NVIDIA GeForce 945M vs NVIDIA GeForce GTX 880M Comparison

# The Verdict

The benchmark data tells a surprisingly clear story despite both GPUs sitting at the 42nd percentile among all GPUs. The NVIDIA GeForce 945M wins the only head-to-head benchmark available, delivering a 44.1% higher Geekbench OpenCL score of 8099 versus the GTX 880M's 5622. That is not a marginal gap—it is a decisive generational leap in compute performance.

However, the GTX 880M is not without its own arguments. Its average benchmark score of 8040 across two tests (including a Geekbench Metal score of 10458) shows it can excel in specific workloads. The 945M has only one benchmark entry, the OpenCL score of 8099, which matches its average. The 880M's Metal score of 10458 is significantly higher than either GPU's OpenCL results, suggesting Apple's Metal API favors the older Kepler architecture in ways OpenCL does not.

For buyers prioritizing raw compute throughput in OpenCL applications, the 945M is the clear choice. For those targeting Metal-based workloads, the 880M's 10458 Metal score indicates superior performance in that specific API. The data also shows the 945M sits within 0.4% of the GTX 950M (score 8135) and 0.4% above the Radeon R9 M360 (score 8129), placing it in a competitive mid-range mobile segment. The 880M's nearest rival is the Quadro P5000 at 8039, a 0% delta, meaning the 880M essentially matches a professional-grade workstation GPU in average score.

# Architecture Differences

The architectural divide here is fundamental: the 945M uses the GM107 chip built on Maxwell, while the 880M uses the GK104 chip built on Kepler. Both are fabricated on TSMC's 28 nm process, but the similarities end there. The GM107 packs 1,870 million transistors into a 148 mm² die, yielding a transistor density of 12.6M per mm². The GK104 is a much larger chip at 294 mm² with 3,540 million transistors, though its density is slightly lower at 12.0M per mm².

Maxwell's architectural efficiency is evident in the 945M's compute capabilities. Despite having only 640 shading units, 40 texture mapping units, and 16 ROPs, it achieves 1,305.6 GFLOPS of FP32 performance. The 880M counters with 1,536 shading units, 128 TMUs, and 32 ROPs, delivering 3.050 TFLOPS—more than double the raw FP32 throughput. Yet in the OpenCL benchmark, the smaller Maxwell chip wins decisively. This suggests Maxwell's per-shader efficiency and scheduling improvements outweigh Kepler's raw hardware count.

Clock speeds tell a similar story of architectural refinement versus brute force. The 945M runs at a 928 MHz base and 1020 MHz boost, while the 880M starts higher at 954 MHz base but only boosts to 993 MHz. The 880M's memory operates at 1250 MHz with 5 Gbps effective, compared to the 945M's 900 MHz with 1800 Mbps effective. The 880M's memory subsystem is far wider (256-bit versus 128-bit) and delivers 160.0 GB/s bandwidth versus 28.80 GB/s—a 5.5x advantage that should matter in bandwidth-bound scenarios.

API support differs slightly: both support DirectX 12 (11_0) and OpenGL 4.6, but the 945M supports Vulkan 1.4 while the 880M is limited to Vulkan 1.2.175. The 880M's Metal benchmark score suggests Apple's API is well-optimized for Kepler, but the 945M's newer Vulkan support offers forward compatibility for modern Linux and Android-style workloads.

# FAQ

Q: Which GPU has the higher average benchmark score?

A: The 945M leads with an average score of 8099 from its single OpenCL test. The 880M averages 8040 across its Metal and OpenCL tests, placing it 59 points behind. Both sit at the 42nd percentile among all GPUs.

Q: How significant is the 880M's Metal benchmark advantage?

A: The 880M scores 10458 in Geekbench Metal, which is substantially higher than either GPU's OpenCL results. This 10458 score is 29.2% above the 945M's best OpenCL score of 8099, indicating Metal-specific workloads could favor the older Kepler GPU.

Q: What is the memory configuration difference?

A: The 945M comes with 2 GB of DDR3 on a 128-bit bus, providing 28.80 GB/s bandwidth. The 880M offers 8 GB of GDDR5 on a 256-bit bus with 160.0 GB/s bandwidth—a 5.5x bandwidth advantage and 4x capacity advantage.

Q: Does the 880M's larger chip translate to better performance?

A: No. Despite having 3,540 million transistors (89% more than the 945M's 1,870 million) and 2.5x more shading units, the 880M loses the OpenCL head-to-head by 44.1%. The Maxwell architecture's efficiency overcomes Kepler's hardware advantage.

Q: Are these GPUs comparable in their competitive positioning?

A: Yes. The 945M's nearest rival is the GTX 950M at 8135 (0.4% ahead), while the 880M's nearest rival is the Quadro P5000 at 8039 (0% difference). Both occupy similar mid-range performance tiers despite their architectural differences.

Q: What do the transistor densities suggest about design philosophy?

A: The 945M's 12.6M transistors per mm² versus the 880M's 12.0M per mm² indicates Maxwell achieves higher density in a much smaller package. The 945M's 148 mm² die is roughly half the 880M's 294 mm², yet delivers superior OpenCL performance.

