GPU Comparison

# NVIDIA GeForce GTX 950M vs Intel UHD Graphics 750

This head-to-head pits a 2015 mobile discrete GPU against a 2021 integrated graphics solution, and the data reveals a surprisingly split outcome. The NVIDIA GeForce GTX 950M dominates in OpenCL compute workloads, while the Intel UHD Graphics 750 turns the tables in Vulkan performance. With each card claiming one benchmark victory, the choice depends heavily on the application environment rather than raw specifications alone.

Head-to-Head Benchmarks

The Geekbench OpenCL results show a decisive win for the NVIDIA GeForce GTX 950M. It scores 9,745 points compared to the Intel UHD Graphics 750's 6,743 points, representing a 44.5% advantage for the discrete GPU. This is a substantial margin that reflects the GTX 950M's dedicated memory and higher raw compute throughput. The data indicates that for OpenCL-based workloads, often found in compute acceleration, video encoding, and some scientific applications, the older NVIDIA part holds a clear edge despite its age.

However, the Vulkan benchmark tells an entirely different story. The Intel UHD Graphics 750 scores 8,138 points, beating the GTX 950M's 6,525 points by 19.8%. This reversal is notable because Vulkan is a modern graphics API that benefits from newer architecture features. The Intel part's Generation 12.1 architecture appears better optimized for this API, despite having fewer shading units and lower memory bandwidth. The deltaPct of -19.8% in the data means the GTX 950M is behind by that amount, confirming the Vulkan advantage is not marginal but substantial.

When averaging both benchmarks, the GTX 950M achieves an average score of 8,135, placing it in the 42nd percentile of all GPUs. The Intel UHD Graphics 750 averages 7,441, sitting at the 40th percentile. The 42nd versus 40th percentile difference is small, yet the individual benchmark splits reveal that neither card is universally superior. The GTX 950M's nearest rival is the AMD Radeon R9 M360 with an average score of 8,129 (a 0.1% delta), while the Intel part's closest competitor is the AMD Radeon HD 8850M at 7,447 (a -0.1% delta). These tight margins suggest both cards sit in a crowded performance tier.

Where Each One Wins

The GTX 950M wins in scenarios that leverage OpenCL compute acceleration. Its 44.5% lead in Geekbench OpenCL indicates strong performance for general-purpose GPU computing tasks such as physics simulations, image processing filters, or data-parallel workloads that utilize this API. The card's 1,438.7 GFLOPS FP32 throughput, compared to the Intel part's 665.6 GFLOPS, more than doubles the raw floating-point capability. Additionally, the GTX 950M's dedicated 4 GB DDR3 memory with 28.80 GB/s bandwidth ensures that memory-intensive compute tasks do not compete with system memory for bandwidth.

The Intel UHD Graphics 750 wins in Vulkan-based scenarios. Its 19.8% advantage in Geekbench Vulkan suggests better driver optimization and architectural support for this modern API. This matters for gaming titles that use Vulkan, particularly recent releases or cross-platform games that favor this low-overhead API. The Intel part also supports DirectX 12 (12_1) whereas the GTX 950M only supports DirectX 12 (11_0), meaning the Intel solution can handle more advanced DirectX 12 features like variable rate shading or mesh shaders that the older NVIDIA architecture lacks.

For users running older OpenGL applications, both cards support OpenGL 4.6, so compatibility is equal there. However, the GTX 950M's higher texture rate (44.96 GTexel/s versus 20.80 GTexel/s) and pixel rate (17.98 GPixel/s versus 10.40 GPixel/s) suggest it would handle traditional graphics workloads better, even if the Vulkan benchmark shows otherwise. The data implies that the GTX 950M is the stronger choice for legacy APIs and compute, while the Intel UHD Graphics 750 is better suited for modern API-based workloads.

Architecture Differences

The architectural gap between these two GPUs is significant. The NVIDIA GeForce GTX 950M uses the GM107 chip built on a 28 nm process at TSMC, with 1,870 million transistors packed into a 148 mm² die. This translates to a transistor density of 12.6 million per square millimeter. The Maxwell architecture was designed for efficient mobile gaming performance when it launched in 2015 as part of the GeForce 900M generation.

In contrast, the Intel UHD Graphics 750 uses a Rocket Lake chip built on Intel's 14 nm+++ process. The architecture, Generation 12.1, is significantly newer, launching in 2021 as part of the HD Graphics (Rocket Lake) family. The process node is larger in physical dimensions, but the architecture benefits from years of design improvements. Interestingly, Intel does not disclose transistor count or die size for this integrated GPU, as it shares the die with the CPU.

The compute resources differ dramatically. The GTX 950M has 640 shading units, 40 texture mapping units, and 16 ROPs. The Intel UHD Graphics 750 has only 256 shading units, 16 TMUs, and 8 ROPs. Despite having 2.5 times more shading units, the GTX 950M's peak FP32 throughput is only about 2.2 times higher (1,438.7 GFLOPS versus 665.6 GFLOPS), reflecting clock speed differences. The Intel part's base clock of 300 MHz boosts to 1,300 MHz, while the GTX 950M runs at 993 MHz base and 1,124 MHz boost.

