NVIDIA GeForce GT 640 vs NVIDIA GeForce GTX 860M Comparison

The NVIDIA GeForce GTX 860M is decisively faster than the NVIDIA GeForce GT 640 across every common benchmark, with the data showing a roughly 2.3x to 2.8x performance advantage depending on the workload. The GT 640, an early Kepler desktop part, is outclassed by the later, higher-end GK104 chip in the GTX 860M, which was designed for gaming notebooks. While both cards are end-of-life products, the benchmark results make the hierarchy clear: the GTX 860M is the superior performer, and the GT 640 sits in a much lower performance tier.

Head-to-Head Benchmarks

The GTX 860M wins all three head-to-head benchmark comparisons, and the margins are substantial. In the Geekbench Metal test, the GTX 860M scores 4900 points against the GT 640's 2086 points. This represents a delta of -57.4% from the perspective of the GT 640, meaning the GTX 860M is roughly 135% faster. In OpenCL, the gap widens further: the GTX 860M posts 10461 points versus 3736 for the GT 640, a delta of -64.3%. This is the largest percentage difference in the head-to-head set, indicating the GTX 860M's compute capabilities scale particularly well in this API.

The Vulkan results tell a similar story. The GTX 860M scores 9648 points, while the GT 640 manages 3809 points, a delta of -60.5%. This consistency across Metal, OpenCL, and Vulkan suggests the GTX 860M's advantage is not tied to a single driver or API optimization but reflects a fundamental hardware superiority. The GT 640's average benchmark score across all tests is 3210 points, while the GTX 860M's average is 2967 points — but this aggregate figure is misleading because the GTX 860M includes several low-scoring Passmark tests in its dataset, which drag its average down. The direct head-to-head comparisons are the more reliable indicator, and they all point decisively in one direction.

Looking at the broader benchmark suite for the GTX 860M, its Passmark scores reveal a mixed profile. It scores 3081 in Passmark G3D and 1251 in GPU Compute, but its DirectX-specific scores are low: 55 in DirectX 9, 23 in DirectX 11, 15 in DirectX 10, and 13 in DirectX 12. These DirectX numbers appear anomalously low compared to its Geekbench scores, suggesting the Passmark tests may be measuring something different, such as specific driver paths or legacy workloads. Regardless, the head-to-head Geekbench data is unambiguous: the GTX 860M beats the GT 640 by a wide margin in every test where both cards have comparable data.

Where Each One Wins

The GTX 860M wins everywhere the data provides a direct comparison. In Geekbench Metal, OpenCL, and Vulkan, it dominates the GT 640 with no contest. The GT 640 has zero benchmark wins in the head-to-head table, and the GTX 860M has three. For users running compute-intensive applications that leverage OpenCL or Vulkan, the GTX 860M is the clear choice. Its 10461 OpenCL score is nearly triple the GT 640's 3736, making it suitable for tasks like video encoding, physics simulation, or any GPU-accelerated compute workload.

The GT 640's only advantage is its form factor and power characteristics. It is a single-slot desktop card with a 65 W TDP and no power connectors, drawing power entirely from the PCIe slot. The GTX 860M, by contrast, is an IGP (integrated graphics processor) designed for laptops, with a 75 W TDP and no standalone dimensions listed. For a desktop user with a low-power system, the GT 640 might be the only physically compatible option — but that is a hardware constraint, not a performance win. In terms of raw output, the GT 640's pixel rate is 7.216 GPixel/s and its texture rate is 28.86 GTexel/s, while the GTX 860M achieves 21.96 GPixel/s and 87.84 GTexel/s. The GTX 860M is roughly 3x faster in both rasterization and texturing throughput.

The GT 640 does have a slight edge in its average benchmark score percentile. It sits at the 20th percentile of all GPUs, while the GTX 860M sits at the 19th percentile. This is a marginal difference and likely reflects the GTX 860M's inclusion of low Passmark scores in its average. Looking at nearest rivals, the GT 640 is closest to the NVIDIA GeForce 920M, which scores 3287 with a delta of -2.3%, and the NVIDIA GeForce GT 730M at 3316 with a delta of -3.2%. The GTX 860M's nearest rival is the NVIDIA GeForce 820A at 2983 with a delta of -0.5%, and the NVIDIA GeForce GTX 750 Ti at 2953 with a delta of 0.5%. These comparisons show that the GT 640 and GTX 860M are in different performance neighborhoods entirely — the GT 640 competes with low-end mobile chips, while the GTX 860M trades blows with the GTX 750 Ti.

The Verdict

The verdict is straightforward: the GTX 860M is the superior GPU for any performance-sensitive task. The data shows it wins all three head-to-head benchmarks with deltas ranging from -57.4% to -64.3%, which translates to roughly 2.3x to 2.8x the performance of the GT 640. The GTX 860M's shading unit count of 1152 versus the GT 640's 384 explains much of this gap, as does its 96 texture mapping units versus 32. The GTX 860M also has a boost clock of 915 MHz, while the GT 640 has no listed base or boost clock, and its memory runs at 5 Gbps effective versus the GT 640's 1782 Mbps effective.