# Specification Differences

| Specification | 945M | GTX 880M |

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

| Chip | GM107 | GK104 |

| Architecture | Maxwell | Kepler |

| Generation | GeForce 900M | GeForce 800M |

| Transistors | 1,870 million | 3,540 million |

| Die Size | 148 mm² | 294 mm² |

| Transistor Density | 12.6M / mm² | 12.0M / mm² |

| Base Clock | 928 MHz | 954 MHz |

| Boost Clock | 1020 MHz | 993 MHz |

| Memory Clock | 900 MHz (1800 Mbps effective) | 1250 MHz (5 Gbps effective) |

| Memory Size | 2 GB | 8 GB |

| Memory Type | DDR3 | GDDR5 |

| Memory Bus Width | 128 bit | 256 bit |

| Memory Bandwidth | 28.80 GB/s | 160.0 GB/s |

| Shading Units | 640 | 1536 |

| TMUs | 40 | 128 |

| ROPs | 16 | 32 |

| Pixel Rate | 16.32 GPixel/s | 31.78 GPixel/s |

| Texture Rate | 40.80 GTexel/s | 127.1 GTexel/s |

| FP32 Performance | 1,305.6 GFLOPS | 3.050 TFLOPS |

| TDP | 75 W | 122 W |

| Vulkan Support | 1.4 | 1.2.175 |

| Release Date | 2015-10-26 | 2014-03-11 |

| Predecessor | GeForce 800M | GeForce 700M |

| Successor | GeForce 10 Mobile | GeForce 900M |

# Head-to-Head Benchmarks

The sole direct comparison is Geekbench OpenCL, and it is a landslide. The 945M scores 8099 against the 880M's 5622, a delta of 44.1%. This is not a subtle victory—the Maxwell GPU outperforms its Kepler predecessor by nearly half again its own score. The result is particularly striking given the 880M's massive hardware specifications: 2.5x the shading units, 3.2x the TMUs, 2x the ROPs, and 5.5x the memory bandwidth.

The 880M's pixel rate of 31.78 GPixel/s and texture rate of 127.1 GTexel/s dwarf the 945M's 16.32 GPixel/s and 40.80 GTexel/s. In fill-rate-bound scenarios, the 880M should theoretically dominate. Yet OpenCL compute does not reward these characteristics. The 945M's 1,305.6 GFLOPS FP32, while less than half the 880M's 3.050 TFLOPS, still wins the compute benchmark decisively. This implies Maxwell's instruction scheduling and memory access patterns are far more efficient for general-purpose compute workloads.

The 880M's Metal benchmark score of 10458 provides the only counterpoint. If we compare that to the 945M's OpenCL score of 8099, the 880M leads by 29.1%. However, these are different APIs testing different workloads. The 880M's OpenCL result of 5622 is 30.6% below its Metal result, suggesting the Kepler architecture responds very differently to API-level optimization.

# Where Each One Wins

The 945M wins in OpenCL compute workloads. Its 8099 score versus the 880M's 5622 shows a 44.1% advantage that cannot be explained by any single specification. The Maxwell architecture's efficiency, combined with its newer Vulkan 1.4 support, makes it the better choice for general-purpose GPU computing, OpenCL acceleration, and cross-platform applications that leverage Vulkan.

The 945M wins in power efficiency. At 75 W TDP versus the 880M's 122 W, the 945M delivers superior OpenCL performance while consuming 38.5% less power. For laptop implementations, this translates to better battery life and less thermal stress. The 945M's smaller 148 mm² die also suggests lower manufacturing costs and better yields, though pricing data is not available.

The 880M wins in memory-heavy scenarios. Its 160.0 GB/s bandwidth and 8 GB capacity are unmatched by the 945M's 28.80 GB/s and 2 GB. Applications that load large textures, datasets, or require high-resolution framebuffers will benefit from the 880M's memory subsystem. The 256-bit bus and GDDR5 memory provide 5.5x the bandwidth, which matters for certain rendering tasks.

The 880M wins in Metal-based workloads. Its 10458 Metal score is the highest single benchmark result between both GPUs. On platforms using Apple's Metal API, the 880M demonstrates superior optimization despite its older Kepler architecture. This suggests macOS or Metal-compatible applications would favor the 880M.

The 880M wins in raw fill-rate throughput. With 31.78 GPixel/s and 127.1 GTexel/s, it offers roughly 2x and 3.1x the rates of the 945M respectively. Traditional rasterization workloads that stress pixel and texture processing would see measurable benefits from the 880M's larger hardware configuration.

The 945M wins in architectural longevity. Its newer release date, Maxwell generation, and Vulkan 1.4 support indicate better forward compatibility with modern APIs. The 880M's Vulkan 1.2.175 support and older Kepler design suggest it is closer to the end of its software optimization lifecycle. The 945M's successor is GeForce 10 Mobile, while the 880M's successor is GeForce 900M, placing the 945M one full generation ahead in NVIDIA's product timeline.