AMD Radeon Vega 8 Mobile vs NVIDIA GeForce GTX 680M Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The AMD Radeon Vega 8 Mobile leads with an average benchmark score of 7203, while the NVIDIA GeForce GTX 680M trails at 7023. This places the AMD part 2.6% ahead on average, despite the head-to-head OpenCL result favoring NVIDIA.

Q: How do the two GPUs rank against all other GPUs?

A: Both GPUs sit at the 39th percentile vs all GPUs. They are effectively peers in overall standing, though their nearest rival lists differ: the AMD part is closest to the NVIDIA GeForce GTX 750 (deltaPct -0.3), while the NVIDIA part is closest to the NVIDIA T600 (deltaPct -0.2).

Q: What is the transistor density difference between the two chips?

A: The AMD Radeon Vega 8 Mobile packs 23.5M transistors per mm² on a 14 nm process, while the NVIDIA GeForce GTX 680M achieves 12.0M transistors per mm² on a 28 nm process. The AMD chip is denser by nearly 2x, reflecting the newer fabrication node.

Q: Which GPU has the higher FP32 compute throughput?

A: The NVIDIA GeForce GTX 680M delivers 2.038 TFLOPS of FP32 performance, which is 81% higher than the AMD Radeon Vega 8 Mobile's 1,127.4 GFLOPS. This is a substantial compute advantage for the Kepler-based part.

Q: What is the memory bandwidth of each GPU?

A: The NVIDIA GeForce GTX 680M has 115.2 GB/s of bandwidth from 4 GB of GDDR5 on a 256-bit bus. The AMD Radeon Vega 8 Mobile uses system-shared memory, so its bandwidth is listed as "System Dependent" with no fixed figure.

Q: Which GPU supports a higher DirectX version?

A: The AMD Radeon Vega 8 Mobile supports DirectX 12 (12_1), while the NVIDIA GeForce GTX 680M supports DirectX 12 (11_0). Both support OpenGL 4.6, but the AMD part also supports Vulkan 1.3 versus Vulkan 1.2.175 on the NVIDIA part.

Architecture Differences

The AMD Radeon Vega 8 Mobile is built on GCN 5.0 architecture using the Raven-M chip, manufactured on a 14 nm process at GlobalFoundries. The NVIDIA GeForce GTX 680M uses the older Kepler architecture with the GK104 chip, fabricated on a 28 nm process at TSMC. This process gap is significant: the AMD chip packs 4,940 million transistors into a 210 mm² die, achieving a transistor density of 23.5M / mm², while the NVIDIA chip contains 3,540 million transistors on a larger 294 mm² die, yielding just 12.0M / mm².

The shader configuration diverges sharply. The AMD part has 512 shading units, 32 texture mapping units, and 8 ROPs. The NVIDIA part has 1344 shading units, 112 TMUs, and 32 ROPs — 2.6x, 3.5x, and 4x more, respectively. This explains the NVIDIA part's higher pixel rate of 21.22 GPixel/s versus 8.808 GPixel/s, and texture rate of 84.90 GTexel/s versus 35.23 GTexel/s.

Clocks also differ in character. The AMD part has a base clock of 300 MHz and a boost clock of 1101 MHz — a wide boost window typical of an integrated part. The NVIDIA part runs at a base of 719 MHz and boosts to 758 MHz, a much tighter range. Memory architecture is fundamentally different: the AMD Vega 8 Mobile uses system-shared memory with no dedicated VRAM, while the GTX 680M has 4 GB of GDDR5 on a 256-bit bus delivering 115.2 GB/s.

The form factors reflect their intended roles. The AMD part is an IGP with no power connectors and a 25 W TDP. The NVIDIA part is an MXM module with a 100 W TDP and no power connectors, using the MXM-B (3.0) bus interface. The AMD part's predecessor is GCN 3.0 IGP, while the NVIDIA part's predecessor is GeForce 500M and successor is GeForce 700M. The AMD part is newer, released on 2019-01-07, versus the NVIDIA part's 2012-06-03 release.

Where Each One Wins

The NVIDIA GeForce GTX 680M wins decisively in raw compute and rasterization throughput. Its FP32 performance of 2.038 TFLOPS is 81% higher than the AMD part's 1,127.4 GFLOPS. The NVIDIA part also dominates in pixel fill rate (21.22 GPixel/s versus 8.808 GPixel/s) and texture fill rate (84.90 GTexel/s versus 35.23 GTexel/s). This makes it the clear choice for gaming workloads that stress fill-rate limits and for compute tasks that leverage large shader counts.

The AMD Radeon Vega 8 Mobile wins on architectural modernity and efficiency. It is fabricated on a 14 nm process with 23.5M transistors per mm², versus 12.0M on the NVIDIA part. It also supports a newer DirectX feature level (12_1 versus 11_0) and a newer Vulkan version (1.3 versus 1.2.175). Its 25 W TDP is a quarter of the NVIDIA part's 100 W, making it suitable for power-constrained portable devices. The AMD part also supports FP16 at 2.255 TFLOPS (2:1), a capability the NVIDIA part lacks entirely.

In benchmark averages, the AMD part edges ahead with an average score of 7203 versus 7023 for the NVIDIA part. The AMD part also has a higher OpenCL score in its individual benchmark list (7435) than the NVIDIA part's Metal score (4815), though these are different test suites. The AMD part's nearest rivals include the NVIDIA GeForce GTX 750 (deltaPct -0.3) and GTX 560 SE (deltaPct 0.4), while the NVIDIA part's nearest rivals include the NVIDIA T600 (deltaPct -0.2) and AMD Radeon R5 M240 (deltaPct 0.7).

