AMD Radeon HD 8870M vs NVIDIA GeForce GTX 960 Comparison

# NVIDIA GeForce GTX 960 vs AMD Radeon HD 8870M

The database places these two mobile and desktop graphics solutions in an interesting position. The GTX 960 is a discrete desktop card from the Maxwell 2.0 generation, while the HD 8870M is a mobile part built on the GCN 1.0 architecture. They are not direct contemporaries in the usual sense, but their benchmark results place them close enough for a meaningful comparison. The recorded data shows a single head-to-head benchmark, and the results are lopsided in favor of NVIDIA.

Head-to-Head Benchmarks

The database includes one direct comparison, the Geekbench OpenCL test. In this workload, the NVIDIA GeForce GTX 960 scores 18,925 points, while the AMD Radeon HD 8870M scores 8,462 points. That is a delta of 123.6% in favor of the GTX 960. This is the largest win in the entire comparison and it is a decisive one. The GTX 960 effectively doubles the performance of the HD 8870M in this compute-oriented test.

The average benchmark score across all recorded workloads further confirms the separation. The GTX 960 achieves an average score of 9,273, placing it in the 45th percentile of all GPUs in the database. The HD 8870M, by contrast, averages 8,462, which lands it in the 43rd percentile. The raw score gap is roughly 811 points, about 9.6% higher for the NVIDIA part. This is a modest but consistent advantage, not a generational leap.

Looking at the nearest rivals, the GTX 960 sits close to the NVIDIA GeForce GTX 465, which averages 9,294 (0.2% higher). It also trades blows with the AMD Radeon R7 M380, which is 0.4% slower, and the AMD Radeon Vega 8, which is 0.6% slower. The HD 8870M, on the other hand, is close to the NVIDIA GeForce GTX 850M, which scores 9,302 (0.3% higher). It is also near the NVIDIA GeForce GTX 675MX, which is 0.4% faster. These are all minor differences, within the range of clock binning and driver variation.

Architecture Differences

The GTX 960 is built on the GM206 chip, using the Maxwell 2.0 architecture. It is fabricated on a 28 nm process at TSMC, with a die size of 228 mm² and 2,940 million transistors. The HD 8870M, meanwhile, uses the Venus chip, based on Graphics Core Next 1.0. It is also a 28 nm TSMC part, but the die is smaller at 123 mm² and has 1,500 million transistors. This makes the transistor density of the NVIDIA part notably higher: 12.9 million per mm² against 12.2 million per mm² for the AMD part.

Clock behavior differs. The GTX 960 has a base clock of 1127 MHz and a boost clock of 1178 MHz. The AMD part runs at a lower clock speed, with a base of 725 MHz and a boost of 775 MHz. Memory clocks are even more distinct: the GTX 960 runs its 2 GB of GDDR5 at 1753 MHz, while the HD 8870M uses 2 GB of GDDR5 at 1125 MHz. Effective memory bandwidth is rated at 7 Gbps for the NVIDIA part versus 4.5 Gbps on the AMD.

The NVIDIA card has a 128 bit memory bus, while AMD uses a 128 bit bus. This yields a bandwidth of 112.2 GB/s for the GTX 960 and 72.00 GB/s for the HD 8870M. This is a clear and substantial memory bandwidth advantage for the GTX 960.

The shading and pixel processing hardware also differ. The GTX 960 has 1024 shader cores, 64 TMUs, and 32 ROPs. The HD 8870M has 640 shader units, 40 TMUs, and 16 ROPs. The NVIDIA part is therefore much larger on every one of these fronts. Pixel and texture rates are correspondingly higher: 37.70 GPixel/s and 75.39 GTexel/s for the GTX 960, versus 12.40 GPixel/s and 31.00 GTexel/s for the HD 8870M. The GTX 960 has a FP32 performance of 2.413 TFLOPS; the HD 8870M is rated at 992.0 GFLOPS.

The NVIDIA chip supports a wider set of APIs. It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The AMD part supports DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170. There is no RT core or Tensor Core on either.

Where Each One Wins

The GTX 960 is the clear compute leader. That 123.6% lead in the OpenCL head-to-head, plus the higher FP32, texture, and pixel rates, makes it the choice for GPU-bound compute. The database shows a higher average benchmark score, a higher percentile rank among all GPUs, and a win in the only direct comparison. If the workload is heavily GPGPU oriented, or if the application can use the faster memory bandwidth and newer API feature levels, the GTX 960 is the one to get.

