GPU Comparison

The Verdict

The data presents a split decision. The AMD Radeon R7 Graphics and NVIDIA GeForce 940MX each claim one benchmark victory, with the overall average scores landing remarkably close: 4998 for the AMD part versus 4844 for the NVIDIA part. That is a mere 3.1% gap in the AMD’s favor, a difference that falls well within the noise of the nearest rival comparisons. Neither GPU escapes the bottom third of the performance distribution, the R7 sits at the 29th percentile of all GPUs, while the 940MX lands at the 28th. These are entry-level parts, and the data consistently shows it.

For a user prioritizing OpenCL compute workloads, the NVIDIA GeForce 940MX is the clear pick. Its 4939 OpenCL score beats the AMD’s 4015 by 18.7%, a decisive margin in that specific test. Conversely, for Vulkan-based applications or games, the AMD Radeon R7 Graphics dominates, posting 5980 against the 940MX’s 4749, a 25.9% swing in the opposite direction. The choice effectively comes down to API preference. If the workload leans on modern Vulkan titles, the integrated AMD part is unexpectedly superior. If the task is general-purpose compute or older DirectX 11-era software, the discrete NVIDIA chip holds the advantage.

The 940MX also offers the practical benefit of dedicated memory. Its 2 GB of GDDR5 on a 64-bit bus delivers 40.10 GB/s of bandwidth, whereas the R7 Graphics must share system memory, making its bandwidth “System Dependent.” For sustained workloads, the NVIDIA part’s fixed memory allocation is a structural advantage that the benchmarks likely understate. Meanwhile, the R7 Graphics, being an IGP, has zero additional power connectors and relies entirely on the motherboard for output, a simpler, lower-power integration story.

Where Each One Wins

The NVIDIA GeForce 940MX wins decisively in OpenCL compute. The 18.7% lead in geekbench_opencl is the largest margin in the entire head-to-head dataset. This suggests the 940MX is better suited for tasks like video encoding, physics simulation, or any OpenCL-accelerated productivity application. The 512 shading units and 32 texture mapping units, combined with 881.7 GFLOPS of FP32 performance, provide a raw compute ceiling that the R7’s 384 shading units and 553.0 GFLOPS cannot match.

The AMD Radeon R7 Graphics wins in Vulkan, and it wins big. The 25.9% advantage in geekbench_vulkan is even larger than the NVIDIA’s OpenCL lead. This is the more surprising result, given that the R7 is an integrated part with lower theoretical throughput. The win likely stems from architectural efficiency in Vulkan’s explicit API model. The R7 also shows better API support in one specific area: it supports DirectX 12_0, while the 940MX only reaches DirectX 12 (11_0). For modern game engines targeting DX12 or Vulkan, the AMD part has a feature-level advantage that translates directly into this benchmark victory.

Where the 940MX wins outside of raw scores is in memory consistency. The dedicated 2 GB GDDR5 frame buffer with 40.10 GB/s bandwidth is a fixed resource. The R7’s memory performance is “System Dependent,” meaning its effective bandwidth fluctuates with the host system’s RAM configuration. In a laptop with slow dual-channel memory, the R7 would likely perform worse than its Vulkan score suggests; in a system with fast memory, it might perform better. The 940MX offers predictable performance, while the R7 offers best-case potential.

Architecture Differences

The two GPUs come from different foundries, different architectures, and different design philosophies. The AMD Radeon R7 Graphics uses the Spectre Lite chip built on GCN 2.0, produced on a 28 nm process at GlobalFoundries. It integrates 2,410 million transistors across a 245 mm² die, yielding a transistor density of 9.8M per mm². The NVIDIA GeForce 940MX uses the GM107 chip on the Maxwell architecture, also 28 nm but fabricated by TSMC. It packs 1,870 million transistors into a smaller 148 mm² die, achieving a higher density of 12.6M per mm². The R7 is the physically larger chip, but the Maxwell design extracts more performance per transistor in OpenCL.

Core counts diverge significantly. The R7 fields 384 shading units, 24 texture mapping units, and 8 ROPs. The 940MX counters with 512 shading units, 32 texture mapping units, and 8 ROPs. That is a 33% advantage in shader count and a 33% advantage in texture units for NVIDIA, yet the R7 still wins Vulkan. The R7’s GCN 2.0 architecture uses a different scheduling model that appears better suited to Vulkan’s low-overhead draw calls. The 940MX’s Maxwell architecture, while efficient in raw FP32 throughput (881.7 GFLOPS vs 553.0 GFLOPS), seems to lose efficiency in this specific API context.

