AMD Radeon RX 5700 XT vs NVIDIA GeForce GTX 690 Comparison

Head-to-Head Benchmarks

The two GPUs split their two shared benchmark results exactly one win apiece, but the margins are anything but symmetrical. In Geekbench OpenCL, the NVIDIA GeForce GTX 690 posts a score of 17,399 against the AMD Radeon RX 5700 XT's 9,524 — a commanding 82.7% lead for the older Kepler card. That is not a marginal victory; it is a near-doubling of the AMD card's compute output in this particular test. The GTX 690's raw FP32 throughput of 3.130 TFLOPS, combined with its dual-GPU configuration, appears to give it a distinct advantage in OpenCL workloads that scale across multiple devices.

The reverse is true, and then some, in Geekbench Vulkan. Here the RX 5700 XT obliterates the GTX 690, scoring 68,310 versus 16,675 — a 75.6% deficit for the NVIDIA card. The delta is almost identical in magnitude to the OpenCL result, but in the opposite direction. This suggests the RX 5700 XT's modern RDNA 1.0 architecture handles Vulkan's explicit multi-threading and low-level GPU control far more efficiently than Kepler's older design. The GTX 690's Vulkan API support is listed at 1.2.175, while the RX 5700 XT supports Vulkan 1.4, which may explain part of the gulf.

Looking at the broader benchmark landscape, the average scores tell a nuanced story. The GTX 690's average benchmark score is 17,037, while the RX 5700 XT sits at 16,361 — a difference of roughly 4.1% in favor of the NVIDIA card. Yet the RX 5700 XT's percentile rank among all GPUs is 59, compared to the GTX 690's 60. These two metrics nearly overlap, placing both cards in the same performance tier despite their seven-year age gap. The GTX 690's nearest rivals include the NVIDIA GeForce RTX 3070 at 17,208 (a 1% delta) and the AMD Radeon RX 7600 XT at 17,083 (a 0.3% delta), while the RX 5700 XT sits near the AMD Radeon Pro 5600M (16,351, 0.1% delta) and the NVIDIA RTX PRO 6000 Blackwell (16,408, 0.3% delta). Neither card is an outlier in its peer group; both are firmly mid-pack performers.

Where Each One Wins

The GTX 690's victory in OpenCL is its only head-to-head win, but it is a decisive one. For compute workloads that rely on OpenCL — think scientific simulation, video encoding pipelines, or general-purpose GPU compute — the data shows the GTX 690 delivering 82.7% more performance than the RX 5700 XT. The dual-GPU design effectively doubles the shading units to 1,536 per GPU (3,072 total across both dies), and while that is not a linear scaling guarantee, the benchmark result suggests OpenCL is one workload where the Kepler architecture's multi-GPU approach pays off. The GTX 690 also holds advantages in raw memory bandwidth per GPU (192.3 GB/s per GPU, though the card is limited to 2 GB total VRAM), which may benefit certain compute patterns.

The RX 5700 XT's Vulkan win is equally one-sided, and it points to a different set of strengths. With 2,560 shading units, 160 texture mapping units, and 64 ROPs, the Navi 10 chip is substantially more powerful on paper than a single GK104 die. Its 9.754 TFLOPS of FP32 performance is more than triple the GTX 690's 3.130 TFLOPS, and its 448.0 GB/s memory bandwidth more than doubles the NVIDIA card's 192.3 GB/s. The Vulkan benchmark likely reflects this raw hardware advantage when the software stack is modern enough to exploit it. The RX 5700 XT also shows strong results in other tests: its PassMark G3D score of 16,277 and PassMark GPU Compute score of 7,650 indicate solid DirectX and compute performance, while its Geekbench Metal score of 74,647 suggests Apple ecosystem compatibility that the GTX 690 lacks entirely.

For gaming specifically, the RX 5700 XT's DirectX 12 (12_1) support versus the GTX 690's DirectX 12 (11_0) is a meaningful differentiator. The newer card can execute feature-level 12_1 workloads, which include features like conservative rasterization and rasterizer-ordered views, while the GTX 690 is limited to the older 11_0 feature set. This architectural gap, combined with the RX 5700 XT's 8 GB of GDDR6 memory versus the GTX 690's 2 GB of GDDR5, makes the AMD card the clear choice for modern, VRAM-hungry titles.