The AMD HD 8870M is not without its merits. It posts a positive delta over the GTX 960 in four rival comparisons: it is 0.1% ahead of the GeForce GTX 465, 0.3% ahead of the Radeon R7 M380, and 0.4% ahead of the Radeon Vega 8. It is also 1.1% ahead of the Intel Arc A380. These are all small wins, however, and they put it in the same performance class as the GTX 960. The Radeon is the more power-efficient part: it has a lower thermal design point, and its power connector is not listed in the database. For a system with limited power delivery, or where the machine is already existing and the card is a drop-in, the AMD part may be a more practical choice. It also provides a lower average score, but that can be a benefit in thermally constrained chassis.

For gaming at 1080p on higher settings, the GTX 960 has the edge in raw shading throughput and memory bandwidth. The extra TMUs and ROPs give it a more headroom in geometry and pixel fill. Its higher memory clock and bandwidth are also useful for high-resolution textures and larger framebuffers.

The HD 8870M is a better match for a light, thermally limited chassis, or for a system where the goal is modest 1080p gaming with lower power draw. It is still a competent GPU, and it will handle a typical e-sports title at medium to high settings. The GTX 960 on the other hand, is the pick for anything more demanding.

FAQ

Q: Which GPU is faster in the recorded benchmark data?

A: The NVIDIA GeForce GTX 960 is faster. It wins the only direct head-to-head benchmark, the Geekbench OpenCL test, with a score of 18,925 versus 8,462 for the AMD Radeon HD 8870M, a 123.6% advantage.

Q: What is the average benchmark score difference?

A: The GTX 960 has an average score of 9,273, versus 8,462 for the HD 8870M. The NVIDIA part averages about 9.6% higher.

Q: How do the memory subsystems compare?

A: The GTX 960 has 2 GB of GDDR5 on a 128-bit bus, running at 1753 MHz, for a bandwidth of 112.2 GB/s. The HD 8870M has 2 GB of GDDR5 on a 128-bit bus at 1125 MHz, giving 72.00 GB/s of bandwidth.

Q: Which GPU supports more modern graphics APIs?

A: The NVIDIA GeForce GTX 960 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The AMD Radeon HD 8870M supports DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170.

Specification Differences

| Specification | NVIDIA GeForce GTX 960 | AMD Radeon HD 8870M |

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

| Chip | GM206 | Venus |

| Architecture | Maxwell 2.0 | GCN 1.0 |

| Process | 28 nm | 28 nm |

| Transistors | 2,940 million | 1,500 million |

| Die size | 228 mm² | 123 mm² |

| Transistor density | 12.9M / mm² | 12.2M / mm² |

| Base clock | 1127 MHz | 725 MHz |

| Boost clock | 1178 MHz | 775 MHz |

| Memory clock | 1753 MHz | 1125 MHz |

| Memory size | 2 GB | 2 GB |

| Memory type | GDDR5 | GDDR5 |

| Memory bus | 128 bit | 128 bit |

| Bandwidth | 112.2 GB/s | 72.00 GB/s |

| Shading units | 1024 | 640 |

| TMUs | 64 | 40 |

| ROPs | 32 | 16 |

| Pixel rate | 37.70 GPixel/s | 12.40 GPixel/s |

| Texture rate | 75.39 GTexel/s | 31.00 GTexel/s |

| FP32 | 2.413 TFLOPS | 992.0 GFLOPS |

| Bus interface | PCIe 3.0 x16 | PCIe 3.0 x16 |

| APIs | DirectX 12 (12_1), OpenGL 4.6, Vulkan 1.4 | DirectX 12 (11_1), OpenGL 4.6, Vulkan 1.2.170 |

The Verdict

The data is unambiguous: the GTX 960 is the higher-performance part in almost every measurable way. It wins the only direct benchmark, has a higher average score, a bigger shading core, more texture units and ROPs, and a much higher memory bandwidth. The Radeon HD 8870M is only competitive in a few near-rival cases, and those wins are within a percentage point or two. The GTX 960 is for the user who wants maximum throughput and the latest API support. The HD 8870M is for the user who wants a capable but less power-hungry GPU in a space-constrained system, and who does not need the extra throughput. From the database's perspective, the NVIDIA is the stronger overall choice.