NVIDIA GeForce GTX 675MX vs NVIDIA GeForce GTX 950A Comparison

Head-to-Head Benchmarks

The only directly comparable benchmark in the database is Geekbench OpenCL, and the results show a narrow victory for the older NVIDIA GeForce GTX 675MX. The GTX 675MX scores 10,723 points, while the GTX 950A trails with 10,273 points. That is a difference of 4.2 percent in favor of the Kepler-based mobile GPU. For context, the GTX 950A sits at the 48th percentile of all GPUs in the database, while the GTX 675MX sits at the 43rd percentile, an interesting inversion where the lower-percentile card wins the head-to-head.

The delta is small enough that real-world applications would likely show run-to-run variance. The GTX 950A's nearest rivals bracket its score tightly: the AMD Radeon RX 6500M is just 0.9 percent ahead, the NVIDIA Tesla C2075 is 1.2 percent ahead, the AMD Radeon R9 M375 is 2 percent behind, and the AMD Radeon RX 550X is 2 percent ahead. The GTX 675MX's nearest rivals similarly cluster: the AMD Radeon 880M is 0.1 percent behind, the NVIDIA GeForce MX330 is 0.4 percent behind, the AMD Radeon HD 8870M is 0.4 percent behind, and the AMD Radeon R9 M375X is 1.2 percent ahead.

Notably, the GTX 675MX has a second benchmark entry, a Geekbench Metal score of 6,131, which the GTX 950A lacks. This means the older card has been tested across two different compute APIs, while the newer card has only a single OpenCL result. The average benchmark score for the GTX 675MX, 8,427 points, is dragged down by the Metal result, but the OpenCL score stands as the only apples-to-apples comparison.

Architecture Differences

The two cards come from different architectural generations, and the data shows a generational shift in design philosophy. The GTX 950A uses the GM107 chip on the Maxwell architecture, built on a 28 nm process at TSMC. It packs 1,870 million transistors into a 148 mm² die, giving a transistor density of 12.6 million per square millimeter. The GTX 675MX uses the GK104 chip on the older Kepler architecture, also on 28 nm at TSMC, but with a substantially larger 294 mm² die holding 3,540 million transistors, yielding 12.0 million per square millimeter.

The GTX 950A has a smaller transistor count and die size, but it runs at higher clock speeds. Its base clock is 993 MHz with a boost clock of 1,124 MHz. The GTX 675MX has no recorded base or boost clock in the database, so clock speed comparison is not possible from the data. However, the GTX 950A achieves its 1,438.7 GFLOPS of FP32 compute with just 640 shading units, while the GTX 675MX manages only 1,255.7 GFLOPS with 960 shading units. This indicates the Maxwell architecture extracts more floating-point throughput per shading unit, a clear architectural efficiency gain.

Memory configurations differ substantially. The GTX 950A uses 2 GB of DDR3 on a 128-bit bus, yielding 32.03 GB/s of bandwidth with memory clocked at 1,001 MHz (2 Gbps effective). The GTX 675MX uses 2 GB of GDDR5 on a 256-bit bus, producing 115.2 GB/s of bandwidth at 900 MHz (3.6 Gbps effective). The GDDR5 memory and wider bus give the GTX 675MX more than three times the memory bandwidth, a major advantage in memory-heavy workloads.

The GTX 950A has 40 texture mapping units and 16 ROPs, while the GTX 675MX has 80 TMUs and 32 ROPs, exactly double in both categories. The pixel rate tells a different story: the GTX 950A achieves 17.98 GPixel/s, while the GTX 675MX achieves 13.08 GPixel/s. The GTX 950A's pixel throughput is higher despite fewer ROPs, thanks to its clock advantage. Texture rate favors the GTX 675MX at 52.32 GTexel/s versus 44.96 GTexel/s, a 16.4 percent lead.

Where Each One Wins

The GTX 675MX wins the compute performance race in Geekbench OpenCL, but the real story is in the underlying metrics. The GTX 675MX's enormous memory bandwidth advantage, 115.2 GB/s versus 31.03 GB/s, makes it the better choice for workloads that stream large datasets, such as texture-heavy game scenes, higher resolution rendering, or data-intensive compute tasks. The 256-bit bus and GDDR5 memory are a classic combination for bandwidth-bound scenarios.

The GTX 950A, with its newer Maxwell architecture, wins in raw compute efficiency. Its FP32 throughput of 1,438.7 GFLOPS exceeds the GTX 675MX's 1,255.7 GFLOPS, despite having 320 fewer shading units. This suggests applications that are shader-bound, rather than memory-bound, may see an advantage on the GTX 950A. The higher pixel rate of 17.98 GPixel/s versus 13.08 GPixel/s also points to faster fill-rate performance, which matters for simple geometry with heavy overdraw.

In terms of API support, both cards support DirectX 12 (11_0) and OpenGL 4.6, so no difference there. The GTX 950A supports Vulkans 1.4, while the GTX 675MX supports Vulkans 1.2.175. This gives the GTX 950A a newer Vulkan version, which may matter for modern games or applications that use newer Vulkan extensions. The GTX 675MX has a Geekbench Metal score of 6,300, indicating it was tested under Apple's Metal API, while the GTX 950X has no Metal score recorded.

