AMD Radeon HD 6770 vs NVIDIA GeForce GT 640 Comparison

The AMD Radeon HD 6770 and NVIDIA GeForce GT 640 are two end-of-life graphics cards from different architectural generations, targeting the entry-level segment of their respective eras. The benchmark data shows a narrow contest overall, with the GT 640 edging out the HD 6770 in the sole head-to-head compute test, yet the two cards present distinctly different profiles in memory technology, power requirements, and API support that could tip a purchasing decision.

Where Each One Wins

The data presents a clear but limited picture: in the only shared benchmark test (Geekbench OpenCL), the NVIDIA GeForce GT 640 scores 3736, which is 2.3% higher than the AMD Radeon HD 6770’s 3649. This makes the GT 640 the winner in the single head-to-head comparison available. However, this does not tell the entire story.

The AMD card’s advantage lies in its raw compute specification. With 800 shading units, 40 texture mapping units, and a peak FP32 throughput of 1,360.0 GFLOPS, the HD 6770 is built for raw parallel number-crunching. In contrast, the GT 640 has 384 shading units and 32 TMUs, delivering 692.7 GFLOPS of FP32 performance. The HD 6770’s compute ceiling is roughly double that of the GT 640, suggesting that in workloads not limited by memory bandwidth or driver optimization, the AMD part should pull ahead. Its pixel rate of 13.60 GPixel/s versus 7.216 GPixel/s also indicates faster fill-rate-bound work.

The GT 640 wins on memory capacity and modern feature support. It carries 2 GB of DDR3 memory versus 1 GB of GDDR5 on the HD 6770. While the GDDR5 is much faster in bandwidth (76.80 GB/s versus 28.51 GB/s), the GT 640’s larger frame buffer allows for higher texture resolutions and more data in flight without spilling to system memory. The GT 640 also supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175, while the HD 6770 tops out at DirectX 11.2 (11_0) and OpenGL 4.4 with no Vulkan support. For modern titles or compute APIs, the GT 640 is the only viable option.

Architecture Differences

The two cards come from fundamentally different design philosophies. The AMD Radeon HD 6770 uses the Juniper chip built on TeraScale 2 architecture, fabricated on a 40 nm process at TSMC. It packs 1,040 million transistors onto a 166 mm² die, yielding a transistor density of 6.3 million per square millimeter. This is the older, larger process node.

NVIDIA’s GeForce GT 640 uses the GK107 chip built on Kepler architecture, also fabricated by TSMC but on a 28 nm process. This allows 1,270 million transistors to fit on a much smaller 118 mm² die, achieving a transistor density of 10.8 million per square millimeter. The difference in process technology explains why the GT 640 has a significantly lower thermal design power of 65 W versus the HD 6770’s 108 W, despite having more transistors.

Memory architecture diverges sharply. The HD 6770 uses 1 GB of GDDR5 on a 128-bit bus, achieving 76.80 GB/s of bandwidth. The GT 640 uses 2 GB of DDR3 on the same 128-bit bus, but only reaches 28.51 GB/s. This is the most consequential spec difference: the AMD card has 2.7 times the memory bandwidth, which is critical for texture-heavy scenes and high-resolution rendering. The GT 640 compensates with double the capacity, which helps with modern game asset sizes but cannot match the raw data throughput.

Other differences include the memory clock: the HD 6770 runs at 1200 MHz (4.8 Gbps effective) while the GT 640 runs at 891 MHz (1782 Mbps effective). The power connectors also differ: the HD 6770 requires a single 6-pin connector and a 300 W power supply, while the GT 640 has no external power connectors and only needs a 250 W PSU. The physical cards differ as well: the HD 6770 is a dual-slot design measuring 198 mm (7.8 inches) in length, while the GT 640 is a single-slot card at 145 mm (5.7 inches). The GT 640 also supports PCIe 3.0 x16 versus PCIe 2.0 x16 on the AMD card, and offers one DisplayPort 1.2 output compared to two mini-DisplayPort 1.1 on the HD 6770.

The Verdict

From the data, the NVIDIA GeForce GT 640 is the safer all-around choice. It wins the only head-to-head OpenCL benchmark, offers double the memory capacity, supports modern APIs including Vulkan and DirectX 12, consumes 43 W less power, requires no external power connector, and fits in a smaller single-slot form factor. Its 28 nm process yields a more efficient design.

The AMD Radeon HD 6770 is for a specific use case: raw compute throughput and memory bandwidth without regard for modern API support. Its 1,360.0 GFLOPS FP32 performance and 76.80 GB/s of memory bandwidth are far superior on paper, and its 2.7 times higher bandwidth could make it faster in bandwidth-bound legacy applications that do not leverage modern APIs. Its 40 TMUs and 13.60 GPixel/s pixel rate indicate stronger texture and pixel fill capabilities.

Benchmark percentiles place the HD 6770 at the 22nd percentile of all GPUs and the GT 640 at the 20th percentile, meaning both are in the lower quartile of performance. However, the HD 6770’s nearest rivals include the NVIDIA GeForce GTX 1050 (scoring 3629, a 0.6% deficit) and the GeForce 825M (3694, a 1.2% surplus), showing it is competitive with a much newer mainstream card. The GT 640’s rivals include the Intel Arc Pro B60 (3182, a 0.9% deficit) and the Quadro P1000 (3163, a 1.5% deficit), placing it in a similar performance band.

