AMD Radeon HD 8870M vs NVIDIA GeForce GTX 650 Ti Comparison

The AMD Radeon HD 8870M and the NVIDIA GeForce GTX 650 Ti are both end-of-life graphics solutions built on a 28 nm TSMC process, yet the benchmark data reveals a clear edge for the AMD mobile part. In the sole head-to-head benchmark available, the Geekbench OpenCL test, the HD 8870M scores 8462 points against the GTX 650 Ti’s 7877 points, a 7.4% advantage. This win places the AMD part in the 43rd percentile of all GPUs, while the NVIDIA card sits just behind at the 42nd percentile. The data indicates that despite being a mobile-oriented chip, the HD 8870M delivers higher raw compute performance than this desktop-class NVIDIA offering.

Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test, and the result is unambiguous. The AMD Radeon HD 8870M achieves a score of 8462, while the NVIDIA GeForce GTX 650 Ti trails at 7877. The delta of 7.4% in favor of AMD is significant enough to separate the two products in the performance hierarchy. Looking at the broader context, the HD 8870M’s score places it nearly even with the NVIDIA GeForce MX330, which averages 8458 (a 0.1% difference), and slightly ahead of the AMD Radeon 880M at 8436. The GTX 650 Ti’s score of 7877, in contrast, puts it in the company of the NVIDIA GeForce GTX 880M (8040) and the NVIDIA Quadro P5000 (8039), both of which it edges out by 0.2%.

The GTX 650 Ti does have a second benchmark result in the Geekbench Vulkan test, scoring 8229, but no corresponding Vulkan score exists for the HD 8870M in the data. This absence means the OpenCL comparison stands as the definitive measure between the two. The AMD card’s higher OpenCL performance is corroborated by its average benchmark score of 8462, compared to the GTX 650 Ti’s average of 8053 — a gap of roughly 5.1% across all recorded tests. While the GTX 650 Ti shows a stronger Vulkan result relative to its own OpenCL score (8229 vs 7877), the lack of a direct AMD Vulkan score prevents a cross-vendor comparison on that API.

In terms of raw throughput, the HD 8870M’s 7.4% lead in OpenCL translates to a tangible performance differential. The AMD part’s nearest rival, the Intel Arc A380, scores 8558 and is 1.1% faster, showing that the HD 8870M sits at the upper edge of its performance tier. The GTX 650 Ti’s closest competitor, the NVIDIA GeForce GTX 650 Ti Boost, is only 0.2% faster at 8067, indicating that the non-Boost model is very near the ceiling of its own class. Overall, the head-to-head data consistently favors AMD, with no benchmark in the pack showing a NVIDIA win.

Architecture Differences

The architectural split between these two GPUs is stark. The AMD Radeon HD 8870M is built on the GCN 1.0 architecture with the Venus chip, while the NVIDIA GeForce GTX 650 Ti uses the Kepler architecture with the GK106S chip. Both are fabricated by TSMC on a 28 nm process, but the silicon itself diverges significantly. AMD’s Venus die is 123 mm² and packs 1,500 million transistors, yielding a transistor density of 12.2 million per mm². NVIDIA’s GK106S is substantially larger at 221 mm², containing 2,540 million transistors, though its density is slightly lower at 11.5 million per mm².

The compute resources tell a story of divergent design priorities. The GTX 650 Ti has more shading units (768 vs 640), more texture mapping units (64 vs 40), but an equal number of raster output units (16 each). This configuration gives NVIDIA a commanding lead in texture throughput — 59.39 GTexel/s versus the AMD’s 31.00 GTexel/s — and a smaller edge in pixel rate (14.85 GPixel/s vs 12.40 GPixel/s). However, the AMD part compensates with higher FP32 compute in a per-clock sense; its 992.0 GFLOPS comes from a 725 MHz base and 775 MHz boost clock, while the GTX 650 Ti’s 1,425.4 GFLOPS is achieved without listed base or boost frequencies in the data.

Memory configurations differ in capacity but not width. The HD 8870M features 2 GB of GDDR5 on a 128-bit bus, running at 1125 MHz with 4.5 Gbps effective speed, yielding 72.00 GB/s of bandwidth. The GTX 650 Ti offers only 1024 MB of GDDR5 on the same 128-bit bus, but at a higher 1350 MHz (5.4 Gbps effective), achieving 86.40 GB/s — a 20% bandwidth advantage. API support is nearly identical, with both supporting DirectX 12 (though AMD lists 11_1 and NVIDIA lists 11_0), OpenGL 4.6, and Vulkan (1.2.170 for AMD, 1.2.175 for NVIDIA). Power characteristics are only disclosed for the NVIDIA card, which lists a 110 W TDP and a 300 W suggested PSU.

FAQ

Q: Which GPU has the higher Geekbench OpenCL score?

A: The AMD Radeon HD 8870M scores 8462, which is 7.4% higher than the NVIDIA GeForce GTX 650 Ti’s 7877.

Q: How does the memory bandwidth compare between the two cards?

A: The GTX 650 Ti has higher memory bandwidth at 86.40 GB/s, while the HD 8870M delivers 72.00 GB/s. Both use a 128-bit GDDR5 interface.

Q: What is the difference in transistor count and die size?

A: The GTX 650 Ti uses 2,540 million transistors on a 221 mm² die, whereas the HD 8870M uses 1,500 million transistors on a smaller 123 mm² die.

