NVIDIA GeForce GTX 780

The NVIDIA GeForce GTX 780 is an end-of-life desktop graphics card built on the Kepler architecture, utilizing the GK110 chip. Fabricated by TSMC on a 28 nm process, the die measures 561 mm² and contains 7,080 million transistors, yielding a density of 12.6 million per square millimeter. The card launched on 2013-05-22 with a launch MSRP of 649 USD, succeeding the GeForce 600 series and preceding the GeForce 900 series. It features 2,304 shading units, 192 TMUs, and 48 ROPs, paired with 3 GB of GDDR5 memory on a 384-bit bus for 288.4 GB/s of bandwidth. Memory runs at 1502 MHz (6 Gbps effective). The bus interface is PCIe 3.0 x16. In the benchmark database, the GTX 780 achieves an average score of 19,405, placing it in the 62nd percentile of all GPUs. This makes it a solid mid-to-high performer within its generation.

Power and Cooling

The GTX 780 has a TDP of 250 W, a figure that directly informs the suggested PSU rating of 600 W. Power delivery relies on a combination of one 6-pin and one 8-pin connector, which together supply the necessary current for the GK110 chip. The dual-slot cooler is a mandatory design choice given the thermal output; a single-slot design would be insufficient. The card's physical footprint is substantial: 267 mm in length (10.5 inches), 111 mm in height (4.4 inches), and 38 mm in width (1.5 inches). This length requires a case with adequate clearance, and the height is standard for a dual-slot PCIe card. The 28 nm process node, while mature, still results in a 561 mm² die, which is large even by modern standards. The transistor density of 12.6 million per square millimeter is a direct consequence of the architecture's complexity. Users should verify that their power supply has both the wattage headroom and the required 6-pin and 8-pin connectors before installation. The 250 W TDP is a hard limit; exceeding it without proper cooling will throttle performance. The dual-slot cooler also affects case airflow, as the card exhausts heat into the chassis. For systems with multiple GPUs, the 600 W PSU recommendation becomes even more critical, though the data only specifies a single-card configuration. The 38 mm width (1.5 inches) is typical for a dual-slot design, ensuring compatibility with most motherboards. The 250 W TDP is a significant draw, and the 600 W PSU recommendation provides a reasonable headroom for the rest of the system. The specific connector arrangement is a fixed requirement, and any PSU lacking both connectors will not be able to power the card.

Ray Tracing and Feature Set

The GTX 780 does not integrate RT cores or tensor cores; these dedicated accelerators are absent from the Kepler architecture. Consequently, hardware-accelerated ray tracing and tensor-based machine learning workloads are not supported. The card's compute capabilities are defined by its 2,304 shading units, 192 texture mapping units, and 48 raster output units. The pixel fill rate is 43.30 GPixel/s, and the texture fill rate is 173.2 GTexel/s, while single-precision FP32 performance reaches 4.156 TFLOPS. Memory operates at 1502 MHz, translating to 6 Gbps effective, across a 384-bit bus for 288.4 GB/s of bandwidth. The bus interface is PCIe 3.0 x16, which provides sufficient bandwidth for the card's data needs. API support spans DirectX 12 (with feature level 11_0), OpenGL 4.6, and Vulkan 1.2.175. The DirectX 12 support is limited to the 11_0 feature level, meaning certain DX12 features are unavailable. Display outputs include two DVI ports, one HDMI 1.4a, and one DisplayPort 1.2, covering standard multi-monitor setups. The absence of RT and tensor cores means the card relies entirely on traditional rasterization and compute shaders for rendering. The 3 GB frame buffer is adequate for its era, and the 384-bit bus ensures high memory throughput. The FP32 throughput of 4.156 TFLOPS is a strong indicator of its compute potential in OpenCL and Vulkan workloads, which is reflected in the benchmark scores. The lack of tensor cores also means no DLSS or similar AI-based upscaling is available, and the DirectX 12 feature level 11_0 restricts certain modern rendering techniques.

