NVIDIA GeForce GTX 675M vs NVIDIA GeForce GTX 950M Comparison

Head-to-Head Benchmarks

The database contains a single head-to-head benchmark between these two mobile graphics processors, and the result is decisive. In the Geekbench OpenCL test, the NVIDIA GeForce GTX 950M records a score of 9745, while the NVIDIA GeForce GTX 675M trails at 6946. That translates to a 40.3% delta in favor of the newer Maxwell-based part. The GTX 950M wins the only direct comparison, and the win column reflects that: one win for the GTX 950M, zero for the GTX 675M.

Context from the nearest rivals list makes the margin more meaningful. The GTX 950M sits just 0.4% behind the NVIDIA GeForce GTX 980 in average benchmark score, a gap that is effectively negligible in real-world performance terms. It is also 0.1% ahead of the AMD Radeon R9 M360 and 0.4% ahead of the NVIDIA GeForce 945M. The GTX 675M, by contrast, trails its nearest listed rival, the NVIDIA T600, by 1.3%, and sits 1.1% behind the GeForce GTX 680M. The older Fermi chip is only 0.4% ahead of the AMD Radeon R5 M240, which puts it in a much tighter competitive bracket.

The 40.3% OpenCL advantage for the GTX 950M is not a marginal improvement; it represents a substantial generational leap. The data shows a clean sweep in the only measurable comparison, with no test where the GTX 675M pulls ahead. The GTX 950M also holds a higher percentile rank among all GPUs in the database: 42 versus 39 for the GTX 675M. That three-point percentile gap aligns with the benchmark delta, suggesting the older card has slipped further down the overall performance distribution.

Architecture Differences

The two GPUs come from different architectural generations, and the specification differences explain much of the benchmark gap. The GTX 950M uses the GM107 chip built on Maxwell architecture, fabricated on a 28 nm process at TSMC. The GTX 675M uses the GF114 chip based on Fermi 2.0 architecture, fabricated on a 40 nm process, also at TSMC. The process node difference is significant: 28 nm versus 40 nm allows the newer chip to pack more transistors into a smaller area. The GTX 950M contains 1,870 million transistors on a 148 mm² die, yielding a transistor density of 12.6M per mm². The GTX 675M carries 1,950 million transistors, slightly more in absolute terms, but spread across a much larger 332 mm² die, resulting in a density of just 5.9M per mm². That is less than half the density of the newer part, a clear indicator of how much more efficient the Maxwell design is.

Clock speeds differ as well. The GTX 950M has a base clock of 993 MHz and a boost clock of 1124 MHz. The GTX 675M has no recorded base or boost clock in the database, so the two cannot be directly compared on that axis. Memory clocks, however, are recorded: the GTX 950M runs its memory at 900 MHz, or 1800 Mbps effective, while the GTX 675M runs at 750 MHz, or 3 Gbps effective. The GTX 675M uses GDDR5 memory with a 256-bit bus, producing 96.00 GB/s of bandwidth. The GTX 950M uses DDR3 memory with a 128-bit bus, producing just 28.80 GB/s. This is a notable inversion: the older card has over three times the memory bandwidth on paper, yet still loses the compute benchmark by a wide margin. The GTX 950M compensates with far higher raw compute throughput.

Shader configuration differs substantially. The GTX 950M has 640 shading units, 40 texture mapping units, and 16 raster output pipelines. The GTX 675M has 384 shading units, 64 TMUs, and 32 ROPs. The newer card has 256 more shaders, but fewer TMUs and ROPs. Despite the lower texture unit count, the GTX 950M achieves a higher texture rate: 44.96 GTexel/s versus 39.68 GTexel/s for the GTX 675M, driven by its higher clock speed. The pixel rate also favors the GTX 950M at 17.98 GPixel/s versus 9.920 GPixel/s. Floating-point performance tells the same story: the GTX 950M delivers 1,438.7 GFLOPS of FP32 compute, while the GTX 675M delivers 952.3 GFLOPS. That is roughly a 51% advantage in raw shader math, which explains the OpenCL result.

Other architectural differences include power and form factor. The GTX 950M has a TDP of 75 W and uses an IGP slot width with no power connectors. The GTX 675M has a TDP of 100 W and uses an MXM Module slot width, also with no power connectors. The bus interface differs: PCIe 3.0 x8 for the GTX 950M versus MXM-B (3.0) for the GTX 675M. API support is largely similar, with both supporting DirectX 12 (11_0) and OpenGL 4.6, but the GTX 950M adds Vulkan 1.4 support while the GTX 675M has no recorded Vulkan support. Neither chip has ray tracing cores or tensor cores.

The Verdict

The benchmark data points to a clear winner. The NVIDIA GeForce GTX 950M outperforms the NVIDIA GeForce GTX 675M by 40.3% in the only direct comparison available, the Geekbench OpenCL test. The GTX 950M also ranks higher in the overall GPU percentile distribution, sitting at the 42nd percentile versus the 39th for the GTX 675M. For compute-oriented workloads that stress shader throughput, the GTX 950M is the superior part.

