NVIDIA GeForce RTX 4070 Mobile vs NVIDIA P104-100 Comparison

Where Each One Wins

The benchmark data splits cleanly by workload type, and it is not subtle. The NVIDIA GeForce RTX 4070 Mobile takes both recorded head-to-head tests, and it does so by enormous margins. The P104-100, by contrast, has no head-to-head win in the database. That does not mean the older card is without merit; it means the only common test points favor the newer architecture overwhelmingly.

The RTX 4070 Mobile wins in OpenCL compute and Vulkan graphics, the two tests where both cards appear. Its OpenCL score of 109,197 versus 52,368 for the P104-100 represents a 52% gap in favor of the mobile card. The Vulkan result is even more lopsided: 108,367 versus 45,165, a 58.3% deficit for the P104-100. For any user prioritizing raw compute throughput or modern graphics API performance, the RTX 4070 Mobile is the only rational choice from this data.

The P104-100, however, is not a typical consumer graphics card. It is a mining-generation product with no display outputs. Its strengths, if any, would have to lie in workloads that do not appear in the shared benchmark suite. The database does record an average benchmark score of 32,982 for the P104-100, which places it slightly above its nearest rivals (NVIDIA T600 Mobile at 32,849, NVIDIA T550 Mobile at 33,161, RTX 3050 Mobile at 33,170, AMD Radeon Pro 570 at 33,207). Its 77th percentile ranking across all GPUs is actually higher than the RTX 4070 Mobile's 73rd percentile, a curious inversion that suggests the P104-100's limited benchmark set may be flattering it relative to its overall capabilities.

In practical terms, the RTX 4070 Mobile is the clear winner for anyone running modern games, compute workloads, or anything that benefits from ray tracing or tensor operations. The P104-100 is a niche product from a prior era, and the data shows it is outclassed in every shared metric.

Architecture Differences

The two GPUs come from different generations and entirely different design philosophies. The P104-100 uses the Pascal architecture on a 16 nm TSMC process, while the RTX 4070 Mobile uses Ada Lovelace on a 5 nm TSMC node. That process shrink alone explains much of the performance gap: the RTX 4070 Mobile packs 22,900 million transistors into a 188 mm² die, achieving a transistor density of 121.8 million per mm². The P104-100 has only 7,200 million transistors spread across a much larger 314 mm² die, for a density of 22.9 million per mm². The Ada chip is smaller physically yet holds more than three times the transistors.

Shading unit counts tell a similar story. The RTX 4070 Mobile has 4,608 shading units, 144 texture mapping units, and 48 ROPs. The P104-100 has 1,920 shading units, 120 TMUs, and 64 ROPs. The newer card has 2.4 times the shaders and 20% more TMUs, but fewer ROPs. That ROP deficit explains why the P104-100 actually has a higher pixel rate (110.9 GPixel/s versus 81.36 GPixel/s), even though the RTX 4070 Mobile wins in texture rate (244.1 GTexel/s versus 208.0 GTexel/s) and FP32 throughput (15.62 TFLOPS versus 6.655 TFLOPS).

The RTX 4070 Mobile also brings features the P104-100 lacks entirely: 36 ray tracing cores and 144 tensor cores. The P104-100 has neither. This is a generational leap, not an incremental improvement. The P104-100's FP16 performance is a token 104.0 GFLOPS (1:64 ratio), while the RTX 4070 Mobile delivers 15.62 TFLOPS FP16 at a full 1:1 ratio. For any AI or machine learning workload, that difference is decisive.

Memory configurations differ as well. The P104-100 has 4 GB of GDDR5X on a 256-bit bus, yielding 320.3 GB/s bandwidth. The RTX 4070 Mobile has 8 GB of GDDR6 on a 128-bit bus, but only 256.0 GB/s bandwidth. The older card actually has more memory bandwidth, a remnant of its mining-oriented design. The newer card compensates with larger capacity and faster effective memory speed (16 Gbps versus 10 Gbps).

Head-to-Head Benchmarks

Only two tests appear in both cards' benchmark sets: Geekbench OpenCL and Geekbench Vulkan. Both are decisive wins for the RTX 4070 Mobile.

In Geekbench OpenCL, the RTX 4070 Mobile scores 109,197 against 52,368 for the P104-100. That is a 52% advantage. The delta is so large that no amount of driver optimization or workload tuning could close it; the Ada Lovelace architecture simply has more than double the FP32 compute capacity (15.62 TFLOPS versus 6.655 TFLOPS) and a far more capable tensor pipeline. The P104-100's nearest rival in this benchmark context, the NVIDIA T600 Mobile, scores 32,849 on average, which is actually below the P104-100's own OpenCL score. The P104-100 is not a weak card for its era; it is just outmatched by a GPU two generations newer.

In Geekbench Vulkan, the gap widens further. The RTX 4070 Mobile scores 108,367, while the P104-100 manages 45,165. That is a 58.3% deficit. Vulkan is a low-level API that rewards modern hardware features, asynchronous compute, and efficient command processing. The Ada Lovelace architecture was designed with these capabilities in mind, while Pascal predates many of them. The RTX 4070 Mobile's nearest rivals in the overall database include the AMD Radeon RX 6700 XT (average score 27,425) and the NVIDIA GeForce RTX 3090 (27,565), both of which sit within 0.5% of the mobile card's average. That context matters: the RTX 4070 Mobile is competing with desktop-class GPUs, while the P104-100 is grouped with mobile workstation cards like the T600 and T550.

