GPU Comparison

The NVIDIA GeForce GTX 965M and the Intel Iris Xe MAX Graphics occupy nearly the same performance tier, with average benchmark scores of 14404 and 14315 respectively, placing both at the 56th percentile of all GPUs. Despite being separated by nearly six years of silicon evolution, the data shows they are direct competitors in overall compute capability, though their architectural approaches and memory configurations reveal starkly different design philosophies. The GTX 965M edges ahead by a razor-thin 0.6% in average score, making the choice between them a matter of specific workload characteristics rather than raw performance dominance.

Where Each One Wins

The benchmark results indicate a single head-to-head comparison in Geekbench OpenCL, where the NVIDIA GeForce GTX 965M claims victory with a score of 14509 against the Intel Iris Xe MAX's 14315, a margin of 1.4%. This makes the GTX 965M the winner in the only direct test available, suggesting that for OpenCL compute workloads, the older Maxwell architecture retains a slight edge over Intel's Generation 12.1 design.

However, the Intel Iris Xe MAX Graphics counters with superior theoretical throughput metrics that do not translate into a benchmark win in this dataset. The Intel part achieves 2.534 TFLOPS of FP32 performance compared to the GTX 965M's 1.946 TFLOPS, a 30% advantage in raw floating-point capability. Similarly, the Iris Xe MAX reaches 39.60 GPixel/s pixel rate versus 30.40 GPixel/s for the GTX 965M, and 79.20 GTexel/s texture rate versus 60.80 GTexel/s. These figures suggest that in workloads that scale with theoretical peak throughput, such as certain rendering tasks or compute shaders, the Intel part should have an advantage, yet the actual OpenCL score contradicts this expectation.

The GTX 965M wins on memory bandwidth with 80.19 GB/s compared to the Iris Xe MAX's 68.26 GB/s, a 17% advantage that likely contributes to its OpenCL victory. The NVIDIA part also carries 2 GB of GDDR5 memory, while Intel provides 4 GB of LPDDR4X, offering double the capacity but with slower bandwidth. For scenarios requiring large working sets, the Iris Xe MAX's 4 GB capacity provides headroom, while the GTX 965M's faster memory benefits latency-sensitive and bandwidth-bound operations.

Architecture Differences

The GTX 965M is built on NVIDIA's Maxwell 2.0 architecture using the GM204 chip, fabricated on a 28 nm process at TSMC. This chip contains 5,200 million transistors on a 398 mm² die, yielding a transistor density of 13.1 million per square millimeter. The Intel Iris Xe MAX uses the DG1 chip based on Generation 12.1 (Xe Graphics) architecture, manufactured on Intel's 10 nm process, with a much smaller 95 mm² die size and no disclosed transistor count. The process node difference is substantial, 28 nm versus 10 nm, representing roughly two generations of lithographic advancement.

Core configurations differ meaningfully: the GTX 965M packs 1024 shading units, 64 texture mapping units, and 32 raster operation units, while the Iris Xe MAX has 768 shading units, 48 TMUs, and 24 ROPs. Despite having fewer cores, the Intel part achieves higher peak rates due to its significantly higher boost clock of 1650 MHz versus the GTX 965M's 950 MHz boost. The base clocks tell a similar story: 300 MHz for Intel versus 924 MHz for NVIDIA, showing that Intel relies on aggressive boost behavior while NVIDIA maintains a more consistent clock profile.

Memory architecture diverges completely. The GTX 965M uses 2 GB of GDDR5 on a 128-bit bus at 1253 MHz (5 Gbps effective) delivering 80.19 GB/s. The Iris Xe MAX uses 4 GB of LPDDR4X on the same 128-bit bus at 2133 MHz (4.3 Gbps effective) achieving 68.26 GB/s. The GTX 965M also supports FP16 with a dedicated 5.069 TFLOPS (2:1) rate, while the GTX 965M lists no FP16 capability. Both support DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4 identically.

Head-to-Head Benchmarks

The sole head-to-head benchmark is Geekbench OpenCL, where the NVIDIA GeForce GTX 965M scores 14509 against the Intel Iris Xe MAX's 14315, producing a 1.4% delta in NVIDIA's favor. This result is notable because the Intel part's theoretical FP32 throughput is 30% higher, yet it loses the actual compute test. The explanation likely lies in the memory subsystem: the GTX 965M's GDDR5 provides 17% more bandwidth, and OpenCL workloads often stress memory access patterns heavily.

Examining the nearest rival data adds context. The GTX 965M's closest competitor is the AMD Radeon RX Vega 11 with an average score of 14385, just 0.1% behind, followed by the NVIDIA GeForce GTX TITAN at 14373 (0.2% gap) and the AMD Radeon Vega 11 at 14352 (0.4% gap). The Intel Iris Xe MAX sits at 14315, only 0.6% behind the GTX 965M. For the Intel part, its nearest rival is the AMD Radeon Vega 11 at 14352, which is 0.3% ahead, while the NVIDIA GeForce GTX 1070 Ti trails by 0.3% at 14277.