Architecture Differences

The two cards represent fundamentally different design philosophies separated by seven years of technological progress. The GTX 690 is built on NVIDIA's Kepler architecture, fabricated on TSMC's 28 nm process node. Its GK104 chip packs 3,540 million transistors into a 294 mm² die, yielding a transistor density of 12.0 million transistors per square millimeter. The RX 5700 XT, by contrast, uses AMD's RDNA 1.0 architecture on TSMC's 7 nm node. The Navi 10 chip contains 10,300 million transistors in a smaller 251 mm² die, achieving a density of 41.0 million transistors per square millimeter — more than three times the GTX 690's density.

The memory subsystems diverge sharply. The GTX 690 ships with 2 GB of GDDR5 across a 256-bit bus, delivering 192.3 GB/s of bandwidth. The RX 5700 XT offers 8 GB of GDDR6 across the same 256-bit bus width, but with 448.0 GB/s of bandwidth — a 133% increase. Memory clock speeds tell the story: the GTX 690 runs at 1502 MHz (6 Gbps effective) while the RX 5700 XT runs at 1750 MHz (14 Gbps effective). The GDDR6 standard's higher data rate per pin is the primary driver here.

Compute resources also favor the newer card dramatically. The RX 5700 XT's 2,560 shading units, 160 TMUs, and 64 ROPs compare favorably to the GTX 690's 1,536 shading units, 128 TMUs, and 32 ROPs — though it is worth remembering the GTX 690 is technically a dual-GPU card, so its per-GPU counts are 768 shading units, 64 TMUs, and 16 ROPs across two dies. The pixel rate difference is stark: 121.9 GPixel/s for the RX 5700 XT versus 32.61 GPixel/s for the GTX 690. Texture rate follows suit at 304.8 GTexel/s versus 130.4 GTexel/s. The RX 5700 XT also supports FP16 compute at 19.51 TFLOPS (2:1 ratio), a feature the GTX 690 lacks entirely.

Power and connectivity further separate the two. The GTX 690 demands 300 W TDP and dual 8-pin power connectors with a suggested 700 W PSU, while the RX 5700 XT draws 225 W from a single 6-pin plus 8-pin configuration and recommends a 550 W PSU. The RX 5700 XT also supports PCIe 4.0 x16, doubling the bandwidth of the GTX 690's PCIe 3.0 x16 interface. Display outputs differ as well: the GTX 690 offers 3x DVI and 1x mini-DisplayPort 1.2, whereas the RX 5700 XT provides 1x HDMI 2.0b and 3x DisplayPort 1.4a.

FAQ

Q: Which GPU is faster in OpenCL compute workloads?

A: The NVIDIA GeForce GTX 690 wins decisively in Geekbench OpenCL with a score of 17,399 versus the AMD Radeon RX 5700 XT's 9,524, a 82.7% advantage. This is likely due to the GTX 690's dual-GPU configuration scaling well in OpenCL.

Q: Does the RX 5700 XT beat the GTX 690 in any benchmark?

A: Yes, in Geekbench Vulkan the RX 5700 XT scores 68,310 against the GTX 690's 16,675, a 75.6% lead. The newer card's RDNA 1.0 architecture and Vulkan 1.4 support give it a major edge in this API.

Q: How do their average benchmark scores compare?

A: The GTX 690 averages 17,037 across benchmarks, while the RX 5700 XT averages 16,361. The GTX 690 holds a roughly 4.1% lead in average score, and both cards sit at nearly identical percentiles (60th and 59th, respectively).

Q: Which card has more memory and bandwidth?

A: The RX 5700 XT has 8 GB of GDDR6 memory with 448.0 GB/s bandwidth, while the GTX 690 has 2 GB of GDDR5 with 192.3 GB/s bandwidth. The AMD card offers four times the capacity and more than double the bandwidth.

Q: What are the power requirements for each card?

A: The GTX 690 has a 300 W TDP and requires two 8-pin power connectors with a suggested 700 W PSU. The RX 5700 XT has a 225 W TDP, uses one 6-pin and one 8-pin connector, and suggests a 550 W PSU.

Q: Which card supports newer PCIe and display standards?