Q: Are there any benchmark tests where the GTX 650 Ti wins?

A: In the data provided, the GTX 650 Ti has no winning benchmark results. It has a Vulkan score of 8229, but no comparable Vulkan score exists for the HD 8870M.

Q: What is the launch MSRP of the GTX 650 Ti?

A: The NVIDIA GeForce GTX 650 Ti has a launch MSRP of 149 USD. The HD 8870M has no listed launch MSRP.

Q: How do the two cards rank in overall GPU percentile?

A: The HD 8870M is in the 43rd percentile of all GPUs, while the GTX 650 Ti is in the 42nd percentile.

The Verdict

Based strictly on the benchmark data, the AMD Radeon HD 8870M is the faster GPU. Its 7.4% lead in Geekbench OpenCL is the only direct comparison available, and it wins decisively. The HD 8870M also holds a higher average benchmark score (8462 vs 8053) and a better overall percentile ranking (43rd vs 42nd). For compute-heavy workloads using OpenCL, the AMD part is the clear choice.

The NVIDIA GeForce GTX 650 Ti does have compensating strengths in the specification sheet, particularly in memory bandwidth (86.40 GB/s vs 72.00 GB/s) and texture fill rate (59.39 GTexel/s vs 31.00 GTexel/s). However, these advantages do not translate into a benchmark victory in the data provided. The GTX 650 Ti’s higher FP32 throughput (1,425.4 GFLOPS vs 992.0 GFLOPS) and larger transistor budget (2,540 million vs 1,500 million) suggest theoretical superiority, yet the real-world OpenCL result contradicts this. The GTX 650 Ti also has a higher TDP of 110 W, which is a consideration for system power draw, but this is not a performance metric.

For users selecting between these two end-of-life parts, the data is unambiguous. The HD 8870M delivers better compute performance in the measured workload, and its higher percentile ranking confirms it sits in a slightly more favorable position relative to the entire GPU market. The GTX 650 Ti’s only potential edge is in its Vulkan score of 8229, but without an AMD counterpart, this cannot be evaluated as a comparative advantage.

Specification Differences

| Specification | AMD Radeon HD 8870M | NVIDIA GeForce GTX 650 Ti |

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

| Architecture | GCN 1.0 | Kepler |

| Chip | Venus | GK106S |

| Transistors | 1,500 million | 2,540 million |

| Die Size | 123 mm² | 221 mm² |

| Transistor Density | 12.2M / mm² | 11.5M / mm² |

| Base Clock | 725 MHz | Not listed |

| Boost Clock | 775 MHz | Not listed |

| Memory Clock | 1125 MHz (4.5 Gbps effective) | 1350 MHz (5.4 Gbps effective) |

| Memory Size | 2 GB | 1024 MB |

| Memory Bandwidth | 72.00 GB/s | 86.40 GB/s |

| Shading Units | 640 | 768 |

| TMUs | 40 | 64 |

| ROPs | 16 | 16 |

| Pixel Rate | 12.40 GPixel/s | 14.85 GPixel/s |

| Texture Rate | 31.00 GTexel/s | 59.39 GTexel/s |

| FP32 | 992.0 GFLOPS | 1,425.4 GFLOPS |

| TDP | Not listed | 110 W |

| Slot Width | Not listed | Single-slot |

| Power Connectors | Not listed | 1x 6-pin |

| Suggested PSU | Not listed | 300 W |

| Display Outputs | Not listed | 2x DVI, 1x mini-HDMI 1.4a |

| Length | Not listed | 145 mm (5.7 inches) |

| DirectX | 12 (11_1) | 12 (11_0) |

| Vulkan | 1.2.170 | 1.2.175 |

| Launch MSRP | Not listed | 149 USD |

| Release Date | 2013-03-31 | 2012-10-08 |

Where Each One Wins

The AMD Radeon HD 8870M wins in the only benchmark category where both have results: Geekbench OpenCL, with a 7.4% margin. This makes it the preferred choice for OpenCL-accelerated compute tasks, such as general-purpose GPU computing or applications that leverage this API for rendering or data processing. Its higher average benchmark score (8462 vs 8053) and better percentile ranking (43rd vs 42nd) further reinforce its position as the overall performance leader in this comparison. The HD 8870M also offers double the memory capacity (2 GB vs 1024 MB), which can be advantageous for workloads with larger memory footprints, even if the bandwidth is lower.

The NVIDIA GeForce GTX 650 Ti wins in several specification categories that could matter for specific use cases. Its memory bandwidth of 86.40 GB/s is 20% higher, which benefits memory-bound workloads like high-resolution texture streaming or certain gaming scenarios. The GTX 650 Ti’s texture fill rate of 59.39 GTexel/s is nearly double the HD 8870M’s 31.00 GTexel/s, giving it a clear edge in texturing-heavy applications. The NVIDIA card also has higher FP32 compute (1,425.4 GFLOPS vs 992.0 GFLOPS) and a larger transistor budget, suggesting better raw arithmetic capability. Its Vulkan score of 8229 indicates solid API 1.2.175 support, though this cannot be compared directly. Finally, the GTX 650 Ti is the only one with a disclosed power profile (110 W TDP) and a launch MSRP of 149 USD, making it the more documented product for system builders.