AMD Radeon HD 8850M vs NVIDIA GeForce GTX 680M Comparison

AMD Radeon HD 8850M and NVIDIA GeForce GTX 680M are both end-of-life mobile graphics solutions from the 28 nm era, but they occupy very different positions in the performance hierarchy. The data shows a clear single benchmark head-to-head, with the NVIDIA part taking the only win, yet the AMD chip’s overall average score is slightly higher due to how the benchmark suites are weighted. This analysis breaks down the measurable differences, architectural philosophies, and which GPU suits which workload based strictly on the provided facts.

Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test, and it is a decisive victory for the NVIDIA GeForce GTX 680M. The NVIDIA part scores 9,230 points, while the AMD Radeon HD 8850M manages 7,447 points. That translates to a 19.3% deficit for the AMD chip, a substantial gap that highlights the raw compute advantage of the GTX 680M in this specific workload. The GTX 680M’s 2.038 TFLOPS of FP32 throughput versus the HD 8850M’s 800.0 GFLOPS explains much of this margin, as OpenCL heavily exercises floating-point shader performance.

However, the overall benchmark picture is more nuanced. The HD 8850M has an average benchmark score of 7,447, while the GTX 680M’s average is 7,023. This discrepancy arises because the GTX 680M’s average includes its weaker Geekbench Metal score of 4,815, which drags down its mean despite the OpenCL win. The AMD part only has the single OpenCL result, so its average equals that score. In practical terms, the GTX 680M is the faster GPU in raw OpenCL compute, but the HD 8850M posts a higher average across all recorded tests.

Looking at the nearest rivals provides context for where each card sits. The HD 8850M’s 7,447 average is essentially tied with the Intel UHD Graphics 750 (7,441, a 0.1% delta) and the AMD Radeon R7 350 (7,425, a 0.3% delta). It is also within 0.3% of the NVIDIA GeForce GTX 1650 (7,472), meaning these four GPUs are functionally interchangeable in this benchmark range. The GTX 680M, with its 7,023 average, sits close to the NVIDIA T600 (7,035, a -0.2% delta) and the AMD Radeon R5 M240 (6,975, a 0.7% delta). Its 1.1% advantage over the GeForce GTX 675M (6,946) shows it is the stronger of the two mobile Kepler parts in this comparison.

The performance percentile rankings reinforce this split. The HD 8850M lands in the 40th percentile of all GPUs, while the GTX 680M is just below at the 39th percentile. Despite losing the head-to-head OpenCL test, the HD 8850M’s higher average score gives it a one-point edge in the overall ranking distribution. This is a reminder that a single benchmark does not define a GPU’s standing, especially when different API tests are involved.

FAQ

Q: Which GPU wins the direct OpenCL benchmark comparison?

A: The NVIDIA GeForce GTX 680M wins decisively, scoring 9,230 versus the AMD Radeon HD 8850M’s 7,447, a 19.3% advantage for NVIDIA.

Q: How do their average benchmark scores compare?

A: The AMD Radeon HD 8850M has a higher average score of 7,447, while the NVIDIA GeForce GTX 680M averages 7,023. The GTX 680M’s average is pulled down by its additional Metal benchmark score of 4,815.

Q: What are the closest rivals to the AMD Radeon HD 8850M?

A: The Intel UHD Graphics 750 (7,441, 0.1% delta) and AMD Radeon R7 350 (7,425, 0.3% delta) are the nearest competitors, with the NVIDIA GeForce GTX 1650 (7,472) just 0.3% ahead.

Q: What about the NVIDIA GeForce GTX 680M’s nearest rivals?

A: The NVIDIA T600 (7,035, -0.2% delta) is the closest, followed by the AMD Radeon R5 M240 (6,975, 0.7% delta) and the NVIDIA GeForce GTX 675M (6,946, 1.1% delta).

Q: How do their overall performance percentiles differ?

A: The AMD Radeon HD 8850M is in the 40th percentile of all GPUs, while the NVIDIA GeForce GTX 680M is in the 39th percentile, a one-point difference despite the NVIDIA card winning the direct benchmark.

Q: Does the GTX 680M have any advantage in the Metal API?

A: Yes, the GTX 680M has a recorded Geekbench Metal score of 4,815, whereas the HD 8850M has no Metal benchmark results in the data.

Architecture Differences

The two GPUs come from fundamentally different design philosophies. The AMD Radeon HD 8850M uses the Venus chip built on GCN 1.0 architecture, part of the Solar System generation (HD 8800M). The NVIDIA GeForce GTX 680M uses the GK104 chip based on Kepler architecture, from the GeForce 600M generation. Both are fabricated on a 28 nm process at TSMC, but their transistor budgets diverge sharply. The NVIDIA chip packs 3,540 million transistors into a 294 mm² die, while the AMD chip uses just 1,500 million transistors on a much smaller 123 mm² die. This gives the GTX 680M a 12.0M transistors per mm² density, slightly lower than the HD 8850M’s 12.2M / mm².

