GPU Comparison

The NVIDIA GeForce GT 640 and the NVIDIA GeForce 820M share the same GK107 chip, but they are separated by nearly three years of mobile-first tuning. The data from the benchmark database shows a clear performance hierarchy, with the desktop-oriented GT 640 holding a commanding lead in the only directly comparable test.

Head-to-Head Benchmarks

The sole head-to-head benchmark available is Geekbench OpenCL, and it delivers a decisive verdict. The GT 640 scores 3,736 points, while the 820M trails significantly with 2,784 points. That is a 34.2% advantage for the GT 640. In practical terms, the older card is roughly one-third faster than the newer mobile part in this compute-oriented workload.

This dominance is consistent with the broader benchmark averages. The GT 640 carries an average benchmark score of 3,210 across three tests (Geekbench Metal, OpenCL, and Vulkan), placing it in the 20th percentile of all GPUs. The 820M, with only an OpenCL result on record, averages 2,784 and sits in the 18th percentile. While both parts occupy the lower end of the performance spectrum, the GT 640’s margin is substantial enough to matter for any compute-light task that offloads work to the GPU.

The wins tally reflects the lopsided nature of this matchup: the GT 640 claims 1 win in head-to-head tests, while the 820M has 0.

Architecture Differences

Despite their different model numbers and generations, both GPUs are built on the same fundamental architecture. Both use the GK107 chip built on Kepler architecture, fabricated on a 28 nm process at TSMC. The transistor count is identical at 1,270 million, and the die size is the same 118 mm², yielding an identical transistor density of 10.8M per mm².

The core configuration is also unchanged. Both parts feature 384 shading units, 32 texture mapping units, and 16 raster output units. Neither GPU includes ray tracing cores or tensor cores, which is expected for Kepler-era parts.

The memory subsystems are nearly identical as well. Both come with 2 GB of DDR3 memory on a 128-bit bus. The GT 640’s memory runs at 891 MHz (1,782 Mbps effective), producing a bandwidth of 28.51 GB/s. The 820M’s memory is clocked marginally higher at 901 MHz (1,802 Mbps effective), yielding 28.83 GB/s of bandwidth. That is a negligible 0.32 GB/s difference in favor of the 820M, far too small to offset the compute gap.

Where the two diverge is in clock-dependent throughput rates. The GT 640 achieves a pixel rate of 7.216 GPixel/s and a texture rate of 28.86 GTexel/s. The 820M, running at lower effective clocks, manages only 6.480 GPixel/s and 25.92 GTexel/s. Similarly, FP32 compute is rated at 692.7 GFLOPS for the GT 640 versus 622.1 GFLOPS for the 820M, a 70.6 GFLOPS gap that explains the OpenCL benchmark delta.

Power and physical design differ notably. The GT 640 has a 65 W TDP, is a single-slot card with no power connectors, and requires a 250 W suggested PSU. The 820M is a leaner 45 W TDP part, but it comes in an MXM Module form factor, meaning it is designed for laptops or modular systems rather than desktop expansion slots. The GT 640’s bus interface is PCIe 3.0 x16; the 820M also uses PCIe 3.0 x16, but its display outputs are listed as “Portable Device Dependent,” reflecting its mobile orientation.

API support is identical: both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The GT 640 offers 1x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2 outputs, while the 820M’s outputs are tied to whatever portable device it is integrated into.

FAQ

Q: Which GPU is faster in OpenCL performance?

A: The NVIDIA GeForce GT 640 is significantly faster, scoring 3,736 in Geekbench OpenCL versus 2,784 for the 820M. That is a 34.2% lead for the GT 640.

Q: Are these two GPUs based on the same chip?

A: Yes. Both the GT 640 and the 820M use the GK107 chip built on the Kepler architecture, with identical transistor counts of 1,270 million and the same 118 mm² die size.

Q: Do both cards have the same memory configuration?