Specification Differences

| Specification | AMD Radeon Vega 8 Mobile | NVIDIA GeForce GTX 680M |

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

| Architecture | GCN 5.0 | Kepler |

| Process Node | 14 nm | 28 nm |

| Foundry | GlobalFoundries | TSMC |

| Transistors | 4,940 million | 3,540 million |

| Die Size | 210 mm² | 294 mm² |

| Transistor Density | 23.5M / mm² | 12.0M / mm² |

| Base Clock | 300 MHz | 719 MHz |

| Boost Clock | 1101 MHz | 758 MHz |

| Memory Size | System Shared | 4 GB |

| Memory Type | System Shared | GDDR5 |

| Memory Bus Width | System Shared | 256 bit |

| Memory Bandwidth | System Dependent | 115.2 GB/s |

| Shading Units | 512 | 1344 |

| TMUs | 32 | 112 |

| ROPs | 8 | 32 |

| Pixel Rate | 8.808 GPixel/s | 21.22 GPixel/s |

| Texture Rate | 35.23 GTexel/s | 84.90 GTexel/s |

| FP32 | 1,127.4 GFLOPS | 2.038 TFLOPS |

| FP16 | 2.255 TFLOPS (2:1) | null |

| TDP | 25 W | 100 W |

| Slot Width | IGP | MXM Module |

| Bus Interface | IGP | MXM-B (3.0) |

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

| Vulkan | 1.3 | 1.2.175 |

| Release Date | 2019-01-07 | 2012-06-03 |

Head-to-Head Benchmarks

The only direct head-to-head benchmark available is Geekbench OpenCL, and the result is unambiguous. The NVIDIA GeForce GTX 680M scores 9230, while the AMD Radeon Vega 8 Mobile scores 7435. This gives NVIDIA a 19.4% win in this test. The deltaPct of -19.4 from the AMD perspective confirms the margin.

This OpenCL result is consistent with the raw compute specifications. The NVIDIA part has 1344 shading units versus 512, and a 2.038 TFLOPS FP32 throughput versus 1,127.4 GFLOPS. The 19.4% gap in OpenCL is actually narrower than the 81% gap in theoretical FP32, suggesting the AMD part extracts more real-world performance per FLOP, likely due to its newer architecture and memory system efficiencies.

Looking at the broader benchmark landscape, the AMD part's individual OpenCL score of 7435 matches its head-to-head result exactly. The NVIDIA part also has a Metal score of 4815, which is substantially lower than its OpenCL score, indicating platform-specific performance variation. The average benchmark scores tell a different story: AMD's 7203 average edges out NVIDIA's 7023, a 2.6% margin that flips the head-to-head narrative. This suggests that in aggregate across multiple test suites, the AMD part is competitive despite losing the single OpenCL comparison.

The nearest rival data further contextualizes these results. The AMD part's closest rival is the NVIDIA GeForce GTX 750 with an average score of 7222 (deltaPct -0.3), meaning the AMD part is within 0.3% of that desktop card. The NVIDIA part's closest rival is the NVIDIA T600 at 7035 (deltaPct -0.2). Both GPUs are also near the NVIDIA GeForce GTX 970, with AMD at deltaPct 0.6 and NVIDIA at deltaPct -1.9.

The Verdict

The data presents a clear split verdict. For users who prioritize raw compute throughput, fill rates, and dedicated memory bandwidth, the NVIDIA GeForce GTX 680M is the superior choice. Its 19.4% OpenCL win, combined with 2.6x more shading units, 4x more ROPs, and 115.2 GB/s of dedicated GDDR5 bandwidth, makes it the stronger performer in graphics-intensive and compute-heavy workloads. The 2.038 TFLOPS FP32 throughput is a decisive advantage.

For users who prioritize architectural modernity, power efficiency, and API support, the AMD Radeon Vega 8 Mobile is the better pick. Its 14 nm process, 23.5M transistors per mm², and 25 W TDP represent a four-generation leap in efficiency over the 28 nm, 12.0M transistors per mm², 100 W NVIDIA part. The AMD part supports DirectX 12_1 and Vulkan 1.3, while the NVIDIA part is limited to DirectX 11_0 and Vulkan 1.2.175. The AMD part also offers FP16 support, which the NVIDIA part lacks.

The average benchmark scores complicate the raw performance narrative. The AMD part's 7203 average is 2.6% higher than the NVIDIA part's 7023, and both sit at the 39th percentile vs all GPUs. This suggests that in typical mixed workloads, the two are far closer than the OpenCL head-to-head implies. The AMD part's nearest rival is the NVIDIA GeForce GTX 750, while the NVIDIA part's nearest rival is the NVIDIA T600 — both are professional-grade comparisons that indicate comparable real-world positioning.

The release dates reinforce the generational gap: the AMD part launched on 2019-01-07, nearly seven years after the NVIDIA part's 2012-06-03 release. This explains why the AMD part achieves competitive average scores despite far lower theoretical peak performance. The NVIDIA part is a legacy product from the Kepler era, while the AMD part represents a modern integrated solution.

The verdict is straightforward: the NVIDIA GeForce GTX 680M wins on peak performance metrics and dedicated memory, making it the choice for users who need maximum fill rates and compute throughput. The AMD Radeon Vega 8 Mobile wins on efficiency, API support, and average performance, making it the choice for modern portable systems where power draw and architectural currency matter more than peak FLOPs. The 39th percentile ranking for both confirms they are mid-range performers, but they achieve that standing through different means — NVIDIA through brute force, AMD through architectural refinement.