The aggregate average benchmark scores reinforce this. The RTX 4070 Mobile averages 27,435 across all its tests, while the P104-100 averages 32,982. That the P104-100 has a higher average is an artifact of its sparse benchmark set: only three tests, all of which are compute-heavy and none of which include the PassMark DirectX suite that drags down the RTX 4070 Mobile's average. The PassMark scores for the RTX 4070 Mobile are odd: DirectX 9 at 223, DirectX 10 at 116, DirectX 11 at 179, DirectX 12 at 85, plus G2D at 763 and G3D at 19,587. These are not comparable to the Geekbench tests, and they illustrate why average scores across different test sets can be misleading.

The Verdict

The data points to a single conclusion: the RTX 4070 Mobile is the superior GPU for virtually any modern workload. It wins both head-to-head benchmarks by margins of 52% and 58.3%, respectively. It has more than double the FP32 throughput, a full complement of ray tracing and tensor cores, twice the memory capacity, and a far more advanced manufacturing process. Its 73rd percentile ranking places it alongside desktop cards like the RTX 3090 and RX 6700 XT, which are only 0.5% and 0% away in average score, respectively.

The P104-100 is a historical artifact. Its 77th percentile ranking is higher than the RTX 4070 Mobile's, but that ranking is based on a much smaller and less demanding benchmark set. Its nearest rivals are low-end mobile workstation GPUs, and it sits within 0.7% of all of them. The P104-100 does have advantages in two narrow areas: memory bandwidth (320.3 GB/s versus 256.0 GB/s) and pixel fill rate (110.9 GPixel/s versus 81.36 GPixel/s). Anyone working with high-resolution pixel-heavy workloads that do not use modern APIs might find those traits useful, but the lack of display outputs makes the P104-100 unsuitable for any interactive graphics work.

Who should pick which? The RTX 4070 Mobile is the choice for gamers, content creators, AI researchers, or anyone who needs a current-generation GPU with ray tracing, tensor acceleration, and broad software compatibility. The P104-100 is only relevant for someone who already owns one and has a specific compute task that favors raw memory bandwidth and pixel throughput, and who does not need video output. The database offers no evidence that the P104-100 can compete with the RTX 4070 Mobile in any shared workload.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The P104-100 has an average benchmark score of 32,982, while the RTX 4070 Mobile averages 27,435. However, the P104-100's average is based on only three tests, all of which are Geekbench compute benchmarks, while the RTX 4070 Mobile's average includes PassMark DirectX tests that score much lower.

Q: Does the RTX 4070 Mobile support ray tracing?

A: Yes. The RTX 4070 Mobile has 36 ray tracing cores and 144 tensor cores. The P104-100 has neither, as it is a Pascal-generation mining GPU with no RT or tensor hardware.

Q: Which card has more memory bandwidth?

A: The P104-100 has higher memory bandwidth at 320.3 GB/s, thanks to a 256-bit bus and GDDR5X memory. The RTX 4070 Mobile has 256.0 GB/s from a 128-bit bus with GDDR6 memory, but it offers 8 GB capacity versus the P104-100's 4 GB.

Q: How do the two GPUs compare in Geekbench Vulkan?

A: The RTX 4070 Mobile scores 108,367 versus 45,165 for the P104-100, a 58.3% advantage. This is the largest recorded gap between the two cards in any shared benchmark.

Q: Is the P104-100 a good choice for gaming?

A: The database shows no gaming benchmarks for the P104-100, and it has no display outputs, meaning it cannot directly drive a monitor. Its architecture lacks ray tracing and tensor cores, so it would not support modern gaming features.

Q: What is the transistor count difference?

A: The RTX 4070 Mobile contains 22,900 million transistors on a 188 mm² die, while the P104-100 has 7,200 million transistors on a 314 mm² die. The RTX 4070 Mobile achieves a density of 121.8 million transistors per mm² versus 22.9 million for the P104-100.

Specification Differences

| Specification | NVIDIA P104-100 | NVIDIA GeForce RTX 4070 Mobile |

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

| Architecture | Pascal | Ada Lovelace |

| Process Node | 16 nm | 5 nm |

| Transistors | 7,200 million | 22,900 million |

| Die Size | 314 mm² | 188 mm² |

| Transistor Density | 22.9M / mm² | 121.8M / mm² |

| Base Clock | 1607 MHz | 1395 MHz |

| Boost Clock | 1733 MHz | 1695 MHz |

| Memory Size | 4 GB | 8 GB |

| Memory Type | GDDR5X | GDDR6 |

| Memory Bus Width | 256 bit | 128 bit |

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

| Memory Speed | 10 Gbps effective | 16 Gbps effective |

| Shading Units | 1920 | 4608 |

| TMUs | 120 | 144 |

| ROPs | 64 | 48 |

| RT Cores | None | 36 |

| Tensor Cores | None | 144 |

| Pixel Rate | 110.9 GPixel/s | 81.36 GPixel/s |

| Texture Rate | 208.0 GTexel/s | 244.1 GTexel/s |

| FP32 Performance | 6.655 TFLOPS | 15.62 TFLOPS |

| FP16 Performance | 104.0 GFLOPS (1:64) | 15.62 TFLOPS (1:1) |

| TDP | Not specified | 115 W |

| Slot Width | Dual-slot | IGP |

| Power Connectors | 1x 8-pin | None |

| Bus Interface | PCIe 1.0 x4 | PCIe 4.0 x8 |

| Display Outputs | No outputs | Portable Device Dependent |

| DirectX Support | 12 (12_1) | 12 Ultimate (12_2) |

| OpenGL Support | 4.6 | 4.6 |

| Vulkan Support | 1.4 | 1.4 |

| Release Date | 2017-12-11 | 2023-01-02 |

| Production Status | End-of-life | Active |

| Predecessor | None specified | GeForce 30 Mobile |

| Successor | None specified | GeForce 50 Mobile |