The deltaPct values reveal that both GPUs sit in an extremely tight performance cluster. The GTX 965M's 1.4% win in the head-to-head test is larger than the 0.6% average score gap, suggesting that the OpenCL workload specifically favors the NVIDIA architecture. Interestingly, the Intel part's average score equals its OpenCL score (14315), indicating no additional benchmark data points exist for it, while the GTX 965M also has a Vulkan score of 14299, which lowers its average slightly from the OpenCL result.

The Verdict

From the data, the NVIDIA GeForce GTX 965M is the safer choice for OpenCL compute workloads, as it won the only direct comparison by 1.4%. Its higher memory bandwidth (80.19 GB/s versus 68.26 GB/s) and larger number of shading units (1024 versus 768) appear to matter more in practice than the Intel part's higher clock speeds and theoretical FP32 throughput. For users running OpenCL-based applications, the GTX 965M's 14509 score provides a measurable, if small, advantage.

The Intel Iris Xe MAX Graphics offers compelling advantages on paper that the benchmark data does not fully validate. Its 4 GB memory capacity is double the GTX 965M's 2 GB, making it more suitable for workloads with large memory footprints. The 1650 MHz boost clock and 2.534 TFLOPS FP32 rate suggest that applications optimized for Intel's architecture could perform differently than what the OpenCL score indicates. Additionally, the Iris Xe MAX's 25 W TDP and integrated form factor (IGP) stand in contrast to the GTX 965M's MXM module design, implying different deployment scenarios.

The percentile rankings are identical at 56, meaning both GPUs outperform roughly 56% of all GPUs in the database. The average score gap of 89 points (14404 versus 14315) is negligible in real-world terms, representing less than 1% difference. Users should choose based on platform compatibility and specific workload requirements rather than expecting a meaningful performance difference between these two parts.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 965M has an average benchmark score of 14404, which is 0.6% higher than the Intel Iris Xe MAX's 14315.

Q: What is the difference in memory bandwidth between the two?

A: The GTX 965M provides 80.19 GB/s of bandwidth from GDDR5 memory, while the Iris Xe MAX provides 68.26 GB/s from LPDDR4X, giving the NVIDIA part a 17% advantage.

Q: How do the FP32 floating-point performances compare?

A: The Intel Iris Xe MAX delivers 2.534 TFLOPS of FP32 performance, which is 30% higher than the GTX 965M's 1.946 TFLOPS.

Q: Which GPU has more shading units?

A: The NVIDIA GeForce GTX 965M has 1024 shading units, compared to 768 on the Intel Iris Xe MAX.

Q: What is the memory capacity difference?

A: The Intel Iris Xe MAX has 4 GB of LPDDR4X memory, while the GTX 965M has 2 GB of GDDR5, giving Intel twice the capacity.

Q: Are there any benchmark tests the Intel GPU wins?

A: In the head-to-head Geekbench OpenCL test, the Intel Iris Xe MAX scored 14315 versus the GTX 965M's 14509, meaning it did not win that test; no other direct comparisons are available in the data.

Specification Differences

| Specification | NVIDIA GeForce GTX 965M | Intel Iris Xe MAX Graphics |

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

| Architecture | Maxwell 2.0 | Generation 12.1 |

| Process Node | 28 nm | 10 nm |

| Foundry | TSMC | Intel |

| Die Size | 398 mm² | 95 mm² |

| Transistors | 5,200 million | Not disclosed |

| Base Clock | 924 MHz | 300 MHz |

| Boost Clock | 950 MHz | 1650 MHz |

| Memory Size | 2 GB | 4 GB |

| Memory Type | GDDR5 | LPDDR4X |

| Memory Bus | 128 bit | 128 bit |

| Memory Bandwidth | 80.19 GB/s | 68.26 GB/s |

| Shading Units | 1024 | 768 |

| TMUs | 64 | 48 |

| ROPs | 32 | 24 |

| Pixel Rate | 30.40 GPixel/s | 39.60 GPixel/s |

| Texture Rate | 60.80 GTexel/s | 79.20 GTexel/s |

| FP32 Performance | 1.946 TFLOPS | 2.534 TFLOPS |

| FP16 Performance | Not listed | 5.069 TFLOPS (2:1) |

| TDP | Not listed | 25 W |

| Slot Width | MXM Module | IGP |

| Bus Interface | MXM-B (3.0) | PCIe 4.0 x8 |

| Power Connectors | None | Not listed |

| Suggested PSU | Not listed | 200 W |

| Display Outputs | Portable Device Dependent | No outputs |

| Release Date | 2015-01-08 | 2020-10-30 |

| Production Status | End-of-life | End-of-life |