DX12, the AMD Radeon RX 6700 scores 2014 against the NVIDIA P104-100's 1413. That is a delta of 29.8% in favor of the AMD card. This test is particularly relevant for modern DirectX 12 workloads, as it stresses the full graphics pipeline including geometry processing and rasterization. The 601-point gap here indicates that the AMD architecture delivers a meaningfully higher sustained graphics throughput in contemporary titles.

The compute-oriented tests show an even wider separation. In Geekbench OpenCL, the AMD Radeon RX 6700 posts 97841, while the NVIDIA P104-100 manages only 52368. The delta is 46.5%, meaning the AMD card is nearly twice as fast in general-purpose compute workloads. This result aligns with the raw FP32 throughput figures in the database, where the AMD card is rated at 11.29 TFLOPS against the NVIDIA card's 6.655 TFLOPS.

Geekbench Vulkan follows the same pattern. The AMD Radeon RX 6700 scores 83974, while the NVIDIA P104-100 scores 45165. The 46.2% delta confirms that the AMD card's advantage persists across both compute APIs tested, not just within a single vendor's optimized path. The Vulkan result is particularly telling, as it reflects cross-platform API performance that is increasingly relevant for both gaming and professional applications.

Looking at the broader database context, the NVIDIA P104-100 holds a 77th percentile ranking among all GPUs, while the AMD Radeon RX 6700 sits at the 75th percentile. This is an interesting inversion: the older NVIDIA card ranks slightly higher in overall percentile despite losing every direct comparison. The explanation lies in the average benchmark score calculation. The P104-100 has an average score of 32982 across its three recorded benchmarks, while the RX 6700 averages 30433 across its eleven recorded tests. The RX 6700's inclusion of several lower-scoring legacy tests, such as PassMark DirectX 9 at 250 and PassMark DirectX 10 at 115, drags its average down despite its superior performance in the tests where both cards were measured head-to-head.

The nearest rivals in the database further contextualize the P104-100's standing. The NVIDIA T600 Mobile averages 32849, just 0.4% below the P104-100. The NVIDIA T550 Mobile averages 33161, 0.5% above. The GeForce RTX 3050 Mobile averages 33170, 0.6% above. The AMD Radeon Pro 570 averages 33207, 0.7% above. These four rivals are all within a single percentage point of the P104-100, indicating that its average score places it in a tightly clustered performance tier.

The RX 6700's nearest rivals tell a different story. The NVIDIA CMP 70HX averages 30476, just 0.1% above the RX 6700. The Tesla M60 averages 30490, 0.2% above. The AMD Radeon RX 6800 averages 30095, which is 1.1% below the RX 6700. The GeForce RTX 3070 Ti averages 29945, 1.6% below. This places the RX 6700 at the top of its immediate cluster, slightly ahead of cards that are generally considered higher-tier products in their respective lineups.

Architecture Differences

The two cards represent fundamentally different design philosophies from different eras of GPU development. The NVIDIA P104-100 is built on the Pascal architecture using the GP104 chip, fabricated on a 16 nm TSMC process. The AMD Radeon RX 6700 uses the RDNA 2.0 architecture with the Navi 22 chip, fabricated on a 7 nm TSMC process. The process node difference is significant, as the 7 nm node allows for substantially higher transistor density.

The transistor counts reflect this manufacturing advantage. The P104-100 packs 7,200 million transistors into a 314 mm² die, yielding a density of 22.9 million transistors per square millimeter. The RX 6700 contains 17,200 million transistors in a 335 mm² die, achieving 51.3 million transistors per square millimeter. The RX 6700 fits more than twice the transistor count into a die that is only 21 mm² larger, a direct consequence of the denser manufacturing process.

Clock speeds also favor the AMD card substantially. The P104-100 has a base clock of 1607 MHz and a boost clock of 1733 MHz. The RX 6700 has a base clock of 1941 MHz, a game clock of 2174 MHz, and a boost clock of 2450 MHz. The boost clock advantage alone is over 700 MHz, which partially explains why the RX 6700 achieves such higher throughput despite having only 2304 shading units against the P104-100's 1920.

Memory configurations differ in capacity and type but are nearly identical in bandwidth. The P104-100 uses 4 GB of GDDR5X on a 256 bit bus, achieving 320.3 GB/s. The RX 6700 uses 10 GB of GDDR6 on a 160 bit bus, achieving 320.0 GB/s. The bandwidth delta is negligible at 0.3 GB/s, but the capacity difference is substantial. The RX 6700 offers 6 GB more memory, which matters increasingly for modern workloads and higher resolution textures. The RX 6700 also runs its memory at 2000 MHz with 16 Gbps effective speed, versus the P104-100's 1251 MHz with 10 Gbps effective speed.

The compute resources show the AMD card's architectural sophistication. The RX 6700 has 2304 shading units, 144 texture mapping units, and 64 ROPs. It also includes 36 ray tracing cores, a feature entirely absent from the P104-100. The RDNA 2.0 architecture supports hardware-accelerated ray tracing, which the Pascal architecture predates. The pixel rate is 156.8 GPixel/s for the RX 6700 versus 110.9 GPixel/s for the P104-100. The texture rate is 352.8 GTexel/s versus 208.0 GTexel/s.

Floating point performance tells a clear story. The RX 6700 delivers 11.29 TFLOPS of FP32 compute, while the P104-100 delivers 6.655 TFLOPS. The FP16 comparison is even more dramatic: the RX 6700 achieves 22.58 TFLOPS at a 2:1 ratio, while the P104-100 manages only 104.0 GFLOPS at a 1:64 ratio. The Pascal card's FP16 throughput is effectively negligible, while RDNA 2.0 offers full-rate half-precision compute.