Memory architecture is fundamentally different. The R7 is an IGP with no dedicated memory; it uses “System Shared” memory with a “System Shared” bus width and “System Dependent” bandwidth. The 940MX is an MXM Module with 2 GB of GDDR5 on a 64-bit bus, delivering 40.10 GB/s. The clock speeds also differ: the 940MX runs at a 795 MHz base and 861 MHz boost, with memory at 1253 MHz (5 Gbps effective). The R7 has no listed base or boost clock, its memory clock is simply “System Shared.” The NVIDIA part’s fixed clocks and dedicated memory provide deterministic performance, while the R7’s performance scales with the host system’s memory subsystem.

The R7 also has a different production timeline. It launched in 2014 as a GCN 2.0 IGP for the Kaveri generation, succeeding TeraScale 3 IGP and preceding GCN 3.0 IGP. The 940MX launched in mid-2016 as part of the GeForce 900M series, succeeding GeForce 800M and preceding GeForce 10 Mobile. Both are end-of-life products, but the R7 is the older design by over two years. The R7 supports Vulkan 1.2.170, while the 940MX supports Vulkan 1.4, a newer API revision that should theoretically improve performance, yet the data shows the older AMD part winning in Vulkan anyway.

FAQ

Q: Which GPU is faster overall?

A: The average benchmark scores are nearly identical: the AMD Radeon R7 Graphics averages 4998, while the NVIDIA GeForce 940MX averages 4844. That is a 3.1% difference in AMD’s favor, but the head-to-head results are split 1-1, with each winning one benchmark.

Q: Why does the AMD R7 Graphics win in Vulkan?

A: The R7 posts a 5980 Vulkan score versus the 940MX’s 4749, a 25.9% advantage. Despite having fewer shading units (384 vs 512) and lower FP32 throughput (553.0 GFLOPS vs 881.7 GFLOPS), the GCN 2.0 architecture appears more efficient in Vulkan’s explicit API model. The R7 also supports DirectX 12_0, while the 940MX only reaches DirectX 12 (11_0).

Q: Should I choose the 940MX for OpenCL workloads?

A: Yes, the data strongly supports this. The 940MX scores 4939 in OpenCL versus 4015 for the R7, an 18.7% lead. The NVIDIA part’s 512 shading units and dedicated 2 GB GDDR5 memory with 40.10 GB/s bandwidth provide a clear advantage for compute tasks.

Q: Does the 940MX’s newer Vulkan support matter?

A: The 940MX supports Vulkan 1.4, while the R7 supports Vulkan 1.2.170. Despite this newer API revision, the 940MX still loses decisively in the Vulkan benchmark. The data suggests that raw API version support matters less than architectural efficiency in this comparison.

Q: Is the R7 Graphics limited by its shared memory?

A: The R7 uses “System Shared” memory with “System Dependent” bandwidth, meaning its performance varies with the host system’s RAM. The 940MX has a fixed 40.10 GB/s from its dedicated GDDR5. This makes the 940MX more predictable, though the R7’s Vulkan score suggests it can excel with a capable memory subsystem.

Q: Which GPU is more power-efficient?

A: The difference is minimal. The R7 has a 25 W TDP, while the 940MX has a 23 W TDP. Both are low-power parts, but the 940MX is slightly more efficient on paper, and it achieves higher OpenCL performance within that similar power envelope.

Head-to-Head Benchmarks

The two GPUs split their head-to-head matchups exactly evenly, but the margins are asymmetric and telling. In geekbench_opencl, the NVIDIA GeForce 940MX wins with a score of 4939 against the AMD Radeon R7 Graphics’ 4015. The delta is 18.7% in NVIDIA’s favor. This is a substantial gap, reflecting the 940MX’s raw compute advantage: 512 shading units versus 384, 32 TMUs versus 24, and 881.7 GFLOPS versus 553.0 GFLOPS. The 940MX also has the benefit of dedicated GDDR5 memory with 40.10 GB/s bandwidth, eliminating any memory contention that might plague the R7’s shared-memory design.

