NVIDIA GeForce 840M vs NVIDIA GeForce GTX 670MX Comparison

# The Verdict

The NVIDIA GeForce GTX 670MX is the clear performance winner in this comparison, taking both head-to-head benchmark wins with a 6.3% lead in Geekbench OpenCL and an 8.9% lead in Geekbench Vulkan. Its average benchmark score of 5721 versus the 840M's 5322 represents a 7.5% overall advantage, placing the 670MX in the 33rd percentile of all GPUs compared to the 840M's 31st percentile.

The GTX 670MX is the choice for users who prioritize raw compute throughput and have the power budget to accommodate a 75 W TDP. It delivers substantially higher pixel rate (12.02 GPixel/s vs 8.992 GPixel/s), texture rate (48.08 GTexel/s vs 17.98 GTexel/s), and FP32 performance (1,153.9 GFLOPS vs 863.2 GFLOPS). The 3 GB GDDR5 memory with a 192-bit bus and 67.20 GB/s bandwidth provides nearly four times the memory bandwidth of the 840M's 2 GB DDR3 setup.

The GeForce 840M is the better choice for thin-and-light portable systems where power efficiency matters more than peak performance. With a 33 W TDP, it draws less than half the power of the GTX 670MX while still delivering respectable OpenCL scores. The 840M also supports Vulkan 1.4, a newer API version than the 670MX's Vulkan 1.2.175, which could matter for future software compatibility.

For gaming at moderate settings in a portable chassis, the 840M's higher boost clock (1124 MHz vs 601 MHz) helps compensate for its smaller core configuration. However, for any workload that stresses memory bandwidth or raw shader throughput, the GTX 670MX dominates without question.

# Architecture Differences

The GTX 670MX uses the GK104 chip based on Kepler architecture, fabricated on TSMC's 28 nm process. It packs 3,540 million transistors across a 294 mm² die, yielding a transistor density of 12.0M per mm². The 840M uses the GM108S chip on Maxwell architecture, also on 28 nm, but with only 1,020 million transistors on a 77 mm² die, achieving a higher density of 13.2M per mm².

Core configuration differs dramatically. The GTX 670MX features 960 shading units, 80 texture mapping units, and 24 ROPs. The 840M has 384 shading units, 16 TMUs, and 8 ROPs. That means the 670MX has 2.5 times the shader count, 5 times the texture units, and 3 times the ROPs.

Clock speeds tell the opposite story. The 840M runs at 1029 MHz base with 1124 MHz boost, while the GTX 670MX is locked at 601 MHz for both base and boost. The 840M's memory runs at 1001 MHz (2 Gbps effective), compared to the 670MX's 700 MHz (2.8 Gbps effective). The 670MX's advantage comes from its wider 192-bit memory bus and GDDR5 memory type, yielding 67.20 GB/s bandwidth versus the 840M's narrow 64-bit bus and DDR3 memory at 16.02 GB/s.

Both chips support DirectX 12 (11_0) and OpenGL 4.6. The 840M supports Vulkan 1.4, while the 670MX supports Vulkan 1.2.175, a notable API generation difference. Neither chip has RT cores or tensor cores.

# Head-to-Head Benchmarks

The Geekbench OpenCL test shows the GTX 670MX scoring 6125 against the 840M's 5764, a 6.3% victory. This margin reflects the 670MX's substantial advantage in raw compute resources—960 shading units versus 384—even though the 840M's higher clocks (1124 MHz boost vs 601 MHz) narrow the gap. The 670MX's FP32 throughput of 1,153.9 GFLOPS versus 863.2 GFLOPS for the 840M translates directly to OpenCL workloads that scale with shader count.

The Geekbench Vulkan test shows a larger gap: 5316 for the 670MX versus 4880 for the 840M, an 8.9% delta. Vulkan workloads often stress memory bandwidth and vertex processing, areas where the 670MX's 67.20 GB/s bandwidth and 24 ROPs give it a decisive edge over the 840M's 16.02 GB/s and 8 ROPs. The 840M's newer Vulkan 1.4 API support doesn't compensate for its hardware limitations in this test.

In the broader competitive landscape, the GTX 670MX sits within 0.2% of the Intel Iris Pro Graphics P6300 (5712) and 0.2% behind the NVIDIA GeForce GTX 550 Ti (5731), while leading the AMD Radeon HD 8790M (5691) by 0.5% and the NVIDIA Quadro M500M (5604) by 2.1%. The 840M is nearly identical to the NVIDIA GeForce 930A (5317) at 0.1% delta, leads the NVIDIA GeForce GTX 980M (5308) by 0.3%, trails the AMD Radeon R7 M445 (5358) by 0.7%, and leads the NVIDIA GeForce 940M (5284) by 0.7%.