API support differs in one notable respect. Both cards support DirectX 12 and Vulkan 1.4. However, the RX 6700 supports DirectX 12 Ultimate (12_2), which includes features like mesh shaders and variable rate shading, while the P104-100 only supports DirectX 12 (12_1). The RX 6700 also supports OpenGL 4.6, matching the P104-100.

The interface and connectivity differences are stark. The P104-100 is a mining-focused card with no display outputs whatsoever. It uses a PCIe 1.0 x4 bus interface, which severely limits host communication bandwidth. The RX 6700 uses PCIe 4.0 x16 and provides 1x HDMI 2.1 and 3x DisplayPort 1.4a outputs. The P104-100's suggested PSU is 200 W, while the RX 6700 requires a 450 W suggested PSU and carries a 175 W TDP. The P104-100 has no recorded TDP in the database. Both cards are dual-slot designs with a single 8-pin power connector and identical 267 mm (10.5 inches) length.

The RX 6700 also has dimensions of 110 mm in height and 40 mm in width, while the P104-100's height and width are not recorded. The P104-100 was released on December 11, 2017, while the RX 6700 came later on June 8, 2021. Both are marked as end-of-life production status. The RX 6700's predecessor is listed as Navi, with Navi III as its successor. The P104-100 has no recorded predecessor or successor.

The Verdict

The data is unambiguous. The AMD Radeon RX 6700 outperforms the NVIDIA P104-100 in every single benchmark where both cards were measured, with deltas ranging from 29.8% to 46.5%. The RX 6700 also offers more memory, hardware ray tracing support, modern API features, and display outputs. The P104-100 has no display outputs at all, making it unsuitable for any interactive use case.

The P104-100 is a mining-oriented card from 2017. Its only advantages in the recorded data are its lower 200 W suggested PSU requirement and its slightly higher percentile ranking of 77 versus 75, which is an artifact of the different benchmark suites applied to each card. The average benchmark score of 32982 for the P104-100 versus 30433 for the RX 6700 reflects test selection, not real-world performance parity.

For any workload that involves rendering to a display, gaming, ray tracing, or modern API features, the RX 6700 is the only viable choice from this data. The P104-100 cannot output video at all, so any comparison involving visual output is moot. For compute workloads, the RX 6700's 46.5% OpenCL advantage and 46.2% Vulkan advantage make it the clear pick. The RX 6700 also has 10 GB of memory versus 4 GB, which matters for larger datasets and textures.

The RX 6700's nearest rivals include the GeForce RTX 3070 Ti at 1.6% behind, which suggests the RX 6700 competes with higher-tier cards in the broader database. The P104-100's nearest rivals are all mobile or professional variants within a 0.7% band, indicating it sits in a much lower performance tier despite its 77th percentile ranking.

FAQ

Q: Which card wins in 3DMark Steel Nomad DX12?

A: The AMD Radeon RX 6700 wins with a score of 2014 against the NVIDIA P104-100's 1413, a 29.8% advantage.

Q: How much faster is the RX 6700 in OpenCL compute?

A: The RX 6700 scores 97841 in Geekbench OpenCL versus 52368 for the P104-100, a delta of 46.5%.

Q: Does the P104-100 support display output?

A: No, the P104-100 has no display outputs. The RX 6700 provides 1x HDMI 2.1 and 3x DisplayPort 1.4a.

Q: What is the memory capacity difference?

A: The RX 6700 has 10 GB of GDDR6, while the P104-100 has 4 GB of GDDR5X. Their bandwidth is nearly identical at 320.0 GB/s versus 320.3 GB/s respectively.

Q: Does the RX 6700 support ray tracing?

A: Yes, the RX 6700 includes 36 ray tracing cores. The P104-100 has no ray tracing support.

Q: How do their percentile rankings compare?

A: The P104-100 ranks at the 77th percentile among all GPUs, while the RX 6700 ranks at the 75th percentile. This is despite the RX 6700 winning every head-to-head test.

Where Each One Wins

The AMD Radeon RX 6700 wins across all measured categories. In 3DMark Steel Nomad DX12, it is 29.8% ahead. In Geekbench Vulkan, it is 46.2% ahead. In Geekbench OpenCL, it is 46.5% ahead. The RX 6700 also wins on memory capacity with 10 GB versus 4 GB, on compute throughput with 11.29 TFLOPS versus 6.655 TFLOPS FP32, and on feature support with DirectX 12 Ultimate, ray tracing cores, and display outputs.

The NVIDIA P104-100 wins in no measured benchmark. Its only advantages in the recorded data are a lower 200 W suggested PSU and a higher percentile ranking of 77 versus 75, with the latter being a statistical artifact of different test suites. The P104-100's average benchmark score of 32982 is higher than the RX 6700's 30433, but this is because the RX 6700 was tested with eleven benchmarks including several low-scoring legacy PassMark tests, while the P104-100 was tested with only three modern benchmarks.

For use cases involving gaming, the RX 6700 is the clear choice. For compute workloads, the RX 6700's near-double OpenCL score makes it dominant. For any workload requiring video output, the RX 6700 is the only option. The P104-100 appears suited only for specialized compute tasks where its 200 W suggested PSU and mining-oriented design might fit a constrained power envelope, but even then, its lower compute scores and 4 GB memory limit make it difficult to recommend from the recorded data.