For a user with a modern operating system and games that use Vulkan or DirectX 12, the GT 640 is the only functional choice. For a user running older software that relies on DirectX 11 and values raw memory bandwidth, the HD 6770’s specifications are superior. The GT 640’s lower power draw and lack of power connectors make it easier to install in legacy or low-wattage systems.

FAQ

Q: Which card is faster in the Geekbench OpenCL benchmark?

A: The NVIDIA GeForce GT 640 scores 3736, which is 2.3% higher than the AMD Radeon HD 6770’s 3649, making the GT 640 the winner in that specific test.

Q: How much memory does each card have, and what type?

A: The AMD Radeon HD 6770 has 1024 MB of GDDR5 memory, while the NVIDIA GeForce GT 640 has 2 GB of DDR3 memory. Both use a 128-bit memory bus.

Q: Does the GT 640 support Vulkan?

A: Yes, the NVIDIA GeForce GT 640 supports Vulkan version 1.2.175. The AMD Radeon HD 6770 has no Vulkan support listed.

Q: What are the power requirements for each card?

A: The AMD Radeon HD 6770 has a TDP of 108 W and requires a single 6-pin power connector and a 300 W power supply. The NVIDIA GeForce GT 640 has a TDP of 65 W, requires no power connectors, and needs only a 250 W power supply.

Q: Which card has higher memory bandwidth?

A: The AMD Radeon HD 6770 has a memory bandwidth of 76.80 GB/s, which is significantly higher than the NVIDIA GeForce GT 640’s 28.51 GB/s.

Q: What are the physical size differences between the two cards?

A: The AMD Radeon HD 6770 is a dual-slot card measuring 198 mm (7.8 inches) in length, while the NVIDIA GeForce GT 640 is a single-slot card measuring 145 mm (5.7 inches) in length.

Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test. The NVIDIA GeForce GT 640 scores 3736, and the AMD Radeon HD 6770 scores 3649. The delta percentage is -2.3% from the perspective of the AMD card, meaning the GT 640 is 2.3% faster. This is a narrow margin, indicating that the two cards are very close in this particular compute workload.

Despite the GT 640’s win, the AMD card’s compute specifications are dramatically higher. The HD 6770 has 1,360.0 GFLOPS of FP32 performance versus 692.7 GFLOPS for the GT 640 — a 96% advantage. Its pixel rate is 13.60 GPixel/s versus 7.216 GPixel/s, an 88% advantage. Its texture rate is 34.00 GTexel/s versus 28.86 GTexel/s, a 17.8% advantage. If the OpenCL test were purely a measure of shader throughput, the HD 6770 should have won by a wide margin.

The fact that it did not suggests the GT 640’s Kepler architecture is more efficient at executing the specific OpenCL workload, or that the GDDR5 memory bandwidth of the HD 6770 is not being fully utilized. The GT 640’s higher OpenCL score (3736) relative to its much lower theoretical FP32 output (692.7 GFLOPS) versus the HD 6770’s score (3649) relative to its FP32 output (1,360.0 GFLOPS) shows a stark efficiency gap. The GT 640 achieves roughly 5.4 points per GFLOPS, while the HD 6770 achieves only 2.7 points per GFLOPS.

The GT 640 also has additional benchmark results: it scores 2086 in Geekbench Metal and 3809 in Geekbench Vulkan. The HD 6770 has no results for those APIs. This means the GT 640 is the only one of the two that can be measured in modern cross-platform compute environments, giving it a functional advantage beyond raw speed.

Specification Differences

The following specifications differ between the two cards:

| Specification | AMD Radeon HD 6770 | NVIDIA GeForce GT 640 |

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

| Architecture | TeraScale 2 | Kepler |

| Process Node | 40 nm | 28 nm |

| Transistors | 1,040 million | 1,270 million |

| Die Size | 166 mm² | 118 mm² |

| Transistor Density | 6.3M / mm² | 10.8M / mm² |

| Memory Size | 1024 MB | 2 GB |

| Memory Type | GDDR5 | DDR3 |

| Memory Clock | 1200 MHz (4.8 Gbps effective) | 891 MHz (1782 Mbps effective) |

| Memory Bandwidth | 76.80 GB/s | 28.51 GB/s |

| Shading Units | 800 | 384 |

| TMUs | 40 | 32 |

| Pixel Rate | 13.60 GPixel/s | 7.216 GPixel/s |

| Texture Rate | 34.00 GTexel/s | 28.86 GTexel/s |

| FP32 Performance | 1,360.0 GFLOPS | 692.7 GFLOPS |

| TDP | 108 W | 65 W |

| Slot Width | Dual-slot | Single-slot |

| Power Connectors | 1x 6-pin | None |

| Suggested PSU | 300 W | 250 W |

| Bus Interface | PCIe 2.0 x16 | PCIe 3.0 x16 |

| Display Outputs | 2x DVI, 1x HDMI 1.3a, 2x mini-DisplayPort 1.1 | 1x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 |

| DirectX | 11.2 (11_0) | 12 (11_0) |

| OpenGL | 4.4 | 4.6 |

| Vulkan | None | 1.2.175 |

| Length | 198 mm (7.8 inches) | 145 mm (5.7 inches) |

| Launch MSRP | 129 USD | 99 USD |