However, the GTX 675M is not without its own advantages in the specification sheet. It carries 2 GB of GDDR5 memory on a 256-bit bus, yielding 96.00 GB/s of bandwidth, versus the GTX 950M's 4 GB of DDR3 on a 128-bit bus at 28.80 GB/s. For memory-bandwidth-sensitive workloads, the older card may have an edge, though the database does not include a benchmark that isolates this factor. The GTX 675M also has more TMUs and ROPs, which could theoretically benefit certain fill-rate-bound scenarios, but the recorded texture and pixel rates actually favor the GTX 950M due to its higher clocks.

The GTX 675M is an end-of-life product from the GeForce 600M generation, released in March 2012, with a successor in the GeForce 700M line. The GTX 950M is also end-of-life, from the GeForce 900M generation, released in March 2015, with a successor in the GeForce 10 Mobile line. The three-year gap in release dates is reflected in the architectural improvements. For any user choosing strictly on measured performance, the GTX 950M is the pick. The GTX 675M only makes sense if memory bandwidth or the MXM form factor is a hard requirement, and no benchmark in the database validates that tradeoff.

FAQ

Q: Which GPU wins the only head-to-head benchmark in the database?

A: The NVIDIA GeForce GTX 950M wins the Geekbench OpenCL test with a score of 9745 against 6946 for the GTX 675M, a 40.3% advantage.

Q: How do the two GPUs compare on memory bandwidth?

A: The GTX 675M has 96.00 GB/s of bandwidth from 2 GB of GDDR5 on a 256-bit bus, while the GTX 950M has 28.80 GB/s from 4 GB of DDR3 on a 128-bit bus.

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

A: The GTX 950M has a density of 12.6M transistors per mm² on a 148 mm² die, while the GTX 675M has 5.9M per mm² on a 332 mm² die.

Q: Does the GTX 950M support Vulkan?

A: Yes, the GTX 950M supports Vulkan 1.4. The GTX 675M has no recorded Vulkan support in the database.

Q: Which GPU has the higher FP32 compute throughput?

A: The GTX 950M delivers 1,438.7 GFLOPS, compared to 952.3 GFLOPS for the GTX 675M.

Q: What are the TDP values for each GPU?

A: The GTX 950M has a TDP of 75 W, while the GTX 675M has a TDP of 100 W.

Where Each One Wins

The GTX 950M wins in raw compute performance. Its OpenCL score is 40.3% higher, its FP32 throughput is 1,438.7 GFLOPS versus 952.3 GFLOPS, and its pixel rate is 17.98 GPixel/s versus 9.920 GPixel/s. The texture rate also favors the newer card at 44.96 GTexel/s versus 39.68 GTexel/s. For general-purpose GPU compute, shader-heavy rendering, or any workload that scales with shader count and clock speed, the GTX 950M is the clear choice. It also wins on power efficiency, drawing 75 W versus 100 W while delivering more performance.

The GTX 675M wins in memory bandwidth. Its 96.00 GB/s is more than triple the GTX 950M's 28.80 GB/s. For workloads that are bandwidth-bound, such as certain texture streaming scenarios or large data transfers, the older card has a theoretical advantage. It also has more TMUs (64 versus 40) and more ROPs (32 versus 16), which could matter for specific fill-rate-limited tasks, though the measured texture and pixel rates do not reflect that advantage. The GTX 675M also uses GDDR5 memory, which has lower latency characteristics than DDR3, though the database does not provide latency measurements.

The GTX 950M wins on API support with Vulkan 1.4, while the GTX 675M has none recorded. Both support DirectX 12 (11_0) and OpenGL 4.6. In terms of form factor, the GTX 950M is an IGP (integrated graphics processor) with no power connectors, while the GTX 675M is an MXM Module, also with no power connectors. The MXM form factor may be more serviceable in certain laptop designs, but the GTX 950M's lower TDP makes it more suitable for thinner systems.

Specification Differences

| Specification | NVIDIA GeForce GTX 950M | NVIDIA GeForce GTX 675M |

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

| Chip | GM107 | GF114 |

| Architecture | Maxwell | Fermi 2.0 |

| Generation | GeForce 900M | GeForce 600M |

| Process Node | 28 nm | 40 nm |

| Transistors | 1,870 million | 1,950 million |

| Die Size | 148 mm² | 332 mm² |

| Transistor Density | 12.6M / mm² | 5.9M / mm² |

| Memory Clock | 900 MHz (1800 Mbps effective) | 750 MHz (3 Gbps effective) |

| Memory Size | 4 GB | 2 GB |

| Memory Type | DDR3 | GDDR5 |

| Memory Bus Width | 128 bit | 256 bit |

| Memory Bandwidth | 28.80 GB/s | 96.00 GB/s |

| Shading Units | 640 | 384 |

| TMUs | 40 | 64 |

| ROPs | 16 | 32 |

| Pixel Rate | 17.98 GPixel/s | 9.920 GPixel/s |

| Texture Rate | 44.96 GTexel/s | 39.68 GTexel/s |

| FP32 Performance | 1,438.7 GFLOPS | 952.3 GFLOPS |

| TDP | 75 W | 100 W |

| Slot Width | IGP | MXM Module |

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

| Vulkan Support | 1.4 | None |

| Release Date | 2015-03-12 | 2012-03-21 |

| Predecessor | GeForce 800M | GeForce 500M |

| Successor | GeForce 10 Mobile | GeForce 700M |