The GTX 950A has a TDP of 75W, while the GTX 675MX has a TDP of 100W. The Maxwell card is more power-efficient per watt, which is a significant advantage for mobile devices with limited cooling or battery capacity. Both cards use an MXM module slot width and have no power connectors, meaning they are designed for laptop or small form factor implementations.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The NVIDIA GeForce GTX 950X has an average benchmark score of 10,273 from its single Geekbench OpenCL test. The GTX 675MX has an average of 8,410 across two tests (Geekbench Metal at 6,300 and Geekbench OpenCL at 10,723), but its average is lowered by the Metal result.

Q: What is the GTX 950X's transistor density compared to the GTX 675MX?

A: The GTX 950X has a transistor density of 12.6 million per square millimeter, while the GTX 675MX has a density of 12.0 million per square millimeter. This means the GTX 950X packs more transistors into each square millimeter of silicon.

Q: How does memory bandwidth differ between the two cards?

A: The GTX 675MX has 115.2 GB/s of memory bandwidth, while the GTX 950X has 31.03 GB/s. The GTX 675MX's GDDR5 memory on a 256-bit bus provides roughly 3.6 times the bandwidth of the GTX 950X's DDR3 on a 128-bit bus.

Q: Which GPU supports a newer Vulkan version?

A: The NVIDIA GeForce GTX 950X supports Vulkan 1.4, while the GTX 675MX supports Vulkans 1.2.175. Both support DirectX 12 (11_0) and OpenGL 4.6.

Q: What is the pixel rate of each GPU?

A: The GTX 950X has a pixel rate of 17.98 GPixel/s, which is higher than the GTX 675MX's 13.08 GPixel/s. This indicates the GTX 950X can output more filled pixels per second.

Q: How many shading units does each card have?

A: The GTX 950X has 640 shading units, while the GTX 675MX has 960 shading units. The GTX 675MX has 50 percent more shading units, but the GTX 950X still achieves higher FP32 performance (1,438.7 GFLOPS versus 1,255.7 GFLOPS).

The Verdict

The data presents a straightforward trade-off. The GTX 675MX wins the only direct head-to-head benchmark by 4.2 percent, and it holds a massive memory bandwidth advantage at 115.2 GB/s versus 31.03 GB/s. Its also has twice the TMUs and ROPs. This makes it the better choice for any workload that is memory-bound or heavily textured, such as high-resolution gaming with large textures or compute tasks that stream data.

The GTX 950X, despite losing the single benchmark, wins on architectural efficiency. It has a higher transistor density, a higher pixel rate (17.98 GPixel/s versus 13.08 GPixel/s), and higher FP32 throughput (1,438.7 GFLOPS versus 1,255.7 GFLOPS) with fewer shading units. It also has a lower 75W TDP compared to 100W, making it more power-efficient. For applications that are compute-bound or fill-rate-bound, and for systems where power consumption is a concern, the GTX 950X is the more appropriate pick.

The percentile rankings reflect this split: the GTX 950X sits at the 48th percentile (higher overall), while the GTX 675MX sits at the 43rd percentile. The newer Maxwell architecture and Vulkan 1.4 support give the GTX 950X a forward-looking edge, though the GTX 675MX has a Metal benchmark score, which the GTX 950X lacks. Users who run Metal workloads on Apple platforms would need the GTX 675MX, while those using Vulkan or requiring lower power draw would prefer the GTX 950X.

Specification Differences

| Specification | NVIDIA GeForce GTX 950X | NVIDIA GeForce GTX 675MX |

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

| Architecture | Maxwell | Kepler |

| Chip | GM107 | GK104 |

| Generation | GeForce 900A | GeForce 600M |

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

| Die Size | 148 mm² | 294 mm² |

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

| Base Clock | 993 MHz | Not recorded |

| Boost Clock | 1,124 MHz | Not recorded |

| Memory Clock | 1,001 MHz (2 Gbps effective) | 900 MHz (3.6 Gbps effective) |

| Memory Type | DDR3 | GDDR5 |

| Memory Bus Width | 128 bit | 256 bit |

| Memory Bandwidth | 32.03 GB/s | 115.2 GB/s |

| Shading Units | 640 | 960 |

| TMUs | 40 | 80 |

| ROPs | 16 | 32 |

| Pixel Rate | 17.98 GPixel/s | 13.08 GPixel/s |

| Texture Rate | 44.96 GTexel/s | 52.32 GTexel/s |

| FP32 | 1,438.7 GFLOPS | 1,255.7 GFLOPS |

| TDP | 75 W | 100 W |

| Vulkan | 1.4 | 1.2.175 |

| Release Date | 2015-03-12 | 2012-09-30 |

| Predecessor | GeForce 800A | GeForce 500M |

| Successor | Not recorded | GeForce 700M |

| Geekbench OpenCL | 10,273 | 10,723 |

| Geekbench Metal | Not recorded | 6,131 |

| Percentile | 48th | 43rd |