# Specification Differences

| Specification | GTX 670MX | GeForce 840M |

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

| Chip | GK104 | GM108S |

| Architecture | Kepler | Maxwell |

| Generation | GeForce 600M | GeForce 800M |

| Transistors | 3,540 million | 1,020 million |

| Die Size | 294 mm² | 77 mm² |

| Transistor Density | 12.0M / mm² | 13.2M / mm² |

| Base Clock | 601 MHz | 1029 MHz |

| Boost Clock | 601 MHz | 1124 MHz |

| Memory Clock | 700 MHz / 2.8 Gbps effective | 1001 MHz / 2 Gbps effective |

| Memory Size | 3 GB | 2 GB |

| Memory Type | GDDR5 | DDR3 |

| Memory Bus Width | 192 bit | 64 bit |

| Memory Bandwidth | 67.20 GB/s | 16.02 GB/s |

| Shading Units | 960 | 384 |

| TMUs | 80 | 16 |

| ROPs | 24 | 8 |

| Pixel Rate | 12.02 GPixel/s | 8.992 GPixel/s |

| Texture Rate | 48.08 GTexel/s | 17.98 GTexel/s |

| FP32 Performance | 1,153.9 GFLOPS | 863.2 GFLOPS |

| TDP | 75 W | 33 W |

| Bus Interface | PCIe 3.0 x16 | PCIe 3.0 x8 |

| Vulkan Support | 1.2.175 | 1.4 |

| Release Date | 2012-09-30 | 2014-03-11 |

| Predecessor | GeForce 500M | GeForce 700M |

| Successor | GeForce 700M | GeForce 900M |

# FAQ

Q: Which GPU is faster in OpenCL workloads?

A: The GTX 670MX scores 6125 in Geekbench OpenCL versus 5764 for the 840M, a 6.3% advantage. This comes from its 960 shading units and 67.20 GB/s memory bandwidth.

Q: Does the 840M have any advantage in API support?

A: Yes, the 840M supports Vulkan 1.4 while the GTX 670MX is limited to Vulkan 1.2.175. Both support DirectX 12 (11_0) and OpenGL 4.6.

Q: Why is the 840M's clock speed much higher?

A: The 840M runs at 1029 MHz base and 1124 MHz boost, compared to the 670MX's fixed 601 MHz. However, the 670MX compensates with 2.5 times more shading units and a much wider memory bus.

Q: How much memory bandwidth difference exists between these two?

A: The GTX 670MX provides 67.20 GB/s from its 192-bit GDDR5 interface, while the 840M offers 16.02 GB/s from a 64-bit DDR3 interface—a 4.2 times difference.

Q: Which GPU is more power efficient?

A: The 840M has a 33 W TDP versus 75 W for the GTX 670MX. The 840M delivers 863.2 GFLOPS at its power level, while the 670MX delivers 1,153.9 GFLOPS at more than double the power draw.

Q: What are the nearest competitors for each GPU?

A: The GTX 670MX sits near the Intel Iris Pro Graphics P6300 (5712, 0.2% delta) and NVIDIA GeForce GTX 550 Ti (5731, -0.2% delta). The 840M is nearly matched with the NVIDIA GeForce 930A (5317, 0.1% delta) and GeForce GTX 980M (5308, 0.3% delta).

# Where Each One Wins

The GTX 670MX wins in every measured benchmark category. Its advantages are most pronounced in memory-intensive workloads—the 67.20 GB/s bandwidth versus 16.02 GB/s makes it ideal for high-resolution textures, large framebuffers, and compute tasks that stream data. The 3 GB GDDR5 frame buffer provides more headroom for modern game assets than the 840M's 2 GB DDR3. The 48.08 GTexel/s texture rate versus 17.98 GTexel/s means texture-heavy scenes will render significantly faster.

The 840M wins in power efficiency, drawing 33 W versus 75 W, making it suitable for ultra-thin laptops without dedicated cooling solutions. Its higher transistor density (13.2M per mm² vs 12.0M per mm²) reflects a more modern design that achieves respectable performance in a much smaller package. The Vulkan 1.4 support means it can run software that requires newer API features, whereas the 670MX is capped at Vulkan 1.2.175.

For gaming, the GTX 670MX is the clear choice for any title that stresses memory bandwidth or relies on heavy pixel throughput. The 840M can handle lighter esports titles and older games at lower settings, where its higher boost clock (1124 MHz) helps mask its smaller core configuration. For compute workloads like OpenCL general-purpose processing, the 670MX's 1,153.9 GFLOPS FP32 performance and 960 shaders provide a substantial advantage.

The production timeline also matters: the 670MX launched as part of the GeForce 600M generation (predecessor: GeForce 500M), while the 840M belongs to the GeForce 800M generation (predecessor: GeForce 700M). Both are now end-of-life products, but the 840M is the more recent release.