Benchmark Performance

The GTX 780's average benchmark score is 19,405, which places it in the 62nd percentile of all GPUs in the database. This percentile indicates that it outperforms the majority of recorded graphics cards, though it does not approach the top tier. In individual tests, the card scores 10,728 in Geekbench Metal, 22,907 in Geekbench OpenCL, and 24,580 in Geekbench Vulkan. The Vulkan result is the highest of the three, followed by OpenCL, and Metal trails significantly. This pattern suggests that the card is better optimized for compute-heavy APIs like Vulkan and OpenCL than for Metal. Comparing to nearest rivals, the GTX 1080 Ti scores 19,402, a delta of 0% — a statistical tie. The AMD Radeon Pro 560X scores 19,426, which is 0.1% higher (delta -0.1%). The NVIDIA Tesla K40m scores 19,519, 0.6% higher (delta -0.6%). The NVIDIA Quadro K5200 scores 19,623, 1.1% higher (delta -1.1%). The data shows the GTX 780 is the slowest of the four, but the maximum deficit is only 1.1%. This tight clustering suggests that the GTX 780's performance is exceptionally well-matched to these professional and high-end consumer cards in compute benchmarks. The 62nd percentile reinforces its position as a solid mid-to-high performer. The average score of 19,405 is the reference point; all rivals fall within a 218-point range (from 19,402 to 19,623). The delta values are all negative for the rivals, indicating that the GTX 780 is the baseline against which the others are measured. In practical terms, a 1.1% difference is imperceptible in real-world usage, making the GTX 780 functionally equivalent to its nearest competitors. The Vulkan score of 24,580 is notably higher than the average, suggesting strong compute performance in that API, while the Metal score of 10,728 drags down the overall average.

FAQ

Q: What is the TDP and recommended power supply for the GTX 780?

A: The TDP is 250 W, and the suggested PSU rating is 600 W.

Q: Does the GTX 780 support ray tracing or tensor cores?

A: No. The card has no RT cores or tensor cores, so hardware ray tracing and tensor acceleration are not available.

Q: What API versions does the GTX 780 support?

A: It supports DirectX 12 (with feature level 11_0), OpenGL 4.6, and Vulkan 1.2.175.

Q: What is the memory configuration of the GTX 780?

A: It has 3 GB of GDDR5 memory on a 384-bit bus, providing 288.4 GB/s of bandwidth, with a memory clock of 1502 MHz (6 Gbps effective).

Q: What are the physical dimensions of the card?

A: The card measures 267 mm in length, 111 mm in height, and 38 mm in width (10.5 x 4.4 x 1.5 inches).

Q: What is the average benchmark score and percentile?

A: The average benchmark score is 19,405, placing it in the 62nd percentile of all GPUs.

How It Compares

NVIDIA GeForce GTX 1080 Ti: The GTX 1080 Ti achieves an average score of 19,402, exactly matching the GTX 780's 19,405 with a delta of 0%. This indicates that the two cards deliver virtually identical compute performance in the benchmark suite, despite the GTX 1080 Ti being a later generation. The GTX 780 holds its ground perfectly.

AMD Radeon Pro 560X: The Radeon Pro 560X posts a score of 19,426, which is 0.1% higher than the GTX 780 (delta -0.1%). This is a negligible difference, meaning the two cards are effectively interchangeable in terms of raw scores. The GTX 780 trails by a hair but remains competitive.

NVIDIA Tesla K40m: The Tesla K40m scores 19,519, a 0.6% advantage over the GTX 780 (delta -0.6%). The Tesla is a compute-focused card, and its slightly higher score reflects a modest edge in the tested workloads. The GTX 780 is 0.6% behind, a small but measurable gap.

NVIDIA Quadro K5200: The Quadro K5200 leads the group with a score of 19,623, which is 1.1% higher than the GTX 780 (delta -1.1%). This is the largest deficit among the four rivals, yet it remains under 2%. The GTX 780 is the slowest of the quartet but only by a narrow margin.