Memory architecture represents the biggest practical difference. The GTX 950M has dedicated 4 GB DDR3 on a 128-bit bus, providing 28.80 GB/s of bandwidth. The Intel UHD Graphics 750 uses system shared memory, with bandwidth described as "System Dependent." This means the Intel solution's performance varies based on the system's RAM configuration and speed, while the NVIDIA card has predictable, fixed bandwidth. The Intel part also lacks dedicated memory resources, which can impact performance in memory-intensive workloads.

Clock speeds reveal an interesting strategy: the Intel part boosts to 1,300 MHz, 15.7% higher than the GTX 950M's 1,124 MHz boost. This higher clock partially compensates for the fewer shading units. Both cards support the same APIs for OpenGL (4.6) and Vulkan (1.4), but differ in DirectX support: the GTX 950M supports DirectX 12 (11_0) while the Intel part supports DirectX 12 (12_1), a meaningful feature-level difference for modern games.

FAQ

Q: Which GPU has higher raw compute performance?

A: The NVIDIA GeForce GTX 950M delivers 1,438.7 GFLOPS FP32 throughput compared to the Intel UHD Graphics 750's 665.6 GFLOPS, more than doubling the Intel solution in raw compute capability.

Q: Why does the Intel UHD Graphics 750 win the Vulkan benchmark?

A: The Intel part scores 8,138 in Geekbench Vulkan versus 6,525 for the GTX 950M, a 19.8% advantage. This likely stems from its newer Generation 12.1 architecture and better driver optimization for modern APIs, despite having fewer shading units.

Q: How much memory bandwidth do these GPUs have?

A: The GTX 950M has dedicated 4 GB DDR3 memory with 28.80 GB/s bandwidth on a 128-bit bus. The Intel UHD Graphics 750 uses system shared memory with bandwidth that is dependent on the system configuration.

Q: What are the power consumption differences?

A: The GTX 950M has a TDP of 75 W while the Intel UHD Graphics 750 has a TDP of 15 W, making the Intel solution significantly more power-efficient for integrated use.

Q: Which GPU supports more advanced DirectX features?

A: The Intel UHD Graphics 750 supports DirectX 12 (12_1), a higher feature level than the GTX 950M's DirectX 12 (11_0) support.

Q: How do these cards compare to their nearest rivals?

A: The GTX 950M averages 8,135, nearly matching the AMD Radeon R9 M360 at 8,129 (0.1% delta). The Intel UHD Graphics 750 averages 7,441, roughly matching the AMD Radeon HD 8850M at 7,447 (-0.1% delta).

The Verdict

Based strictly on benchmark data, the NVIDIA GeForce GTX 950M is the better choice for OpenCL compute workloads and applications that rely on dedicated memory. Its 44.5% OpenCL advantage and double the FP32 throughput make it clearly superior for GPU-accelerated computing tasks. The card also offers more predictable memory performance with its dedicated 4 GB VRAM.

The Intel UHD Graphics 750 is the better choice for Vulkan-based gaming and modern API workloads. Its 19.8% Vulkan lead, higher DirectX 12 feature level, and dramatically lower 15 W power draw make it attractive for systems prioritizing energy efficiency. The system-shared memory is a drawback, but the architecture's newer design compensates in API-specific scenarios.

The percentile data shows both cards are near the middle of the GPU performance distribution: the GTX 950M at the 42nd percentile and the Intel part at the 40th. Neither is a high-performance part, but each has clear strengths. Users with compute-heavy workloads or older games should prefer the GTX 950M. Users running Vulkan titles or seeking an integrated solution with modern API support should prefer the Intel UHD Graphics 750. The data does not support a universal winner; instead, the correct choice depends entirely on the workload.

Specification Differences

| Specification | NVIDIA GeForce GTX 950M | Intel UHD Graphics 750 |

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

| Architecture | Maxwell | Generation 12.1 |

| Process Node | 28 nm | 14 nm+++ |

| Foundry | TSMC | Intel |

| Shading Units | 640 | 256 |

| TMUs | 40 | 16 |

| ROPs | 16 | 8 |

| Base Clock | 993 MHz | 300 MHz |

| Boost Clock | 1124 MHz | 1300 MHz |

| Memory Size | 4 GB | System Shared |

| Memory Type | DDR3 | System Shared |

| Memory Bus Width | 128 bit | System Shared |

| Memory Bandwidth | 28.80 GB/s | System Dependent |

| Pixel Rate | 17.98 GPixel/s | 10.40 GPixel/s |

| Texture Rate | 44.96 GTexel/s | 20.80 GTexel/s |

| FP32 Performance | 1,438.7 GFLOPS | 665.6 GFLOPS |

| FP16 Performance | N/A | 1,331.2 GFLOPS (2:1) |

| TDP | 75 W | 15 W |

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

| Bus Interface | PCIe 3.0 x8 | Ring Bus |

| Transistors | 1,870 million | Not disclosed |

| Die Size | 148 mm² | Not disclosed |