A: Nearly identical. Both have 2 GB of DDR3 on a 128-bit bus. The 820M has a marginally higher memory clock (901 MHz versus 891 MHz), giving it 28.83 GB/s bandwidth versus 28.51 GB/s for the GT 640.

Q: What are the TDP differences?

A: The GT 640 has a 65 W TDP, while the 820M is rated at 45 W. The 820M is the more power-efficient part, though it is also a mobile-oriented MXM module rather than a desktop card.

Q: Which GPU has higher compute throughput?

A: The GT 640 leads with 692.7 GFLOPS FP32 performance, 7.216 GPixel/s pixel rate, and 28.86 GTexel/s texture rate. The 820M trails at 622.1 GFLOPS, 6.480 GPixel/s, and 25.92 GTexel/s, respectively.

Q: Do both support the same modern APIs?

A: Yes. Both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. This means software compatibility is effectively identical despite the performance gap.

The Verdict

The data presents a straightforward choice for anyone comparing these two parts side by side. The GT 640 is the stronger performer in every measured compute metric, and its 34.2% lead in Geekbench OpenCL is the definitive statistic. It also has a higher average benchmark score (3,210 versus 2,784) and a better percentile ranking (20th versus 18th). For any workload that leverages OpenCL, the GT 640 is the clear pick.

However, context matters. The 820M is a mobile part with a 45 W TDP, designed for portable devices where power draw and form factor are primary constraints. Its lower clocks and reduced throughput are the price of that mobility. The GT 640, at 65 W, is a desktop card with a 145 mm length and single-slot profile, requiring a 250 W PSU. It is not a drop-in replacement for a laptop GPU.

The 820M’s nearest rivals in the database are a mixed bag: it trails the GeForce RTX 3060 Ti GDDR6X by 0.4% and the AMD Radeon 550 by 0.6%, while leading the Intel UHD Graphics 610 by 1.2% and the RTX 5070 SUPER by 3.5%. Those deltas are small, indicating the 820M sits in a narrow performance band. The GT 640, meanwhile, is within 0.9% of the Intel Arc Pro B60 and 1.5% of the NVIDIA Quadro P1000, while trailing the GeForce 920M by 2.3% and the GT 730M by 3.2%.

For a desktop build where a dedicated low-profile card is acceptable, the GT 640 is the better choice based on raw performance. Its launch MSRP was 99 USD. For a laptop or MXM-based system, the 820M is the only viable option of the two, and its 45 W TDP makes it a reasonable fit for power-constrained environments. The verdict is not close on performance: the GT 640 wins decisively, but the 820M wins on portability and power efficiency.

Specification Differences

The following table highlights only the fields where the two GPUs differ:

| Specification | NVIDIA GeForce GT 640 | NVIDIA GeForce 820M |

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

| Generation | GeForce 600 | GeForce 800M |

| Memory Clock | 891 MHz (1,782 Mbps effective) | 901 MHz (1,802 Mbps effective) |

| Memory Bandwidth | 28.51 GB/s | 28.83 GB/s |

| Pixel Rate | 7.216 GPixel/s | 6.480 GPixel/s |

| Texture Rate | 28.86 GTexel/s | 25.92 GTexel/s |

| FP32 Performance | 692.7 GFLOPS | 622.1 GFLOPS |

| TDP | 65 W | 45 W |

| Slot Width | Single-slot | MXM Module |

| Power Connectors | None | Not specified |

| Suggested PSU | 250 W | Not specified |

| Display Outputs | 1x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 | Portable Device Dependent |

| Dimensions | 145 mm (5.7 inches) length | Not specified |

| Release Date | 2012-06-04 | 2015-03-20 |

| Predecessor | GeForce 500 | GeForce 700M |

| Successor | GeForce 700 | GeForce 900M |

| Launch MSRP | 99 USD | Not specified |

| Geekbench OpenCL Score | 3,736 | 2,784 |

| Average Benchmark Score | 3,210 | 2,784 |

| Percentile vs All GPUs | 20th | 18th |