AMD Radeon RX 590 GME vs NVIDIA P104-100 Comparison

The NVIDIA P104-100 and AMD Radeon RX 590 GME are both end-of-life graphics cards that land in the same performance tier, yet they achieve parity through radically different design priorities. The data shows a split decision: the P104-100 dominates in two of three benchmarks, while the RX 590 GME counters with a decisive victory in the Vulkan workload. Their average benchmark scores are nearly identical — 32,982 for the NVIDIA part versus 32,601 for the AMD part — placing both at the 77th percentile of all GPUs. This is a comparison of two very different interpretations of the same performance class.

Head-to-Head Benchmarks

The NVIDIA P104-100 wins the 3DMark Steel Nomad DX12 test decisively, scoring 1,413 points against the RX 590 GME's 1,001 points. That is a 41.2% advantage, a massive gap in a modern DirectX 12 workload. The P104-100 also takes the Geekbench OpenCL test with a score of 52,368 versus 36,294 for the AMD card, a 44.3% margin. In raw compute throughput, the P104-100 is the clear leader, and these two benchmark victories show that its architecture is far more efficient in these specific workloads.

The RX 590 GME's lone win comes in the Geekbench Vulkan test, where it scores 60,508 against the P104-100's 45,165. That is a 25.4% reversal, and it is a significant one. Vulkan is a low-level API that rewards different architectural strengths, and the AMD card's GCN 4.0 design clearly responds better to that workload. This single result prevents the NVIDIA card from taking a clean sweep and shows that the RX 590 GME has genuine competitive advantages in the right conditions.

The overall averages tell a slightly different story. The P104-100's average benchmark score of 32,982 puts it just 0.4% ahead of the NVIDIA T600 Mobile (32,849) and 0.5% behind the T550 Mobile (33,161). The RX 590 GME's average of 32,601 places it 0.2% ahead of the AMD FirePro S9300 X2 (32,540) and 0.4% ahead of the RX 7900 GRE (32,456). Both cards sit in a tight cluster of mid-range performers, but the head-to-head deltas show that the P104-100 has the higher ceiling in most tests.

Architecture Differences

The architectural divide is stark. The P104-100 uses the GP104 chip on TSMC's 16 nm process, packing 7,200 million transistors into a 314 mm² die. The RX 590 GME uses the Polaris 20 chip on GlobalFoundries' 14 nm process, with 5,700 million transistors on a 232 mm² die. The NVIDIA chip has a higher transistor density of 22.9M per mm², while AMD's Polaris 20 achieves a denser 24.6M per mm² despite having fewer total transistors.

Clock speeds favor the NVIDIA card significantly. The P104-100 runs at a 1,607 MHz base clock and boosts to 1,733 MHz, while the RX 590 GME operates at 1,257 MHz base and 1,420 MHz boost. Memory configurations also diverge: the P104-100 uses 4 GB of GDDR5X memory at 1,251 MHz (10 Gbps effective) delivering 320.3 GB/s of bandwidth, while the RX 590 GME uses 8 GB of GDDR5 at 2,000 MHz (8 Gbps effective) for 256.0 GB/s. Both use a 256-bit memory bus, but the NVIDIA card's faster memory type gives it a 25% bandwidth advantage.

Shader configurations differ in composition. The P104-100 has 1,920 shading units, 120 texture mapping units, and 64 ROPs. The RX 590 GME has more shading units at 2,304, more TMUs at 144, but only 32 ROPs. This explains the pixel rate disparity: the P104-100 delivers 110.9 GPixel/s versus the AMD card's 45.44 GPixel/s. Texture rates are nearly identical — 208.0 GTexel/s for NVIDIA versus 204.5 GTexel/s for AMD. FP32 performance is essentially a tie at 6.655 TFLOPS for the P104-100 and 6.543 TFLOPS for the RX 590 GME, but FP16 is where they diverge completely: the P104-100 manages only 104.0 GFLOPS (1:64 ratio), while the RX 590 GME delivers 6.543 TFLOPS (1:1 ratio).

The cards also differ in connectivity and power. The P104-100 uses a PCIe 1.0 x4 interface and has no display outputs — it is a mining-focused card. The RX 590 GME uses PCIe 3.0 x16 and has 1x HDMI 2.0b plus 3x DisplayPort 1.4a outputs. The P104-100 has no official TDP listed but suggests a 200 W power supply, while the RX 590 GME has a 175 W TDP and recommends a 450 W power supply. Both are dual-slot cards with a single 8-pin power connector. API support is similar, with both offering DirectX 12 (12_1 for NVIDIA, 12_0 for AMD), OpenGL 4.6, and Vulkan 1.4 for NVIDIA versus 1.3 for AMD.

Where Each One Wins

The P104-100 is the clear choice for DirectX 12 workloads and general compute tasks. Its 41.2% lead in 3DMark Steel Nomad DX12 and 44.3% lead in Geekbench OpenCL demonstrate that it handles modern APIs and raw compute with far greater efficiency. The card's higher clock speeds, superior memory bandwidth, and double the ROP count make it a stronger performer in rasterization-heavy scenarios. Its lack of display outputs means it is strictly a compute or mining card, but within that scope, it is the more powerful option.