A: The RX 5700 XT supports PCIe 4.0 x16 and offers HDMI 2.0b plus DisplayPort 1.4a outputs. The GTX 690 uses PCIe 3.0 x16 and provides DVI and mini-DisplayPort 1.2 connections.

The Verdict

The data presents a clear split based on workload and era. If the task is OpenCL compute, the NVIDIA GeForce GTX 690 is the surprising winner — its 82.7% margin over the RX 5700 XT in Geekbench OpenCL is not a small edge but a dominant one. The GTX 690's average benchmark score of 17,037 also edges out the RX 5700 XT's 16,361, and its 60th percentile rank slightly exceeds the AMD card's 59th. For legacy DirectX 9/10/11 workloads or multi-GPU compute tasks that can leverage dual dies, the Kepler card remains a viable, if power-hungry, option.

The RX 5700 XT, however, is the card for modern graphics. Its Vulkan performance is transformative — 75.6% faster than the GTX 690 — and its 8 GB of GDDR6 memory, 448.0 GB/s bandwidth, and 9.754 TFLOPS of FP32 compute make it the obvious choice for contemporary gaming and GPU-accelerated applications. The 7 nm process node, 41.0M / mm² transistor density, and PCIe 4.0 support indicate a fundamentally more advanced platform. Its DirectX 12 (12_1) feature level and Vulkan 1.4 support ensure compatibility with current and near-future software, while the GTX 690's DirectX 12 (11_0) and Vulkan 1.2.175 support are increasingly dated.

The power story also favors AMD: 225 W TDP versus 300 W, a single 6-pin plus 8-pin versus dual 8-pin, and a 550 W PSU recommendation versus 700 W. The RX 5700 XT delivers more performance per watt in modern APIs while being physically smaller (272 mm versus 279 mm in length, 36 mm versus 38 mm in width). For anyone building a system today, the RX 5700 XT is the rational choice. For a collector or researcher interested in peak OpenCL throughput from a historical dual-GPU flagship, the GTX 690's 82.7% OpenCL lead is a data point worth noting. Both cards are end-of-life, so either purchase is a legacy decision — but the benchmarks clearly favor the AMD card for anything beyond niche compute scenarios.

Specification Differences

| Specification | NVIDIA GeForce GTX 690 | AMD Radeon RX 5700 XT |

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

| Architecture | Kepler | RDNA 1.0 |

| Process Node | 28 nm | 7 nm |

| Transistors | 3,540 million | 10,300 million |

| Die Size | 294 mm² | 251 mm² |

| Transistor Density | 12.0M / mm² | 41.0M / mm² |

| Base Clock | 915 MHz | 1605 MHz |

| Boost Clock | 1019 MHz | 1905 MHz |

| Game Clock | — | 1755 MHz |

| Memory Clock | 1502 MHz (6 Gbps effective) | 1750 MHz (14 Gbps effective) |

| Memory Size | 2 GB | 8 GB |

| Memory Type | GDDR5 | GDDR6 |

| Memory Bus Width | 256 bit | 256 bit |

| Memory Bandwidth | 192.3 GB/s | 448.0 GB/s |

| Shading Units | 1536 | 2560 |

| TMUs | 128 | 160 |

| ROPs | 32 | 64 |

| Pixel Rate | 32.61 GPixel/s | 121.9 GPixel/s |

| Texture Rate | 130.4 GTexel/s | 304.8 GTexel/s |

| FP32 Performance | 3.130 TFLOPS | 9.754 TFLOPS |

| FP16 Performance | — | 19.51 TFLOPS (2:1) |

| TDP | 300 W | 225 W |

| Power Connectors | 2x 8-pin | 1x 6-pin + 1x 8-pin |

| Suggested PSU | 700 W | 550 W |

| Bus Interface | PCIe 3.0 x16 | PCIe 4.0 x16 |

| Display Outputs | 3x DVI, 1x mini-DisplayPort 1.2 | 1x HDMI 2.0b, 3x DisplayPort 1.4a |

| DirectX Support | 12 (11_0) | 12 (12_1) |

| Vulkan Support | 1.2.175 | 1.4 |

| Length | 279 mm (11 inches) | 272 mm (10.7 inches) |

| Width | 38 mm (1.5 inches) | 36 mm (1.4 inches) |

| Release Date | 2012-05-02 | 2019-07-06 |

| Launch MSRP | 999 USD | 399 USD |