The compute resources tell the story of the performance gap. NVIDIA’s GK104 features 1,344 shading units, 112 texture mapping units, and 32 ROPs. AMD’s Venus is far leaner with 640 shading units, 40 TMUs, and 16 ROPs. This 2.1x advantage in shaders and 2.8x in TMUs directly translates to the GTX 680M’s higher pixel rate of 21.22 GPixel/s versus 10.00 GPixel/s, and its texture rate of 84.90 GTexel/s versus 25.00 GTexel/s. The FP32 compute gap is even more pronounced: 2.038 TFLOPS for NVIDIA versus 800.0 GFLOPS for AMD. Neither GPU has ray tracing or tensor cores.

Memory architecture also differs substantially. The GTX 680M uses a 256-bit memory bus with 4 GB of GDDR5, yielding 115.2 GB/s of bandwidth. The HD 8850M uses a 128-bit bus with 2 GB of GDDR5, capping at 64.00 GB/s. Clock speeds favor NVIDIA as well, with a base of 719 MHz and boost of 758 MHz, versus AMD’s 575 MHz base and 625 MHz boost. Memory clocks are closer, with NVIDIA at 900 MHz (3.6 Gbps effective) and AMD at 1000 MHz (4 Gbps effective), but the wider bus gives NVIDIA the bandwidth win.

API support shows a notable divergence. The HD 8850M supports DirectX 12 (11_1), OpenGL 4.6, and Vulkan 1.2.170. The GTX 680M supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. While both reach DirectX 12, the AMD part’s 11_1 feature level is slightly higher than NVIDIA’s 11_0, which could matter for certain legacy titles. The Vulkan versions are nearly identical, with NVIDIA having a marginal edge. The GTX 680M is specified with a 100 W TDP and uses an MXM-B (3.0) bus interface, while the HD 8850M uses PCIe 3.0 x16 and has no TDP listed in the data.

The Verdict

The data points to a clear performance hierarchy, but the right choice depends on the workload. The NVIDIA GeForce GTX 680M is the superior GPU for raw OpenCL compute, winning the direct head-to-head by 19.3% with a score of 9,230 versus 7,447. Its 2.038 TFLOPS of FP32 performance, 115.2 GB/s memory bandwidth, and 1,344 shading units make it the heavier hitter for demanding compute tasks. Users running OpenCL-accelerated applications should favor the GTX 680M without hesitation.

However, the AMD Radeon HD 8850M posts a higher average benchmark score of 7,447 versus 7,023, and it sits one percentile higher at 40 versus 39. This is largely a function of the benchmark set, as the HD 8850M only has an OpenCL result while the GTX 680M also carries a weaker Metal score. For users who care about the aggregate ranking across all available tests, the AMD part edges ahead. Its smaller die (123 mm² versus 294 mm²) and lower transistor count also suggest it is the more efficient design per square millimeter, though no TDP data exists for the AMD chip to confirm power behavior.

The GTX 680M’s 4 GB VRAM and 256-bit bus make it the better option for high-resolution textures and bandwidth-hungry workloads. The HD 8850M’s 2 GB and 128-bit bus are more modest. In terms of API support, the HD 8850M has a slightly higher DirectX feature level (11_1 versus 11_0), which could be a deciding factor for specific older games. The GTX 680M’s Vulkan 1.2.175 support is marginally newer than the HD 8850M’s 1.2.170.

For a straightforward compute workload, the GTX 680M wins. For a broader benchmark average and a slight percentile edge, the HD 8850M takes the crown. The GTX 680M is the more capable mobile GPU in absolute terms, while the HD 8850M offers a more compact implementation with a higher average score in the recorded tests. Neither card is competitive with modern parts, but among these two end-of-life options, the NVIDIA GK104-based chip is the stronger performer where it matters most.

Specification Differences

| Specification | AMD Radeon HD 8850M | NVIDIA GeForce GTX 680M |

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

| Chip | Venus | GK104 |

| Architecture | GCN 1.0 | Kepler |

| Generation | Solar System (HD 8800M) | GeForce 600M |

| Transistors | 1,500 million | 3,540 million |

| Die Size | 123 mm² | 294 mm² |

| Transistor Density | 12.2M / mm² | 12.0M / mm² |

| Base Clock | 575 MHz | 719 MHz |

| Boost Clock | 625 MHz | 758 MHz |

| Memory Clock | 1000 MHz (4 Gbps effective) | 900 MHz (3.6 Gbps effective) |

| Memory Size | 2 GB | 4 GB |

| Memory Bus Width | 128 bit | 256 bit |

| Memory Bandwidth | 64.00 GB/s | 115.2 GB/s |

| Shading Units | 640 | 1,344 |

| TMUs | 40 | 112 |

| ROPs | 16 | 32 |

| Pixel Rate | 10.00 GPixel/s | 21.22 GPixel/s |

| Texture Rate | 25.00 GTexel/s | 84.90 GTexel/s |

| FP32 | 800.0 GFLOPS | 2.038 TFLOPS |

| TDP | Not listed | 100 W |

| Slot Width | Not listed | MXM Module |

| Power Connectors | Not listed | None |

| Bus Interface | PCIe 3.0 x16 | MXM-B (3.0) |

| Display Outputs | Not listed | Portable Device Dependent |

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

| Vulkan | 1.2.170 | 1.2.175 |

| Release Date | 2013-03-31 | 2012-06-03 |

| Predecessor | London | GeForce 500M |

| Successor | Gem System | GeForce 700M |