In geekbench_vulkan, the roles reverse dramatically. The AMD Radeon R7 Graphics scores 5980, while the NVIDIA GeForce 940MX falls to 4749. The delta is 25.9% in AMD’s favor, a larger margin than NVIDIA’s OpenCL win. This is the standout result of the entire comparison. The R7, with its older GCN 2.0 architecture and fewer shaders, outperforms a newer, larger-shader-count GPU by over a quarter in this API. The R7’s DirectX 12_0 support (versus 12 (11_0) for the 940MX) and its Vulkan 1.2.170 implementation evidently translate into better driver-level efficiency for modern explicit APIs.

The average benchmark scores reflect this split. The R7 averages 4998 across both tests, while the 940MX averages 4844. The R7’s Vulkan win (5980) is so large that it overcomes the 940MX’s OpenCL win (4939 vs 4015) in the overall average. The nearest rival data reinforces how close these parts are: the R7’s nearest rival is the NVIDIA Quadro 4000 at 4979 (0.4% delta), while the 940MX’s nearest rival is the GeForce GTX 560M at 4855 (-0.2% delta). Both GPUs sit in a performance cluster where single-digit percentage differences separate them from their peers.

The percentile rankings tell a similar story of mediocrity. The R7 sits at the 29th percentile of all GPUs; the 940MX sits at the 28th. Neither part is competitive with modern discrete graphics, and both are firmly in entry-level territory. The head-to-head data shows that the choice between them is not about overall superiority but about workload-specific suitability. For OpenCL-heavy tasks, the 940MX is the only rational choice. For Vulkan-based gaming or applications, the R7 is surprisingly superior. The data does not support a universal winner, only a context-dependent one.

Specification Differences

The two GPUs differ across nearly every measurable specification, with the exception of process node and ROP count. Both use a 28 nm process, though from different foundries: GlobalFoundries for the AMD chip and TSMC for the NVIDIA chip. Both have 8 ROPs. Beyond those similarities, the differences are extensive.

| Specification | AMD Radeon R7 Graphics | NVIDIA GeForce 940MX |

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

| Chip | Spectre Lite | GM107 |

| Architecture | GCN 2.0 | Maxwell |

| Generation | GCN 2.0 IGP (Kaveri) | GeForce 900M |

| Foundry | GlobalFoundries | TSMC |

| Transistors | 2,410 million | 1,870 million |

| Die Size | 245 mm² | 148 mm² |

| Transistor Density | 9.8M / mm² | 12.6M / mm² |

| Base Clock | None listed | 795 MHz |

| Boost Clock | None listed | 861 MHz |

| Memory Clock | System Shared | 1253 MHz (5 Gbps effective) |

| Memory Size | System Shared | 2 GB |

| Memory Type | System Shared | GDDR5 |

| Memory Bus Width | System Shared | 64 bit |

| Memory Bandwidth | System Dependent | 40.10 GB/s |

| Shading Units | 384 | 512 |

| TMUs | 24 | 32 |

| Pixel Rate | 5.760 GPixel/s | 6.888 GPixel/s |

| Texture Rate | 17.28 GTexel/s | 27.55 GTexel/s |

| FP32 | 553.0 GFLOPS | 881.7 GFLOPS |

| TDP | 25 W | 23 W |

| Slot Width | IGP | MXM Module |

| Power Connectors | None | None |

| Bus Interface | IGP | PCIe 3.0 x8 |

| Display Outputs | Motherboard Dependent | Portable Device Dependent |

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

| OpenGL | 4.6 | 4.6 |

| Vulkan | 1.2.170 | 1.4 |

| Release Date | 2014-02-16 | 2016-06-27 |

| Predecessor | TeraScale 3 IGP | GeForce 800M |

| Successor | GCN 3.0 IGP | GeForce 10 Mobile |

The NVIDIA part wins on raw specifications: more shading units, more TMUs, higher pixel and texture rates, higher FP32 throughput, dedicated memory, and a newer Vulkan revision. The AMD part counters with a larger die, more transistors, a higher transistor count overall, and superior DirectX 12 feature level. The 940MX also has a lower TDP (23 W vs 25 W) despite its higher performance in OpenCL. The R7’s specification sheet is notably sparse in memory-related fields, all reading “System Shared” or “System Dependent,” which underscores its fundamental design as an integrated part with no control over its own memory subsystem. The 940MX, as an MXM module, is a self-contained unit with fixed clocks and dedicated memory, making it the more predictable and, in OpenCL, the more capable part.