NVIDIA GeForce GTX 670M vs NVIDIA GeForce GTX 770M Comparison

Head-to-Head Benchmarks

The data is unambiguous: the NVIDIA GeForce GTX 770M decisively outperforms the GTX 670M in the only direct benchmark where both were measured. In Geekbench OpenCL, the GTX 770M scores 8,552 points against the GTX 670M's 6,513 points. That works out to a 23.8% advantage for the newer card — a substantial margin that separates these two mobile GPUs by nearly a full performance tier. The GTX 670M's score, by contrast, places it in a tight pack of competitors: it sits just 0.3% ahead of the NVIDIA GeForce GT 555M, 0.5% ahead of the NVIDIA Quadro M5000M, and 0.6% ahead of the AMD Radeon Vega 10 Mobile. It trails the Intel UHD Graphics P750 by 0.6%, meaning the 670M is effectively surrounded by a cluster of GPUs within a 1.2% spread.

The GTX 770M's 8,552 OpenCL result is not just a win over its predecessor; it also positions the card differently in the competitive landscape. Its average benchmark score across all tests is 6,000, but the OpenCL result alone shows the card's peak compute capability. The GTX 770M finds itself in the company of the AMD Radeon RX 6400 (0% delta), the NVIDIA RTX PRO 6000 Blackwell Server (0.1% ahead), the AMD FirePro W4100 (0.2% ahead), and the NVIDIA Quadro K4000M (0.2% ahead). The fact that the GTX 770M matches the RX 6400 — a much newer discrete GPU — says something about the efficiency of Kepler's compute design. Meanwhile, the GTX 670M's percentile ranking among all GPUs is 38, which is actually higher than the GTX 770M's 35. This is a statistical artifact of the 770M's average being dragged down by its additional Metal benchmark score of 3,448, which is far lower than its OpenCL result. When looking solely at the head-to-head OpenCL comparison, the 770M is clearly the superior compute part.

Architecture Differences

The two GPUs come from entirely different architectural generations, and that explains the performance gap. The GTX 670M is built on Fermi 2.0, using the GF114 chip fabricated on a 40 nm process at TSMC. The GTX 770M moves to the Kepler architecture with the GK106 chip, also from TSMC but on a more advanced 28 nm node. The process shrink is significant: the 670M packs 1,950 million transistors into a 332 mm² die, yielding a transistor density of 5.9 million per square millimeter. The 770M crams 2,540 million transistors into a smaller 221 mm² die, which works out to 11.5 million transistors per square millimeter — nearly double the density. This is the clearest evidence of how much architectural efficiency improved between the 600M and 700M generations.

Core counts tell a similar story of generational leap. The GTX 670M has 336 shading units, 56 texture mapping units, and 24 ROPs. The GTX 770M more than doubles the shading units to 960, increases TMUs to 80, and keeps ROPs at 24. That massive increase in shading hardware is the primary driver behind the 770M's 23.8% OpenCL lead. Clock speeds also differ: the 670M has no listed base or boost clock, while the 770M runs at a 706 MHz base and 797 MHz boost. Memory clocks show a similar pattern — the 670M's memory runs at 750 MHz (3 Gbps effective), while the 770M's runs at 1002 MHz (4 Gbps effective). Both cards use GDDR5 on a 192-bit bus, but the 770M's higher memory clock pushes bandwidth from 72.00 GB/s to 96.19 GB/s, a 33.6% increase.

The compute throughput numbers illustrate the scale of the difference. The GTX 670M delivers 803.7 GFLOPS of FP32 performance, a pixel rate of 8.372 GPixel/s, and a texture rate of 33.49 GTexel/s. The GTX 770M nearly doubles FP32 throughput to 1.530 TFLOPS, raises pixel rate to 15.94 GPixel/s, and almost doubles texture rate to 63.76 GTexel/s. The TDP for both cards is identical at 75 W, so the 770M achieves this performance without drawing more power — a direct payoff of the 28 nm node. Both are MXM modules with no power connectors, and both use the MXM-B (3.0) bus interface. Display outputs are portable-device-dependent on both. In terms of API support, both cards support DirectX 12 (11_0) and OpenGL 4.6, but the 770M adds Vulkan 1.2.175 support, which the 670M lacks entirely.

Where Each One Wins

The GTX 770M wins on raw compute, and the data shows it wins across every measurable performance metric. In OpenCL, it leads by 23.8%. In shading throughput, it more than doubles the 670M's FP32 output. In texture fill, it nearly doubles the rate. In pixel fill, it improves by roughly 90%. In memory bandwidth, it leads by 33.6%. The 770M also has 3 GB of VRAM versus the 670M's 1536 MB, which matters for larger textures and higher resolutions. If a workload is compute-bound, memory-bound, or texture-bound, the 770M is the clear choice.