For a gamer or content creator, the GTX 860M is the only rational choice. Its 2 GB of GDDR5 memory provides 80.00 GB/s of bandwidth, nearly triple the GT 640's 28.51 GB/s from its 2 GB of DDR3. This bandwidth advantage is critical for modern game textures and compute workloads. The GT 640's sole purpose in a modern context would be as a basic display adapter for a legacy desktop system, where its 65 W TDP and single-slot design are assets. But even then, the GTX 860M's IGP form factor means it is not a drop-in replacement — it is a mobile part. The choice is therefore not between two competing products but between a low-end desktop card and a mid-range laptop chip, and the benchmark data consistently favors the latter.

FAQ

Q: Which GPU has a higher OpenCL benchmark score?

A: The NVIDIA GeForce GTX 860M scores 10461 in Geekbench OpenCL, while the NVIDIA GeForce GT 640 scores 3736, making the GTX 860M approximately 180% faster in this test.

Q: How do the two cards compare in terms of memory bandwidth?

A: The GTX 860M has 80.00 GB/s of bandwidth from its GDDR5 memory, while the GT 640 has 28.51 GB/s from DDR3 memory, a difference of roughly 2.8x.

Q: What is the GTX 860M's closest rival according to benchmark averages?

A: The GTX 860M's nearest rival is the NVIDIA GeForce 820A, which has an average score of 2983 and a delta of -0.5%, meaning the 820A is slightly slower.

Q: Does the GT 640 win any benchmark comparisons against the GTX 860M?

A: No, the head-to-head data shows the GT 640 wins zero benchmarks, while the GTX 860M wins all three tests (Geekbench Metal, OpenCL, and Vulkan).

Q: What are the pixel rates of these two GPUs?

A: The GTX 860M has a pixel rate of 21.96 GPixel/s, while the GT 640 has a pixel rate of 7.216 GPixel/s, making the GTX 860M about 3x faster in pixel fill rate.

Q: Which GPU has a higher transistor count?

A: The GTX 860M uses the GK104 chip with 3,540 million transistors, while the GT 640 uses the GK107 chip with 1,270 million transistors, a difference of roughly 2.8x.

Architecture Differences

Both GPUs are built on NVIDIA's Kepler architecture and manufactured by TSMC on a 28 nm process, but they use different chips with significant differences. The GT 640 uses the GK107 chip with 1,270 million transistors on a die size of 118 mm², resulting in a transistor density of 10.8M per mm². The GTX 860M uses the much larger GK104 chip with 3,540 million transistors on a 294 mm² die, giving it a transistor density of 12.0M per mm². This means the GTX 860M has nearly 2.8x the transistor count of the GT 640, which explains its substantially higher shading unit count: 1152 versus 384.

The memory subsystems differ in type and speed. The GT 640 uses DDR3 memory running at 891 MHz with 1782 Mbps effective, while the GTX 860M uses GDDR5 memory at 1250 MHz with 5 Gbps effective. Both have a 128-bit memory bus, but the GDDR5's higher data rate gives the GTX 860M 80.00 GB/s of bandwidth versus the GT 640's 28.51 GB/s. In terms of compute resources, the GTX 860M has 96 TMUs versus the GT 640's 32, and both have 16 ROPs. The GTX 860M's floating point performance is 2.108 TFLOPS, while the GT 640 achieves 692.7 GFLOPS.

Both cards support the same API set: DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. They also share the same PCIe 3.0 x16 bus interface. The GT 640 is a desktop card with display outputs of 1x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2, while the GTX 860M's display outputs are listed as "Portable Device Dependent," reflecting its laptop-oriented design.

Specification Differences

The most notable specification differences are in the core config and memory. The GTX 860M has 1152 shading units and 96 TMUs, versus the GT 640's 384 shading units and 32 TMUs. Both have 16 ROPs, but the GTX 860M's clock speeds are specified — 797 MHz base and 915 MHz boost — while the GT 640 has no listed base or boost clock. The GTX 860M's memory runs at 1250 MHz with 5 Gbps effective data rate, compared to the GT 640's 891 MHz with 1782 Mbps effective.

The TDPs are close: the GTX 860M is rated at 75 W, while the GT 640 is rated at 65 W. The GT 640 has a suggested PSU of 250 W and is a single-slot card with a length of 145 mm (5.7 inches), while the GTX 860M has no listed dimensions and is classified as an IGP. Neither card requires power connectors. The GT 640 was released on 2012-06-04 and has a launch MSRP of 99 USD; the GTX 860M was released on 2014-03-09 with no listed launch MSRP. The GT 640's successor is GeForce 700, while the GTX 860M's successor is GeForce 900M. Both cards are end-of-life products, and the GTX 860M has a slightly higher transistor density at 12.0M per mm² versus 10.8M per mm².