The RX 590 GME wins in Vulkan workloads, with a 25.4% advantage in Geekbench Vulkan. This suggests that its GCN 4.0 architecture has strengths in low-level API scenarios that the Pascal-based P104-100 cannot match. The AMD card also offers 8 GB of memory versus 4 GB, which is a significant capacity advantage for workloads that exceed the smaller frame buffer. Its full display output suite — HDMI 2.0b and three DisplayPort 1.4a connections — makes it a functional graphics card for actual use, whereas the P104-100 cannot drive a monitor at all. The RX 590 GME also has full-rate FP16 performance at 6.543 TFLOPS, which could benefit certain compute applications that utilize half-precision math.

FAQ

Q: Which card has the higher average benchmark score?

A: The NVIDIA P104-100 has a slightly higher average benchmark score of 32,982 compared to the AMD Radeon RX 590 GME's 32,601, a difference of roughly 1.2%.

Q: How much faster is the P104-100 in DirectX 12?

A: The P104-100 scores 1,413 in 3DMark Steel Nomad DX12, which is 41.2% higher than the RX 590 GME's score of 1,001.

Q: Does the RX 590 GME win any benchmarks?

A: Yes, the RX 590 GME wins the Geekbench Vulkan test with a score of 60,508, beating the P104-100's 45,165 by 25.4%.

Q: What are the memory capacities of each card?

A: The P104-100 has 4 GB of GDDR5X memory, while the RX 590 GME has 8 GB of GDDR5 memory. Both use a 256-bit memory bus.

Q: Can either card be used for display output?

A: Only the RX 590 GME has display outputs, featuring 1x HDMI 2.0b and 3x DisplayPort 1.4a. The P104-100 has no display outputs.

Q: How do the transistor counts compare?

A: The P104-100 contains 7,200 million transistors on a 314 mm² die, while the RX 590 GME contains 5,700 million transistors on a 232 mm² die.

The Verdict

The data points to a clear split: the NVIDIA P104-100 is the stronger compute performer, winning the two most demanding benchmarks by margins of 41.2% and 44.3%. Its higher clocks, faster GDDR5X memory, and double ROP count give it a decisive edge in DirectX 12 and OpenCL scenarios. For any workload that prioritizes raw compute performance or frame throughput, the P104-100 is the superior card.

The AMD Radeon RX 590 GME is the more versatile card. It wins the Vulkan benchmark decisively, offers double the memory capacity at 8 GB, and includes full display outputs for actual graphics use. Its full-rate FP16 capability is a distinct architectural advantage that the P104-100 lacks entirely. For users who need a functional graphics card with modern display connectivity, the RX 590 GME is the only viable option between the two.

Both cards share the same 77th percentile ranking and similar average scores, so the choice comes down to workload. The P104-100 is a specialized compute accelerator with no display support, excelling in performance metrics but lacking usability. The RX 590 GME is a balanced graphics card that sacrifices some raw speed in DirectX and OpenCL for Vulkan superiority, more memory, and full display functionality. The data does not crown an overall winner — it identifies two cards with complementary strengths.

Specification Differences

| Specification | NVIDIA P104-100 | AMD Radeon RX 590 GME |

|---|---|---|

| Process Node | 16 nm | 14 nm |

| Transistors | 7,200 million | 5,700 million |

| Die Size | 314 mm² | 232 mm² |

| Base Clock | 1607 MHz | 1257 MHz |

| Boost Clock | 1733 MHz | 1420 MHz |

| Memory Size | 4 GB | 8 GB |

| Memory Type | GDDR5X | GDDR5 |

| Memory Clock | 1251 MHz (10 Gbps effective) | 2000 MHz (8 Gbps effective) |

| Memory Bandwidth | 320.3 GB/s | 256.0 GB/s |

| Shading Units | 1920 | 2304 |

| TMUs | 120 | 144 |

| ROPs | 64 | 32 |

| Pixel Rate | 110.9 GPixel/s | 45.44 GPixel/s |

| Texture Rate | 208.0 GTexel/s | 204.5 GTexel/s |

| FP32 | 6.655 TFLOPS | 6.543 TFLOPS |

| FP16 | 104.0 GFLOPS (1:64) | 6.543 TFLOPS (1:1) |

| TDP | Not specified | 175 W |

| Suggested PSU | 200 W | 450 W |

| Bus Interface | PCIe 1.0 x4 | PCIe 3.0 x16 |

| Display Outputs | No outputs | 1x HDMI 2.0b, 3x DisplayPort 1.4a |

| DirectX | 12 (12_1) | 12 (12_0) |

| Vulkan | 1.4 | 1.3 |

| Length | 267 mm (10.5 inches) | 241 mm (9.5 inches) |

| Release Date | 2017-12-11 | 2020-03-08 |

| Predecessor | Not specified | Arctic Islands |

| Successor | Not specified | Vega |