NVIDIA GeForce GTX 675M vs NVIDIA GeForce GTX 680M Comparison
NVIDIA GeForce GTX 675M
GeForce GTX 680M
PERFORMANCE BENCHMARKS
Analysis: NVIDIA GeForce GTX 675M vs NVIDIA GeForce GTX 680M
The NVIDIA GeForce GTX 680M and the NVIDIA GeForce GTX 675M are both end-of-life mobile graphics solutions from the GeForce 600M generation, yet they represent two distinct design philosophies from NVIDIA. The data shows a clear performance hierarchy, but the underlying architectural shifts between the two parts reveal a more nuanced story about efficiency and generational progress. Both cards target the same MXM form factor and share a 100 W TDP, but the 680M achieves its dominance through a radically different silicon design.
FAQ
Q: Which GPU is faster based on the benchmark data?
A: The NVIDIA GeForce GTX 680M is the clear winner in the head-to-head comparison. In the single available Geekbench OpenCL test, the 680M scored 9230 against the 675M's 6946, resulting in a 32.9% performance advantage.
Q: How do these GPUs compare to other mobile and desktop parts in the database?
A: Both GPUs sit at the 39th percentile among all GPUs. The 680M's average benchmark score of 7023 places it nearly level with the NVIDIA T600 (7035, a -0.2% delta) and slightly above the AMD Radeon R5 M240 (6975, a 0.7% delta). The 675M, with an average score of 6946, trails the R5 M240 by -0.4% but leads the AMD FirePro M5100 by 1.7%.
Q: What is the primary architectural difference between the two chips?
A: The 680M uses the GK104 chip built on the Kepler architecture, fabricated on a 28 nm process at TSMC. The 675M uses the GF114 chip based on the older Fermi 2.0 architecture, manufactured on a 40 nm process. This node shrink is a fundamental difference in their design.
Q: Do both GPUs support the same modern APIs?
A: The API support is nearly identical, with a key exception. Both support DirectX 12 (11_0) and OpenGL 4.6. However, the 680M supports Vulkan 1.2.175, while the 675M has no listed Vulkan support.
Q: Which GPU has a higher transistor density?
A: The 680M has a significantly higher transistor density. It packs 3,540 million transistors into a 294 mm² die, yielding a density of 12.0M transistors per mm². The 675M has 1,950 million transistors on a larger 332 mm² die, resulting in a density of just 5.9M per mm².
Q: What are the memory specifications for each card?
A: The 680M comes with 4 GB of GDDR5 memory on a 256-bit bus, achieving 115.2 GB/s of bandwidth with an effective speed of 3.6 Gbps. The 675M has 2 GB of GDDR5 memory on the same 256-bit bus, but its bandwidth drops to 96.00 GB/s due to a lower 3 Gbps effective memory clock.
Architecture Differences
The generational leap between these two mobile GPUs is stark. The GTX 680M is built on the Kepler architecture using the GK104 chip, while the GTX 675M relies on the Fermi 2.0 architecture with the GF114 chip. This is not a minor revision; it is a complete overhaul of the GPU's compute paradigm. The most obvious indicator is the manufacturing process: the 680M uses a 28 nm node at TSMC, whereas the 675M uses a larger 40 nm process. This node difference directly explains the massive disparity in transistor counts and density. The 680M contains 3,540 million transistors on a 294 mm² die, while the 675M houses only 1,950 million on a larger 332 mm² die. Consequently, the 680M achieves a transistor density of 12.0M per mm², more than double the 675M's 5.9M per mm².
The compute configuration also diverges dramatically. The 680M features 1344 shading units, 112 texture mapping units (TMUs), and 32 ROPs. The 675M, in contrast, has only 384 shading units, 64 TMUs, and 32 ROPs. While the ROP count is identical, the sheer number of shaders and TMUs in the 680M points to a much wider execution engine. This is reflected in the theoretical throughput figures. The 680M has a pixel rate of 21.22 GPixel/s and a texture rate of 84.90 GTexel/s, compared to the 675M's 9.920 GPixel/s and 39.68 GTexel/s. More tellingly, the FP32 performance of the 680M is 2.038 TFLOPS, which is more than double the 675M's 952.3 GFLOPS.
Clock speeds also differ, though the 680M's advantage is not solely due to raw clock rate. The 680M has a base clock of 719 MHz and a boost clock of 758 MHz. The 675M does not list base or boost clocks, but its memory clock runs at 750 MHz (3 Gbps effective) compared to the 680M's 900 MHz (3.6 Gbps effective). The combination of a newer architecture, higher clocks, and a more robust feature set makes the 680M a fundamentally different product. The 675M is a mature Fermi design pushed into the 600M lineup, while the 680M is a next-generation Kepler part designed for higher efficiency and performance per watt.