The GTX 670M's advantages are narrower but real. Its percentile ranking of 38 versus the 770M's 35 means that, across the entire benchmark database, the 670M sits in a slightly higher relative position — though this is influenced by the 770M's weaker Metal score of 3,448, which drags down its average. The 670M also has a launch date of March 2012, making it a year older, which could mean better driver maturity in legacy applications. Its 40 nm node is less efficient, but the card's lower shading unit count (336) means it draws the same 75 W TDP while producing less heat density in some scenarios — though the data does not quantify thermals.

For gaming, the 770M's 3 GB frame buffer and 96.19 GB/s bandwidth make it better suited to modern titles with high-resolution textures. The 670M's 1536 MB buffer will hit capacity limits far sooner. For compute tasks like OpenCL-based rendering, physics simulation, or machine learning inference, the 770M's 960 shading units and 1.530 TFLOPS are in a different league. The 670M's only realistic win scenario is in very old or lightweight workloads where the extra shading units go unused, or in systems where the 670M's lower transistor count (1,950 million vs 2,540 million) might theoretically correlate with better yields — but that is not a performance metric. The verdict is straightforward: the 770M wins the head-to-head and wins it by a wide margin.

FAQ

Q: How much faster is the GTX 770M than the GTX 670M in OpenCL?

A: The GTX 770M scores 8,552 in Geekbench OpenCL versus 6,513 for the GTX 670M, giving the 770M a 23.8% lead.

Q: Do both cards have the same power draw?

A: Yes, both are rated at 75 W TDP, and both are MXM modules with no power connectors.

Q: Which card has more VRAM?

A: The GTX 770M has 3 GB of GDDR5, while the GTX 670M has 1536 MB of GDDR5. Both use a 192-bit memory bus.

Q: Does the GTX 670M support Vulkan?

A: No. The GTX 670M supports DirectX 12 (11_0) and OpenGL 4.6, but Vulkan support is only listed for the GTX 770M (Vulkan 1.2.175).

Q: How does the GTX 670M compare to its nearest rivals?

A: The GTX 670M is 0.3% ahead of the GT 555M, 0.5% ahead of the Quadro M5000M, 0.6% ahead of the Radeon Vega 10 Mobile, and 0.6% behind the Intel UHD Graphics P750.

Q: What is the GTX 770M's average benchmark score?

A: The GTX 770M has an average benchmark score of 6,000, which comes from its OpenCL score of 8,552 and its Metal score of 3,448.

The Verdict

The data supports only one conclusion: the NVIDIA GeForce GTX 770M is the superior GPU. Its 23.8% OpenCL advantage over the GTX 670M is decisive, and its architecture is fundamentally more capable. The 770M's Kepler architecture on 28 nm delivers 2,540 million transistors in a smaller 221 mm² die, whereas the 670M's Fermi 2.0 on 40 nm uses 1,950 million transistors in a larger 332 mm² package. The 770M has 960 shading units against 336, 80 TMUs against 56, and identical 24 ROPs. It has 3 GB of VRAM versus 1536 MB, and 96.19 GB/s bandwidth versus 72.00 GB/s. It doubles FP32 throughput to 1.530 TFLOPS. It does all this at the same 75 W TDP. There is no workload category in the data where the 670M wins.

For a user choosing between these two, the GTX 770M is the pick for anyone who runs modern games, OpenCL compute, or any graphics-intensive application. The extra VRAM alone justifies it — 3 GB versus 1536 MB is a 100% increase in capacity. The 670M's only distinction is its slightly higher percentile rank of 38 versus 35, which reflects its tighter clustering among similar-performance rivals rather than any actual performance advantage. The GTX 670M makes sense only if a system is physically limited to an older MXM-B (3.0) module with specific power delivery, but even then, the 770M uses the same slot and power envelope. The GTX 770M is the clear winner in this comparison, and the benchmark data says so with a 23.8% margin that is hard to argue against.

Specification Differences

| Specification | NVIDIA GeForce GTX 670M | NVIDIA GeForce GTX 770M |

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

| Architecture | Fermi 2.0 | Kepler |

| Chip | GF114 | GK106 |

| Process Node | 40 nm | 28 nm |

| Transistors | 1,950 million | 2,540 million |

| Die Size | 332 mm² | 221 mm² |

| Transistor Density | 5.9M / mm² | 11.5M / mm² |

| Base Clock | — | 706 MHz |

| Boost Clock | — | 797 MHz |

| Memory Clock | 750 MHz (3 Gbps) | 1002 MHz (4 Gbps) |

| Memory Size | 1536 MB | 3 GB |

| Memory Bandwidth | 72.00 GB/s | 96.19 GB/s |

| Shading Units | 336 | 960 |

| TMUs | 56 | 80 |

| Pixel Rate | 8.372 GPixel/s | 15.94 GPixel/s |

| Texture Rate | 33.49 GTexel/s | 63.76 GTexel/s |

| FP32 | 803.7 GFLOPS | 1.530 TFLOPS |

| Vulkan | — | 1.2.175 |

| Release Date | 2012-03-21 | 2013-05-29 |

| Generation | GeForce 600M | GeForce 700M |

| Predecessor | GeForce 500M | GeForce 600M |

| Successor | GeForce 700M | GeForce 800M |