Head-to-Head Benchmarks
The only direct benchmark comparison available is the Geekbench OpenCL test, and the results are decisively in favor of the GTX 680M. In this test, the 680M scored 9230, while the 675M scored 6946. This yields a delta of 32.9%, meaning the 680M is nearly a third faster than the 675M in this compute-oriented workload. This is a substantial gap that aligns with the architectural differences: the 680M's FP32 throughput of 2.038 TFLOPS dwarfs the 675M's 952.3 GFLOPS, and the 680M's higher memory bandwidth of 115.2 GB/s versus 96.00 GB/s helps feed that extra compute power.
Looking at the broader database context, the 680M's average benchmark score of 7023 is only -1.9% behind the NVIDIA GeForce GTX 970, a desktop-class part, and -0.2% behind the NVIDIA T600. The 675M's average score of 6946 is 1.7% ahead of the AMD FirePro M5100 but trails the AMD Radeon R5 M240 by -0.4%. When the two are pitted directly, the 680M's advantage is clear. The 675M's best-case scenario is to match the performance of the R5 M240, while the 680M sits in a higher performance tier altogether. The data suggests that the 680M is not just a refresh; it is a full generational step that places it closer to entry-level desktop GPUs than to its mobile sibling.
The Verdict
The data presents an unambiguous verdict: the NVIDIA GeForce GTX 680M is the superior GPU in every measurable way. It wins the only head-to-head benchmark by 32.9%, has a higher average benchmark score (7023 vs. 6946), and offers more than double the FP32 compute power and texture fill rate. For users seeking the highest available mobile performance in this generation, the 680M is the clear choice. Its 4 GB memory capacity and higher bandwidth also suggest better longevity for memory-intensive workloads.
The GTX 675M, however, is not without merit. It maintains the same TDP of 100 W and the same MXM-B (3.0) bus interface, making it a drop-in replacement for systems designed for the 680M. If power consumption is a primary concern, the 675M offers a way to get a discrete GeForce 600M experience at a lower performance tier. Yet, the 680M achieves its significant performance lead while keeping the TDP at the same 100 W, which is a remarkable efficiency gain from the 28 nm Kepler architecture. The 675M is a fallback option, but the 680M is the definitive product.
Specification Differences
The two GPUs differ on nearly every core specification, reflecting their different architectures. The 680M uses the GK104 chip (Kepler) on a 28 nm process, while the 675M uses the GF114 chip (Fermi 2.0) on a 40 nm process. Transistor counts differ massively: 3,540 million for the 680M versus 1,950 million for the 675M, with die sizes of 294 mm² and 332 mm² respectively. The 680M has a transistor density of 12.0M / mm², whereas the 675M has 5.9M / mm².
The memory subsystems are also different. The 680M offers 4 GB of GDDR5, while the 675M offers 2 GB. Both have a 256-bit bus, but the 680M's memory clock is 900 MHz (3.6 Gbps effective) versus 750 MHz (3 Gbps effective) on the 675M. This results in bandwidth of 115.2 GB/s for the 680M and 96.00 GB/s for the 675M. Compute resources differ: the 680M has 1344 shading units and 112 TMUs, compared to 384 shading units and 64 TMUs on the 675M. The ROP count is identical at 32. Pixel rate (21.22 vs. 9.920 GPixel/s), texture rate (84.90 vs. 39.68 GTexel/s), and FP32 (2.038 TFLOPS vs. 952.3 GFLOPS) are all dramatically higher on the 680M. The 680M also lists Vulkan 1.2.175 support, which the 675M lacks. Both share the same TDP, slot width, bus interface, display outputs, and API support for DirectX and OpenGL.
Where Each One Wins
NVIDIA GeForce GTX 680M: This GPU wins in every performance category derived from the data. It excels in compute-heavy tasks like OpenCL workloads, as evidenced by its 32.9% lead over the 675M. Its higher FP32 throughput (2.038 TFLOPS) makes it suitable for general-purpose GPU computing and physics simulations. The larger 4 GB frame buffer and higher memory bandwidth (115.2 GB/s) give it an edge in high-resolution texture loading and memory-intensive applications. For users who prioritize raw performance and future-proofing within the GeForce 600M generation, the 680M is the only logical choice.
NVIDIA GeForce GTX 675M: The 675M does not win any benchmark in the data, but it has specific use cases. It holds the advantage in terms of simpler compatibility with older systems, as it is based on the more mature Fermi architecture. For a laptop that requires a 100 W MXM module but does not need the absolute maximum performance, the 675M represents a more modest option that still delivers a 39th percentile standing. Its lower transistor count and simpler design might imply lower manufacturing complexity, though this is not quantified. In scenarios where Vulkan API support is irrelevant and memory capacity beyond 2 GB is unnecessary, the 675M functions adequately, but it is a compromise compared to the 680M.