AMD Radeon RX 560X vs NVIDIA GeForce GTX 690 Comparison

Head-to-Head Benchmarks

The recorded data splits cleanly across the two benchmark tests, with each card claiming one decisive victory. In Geekbench OpenCL, the NVIDIA GeForce GTX 690 scores 17399 against the AMD Radeon RX 560X's 17020, a narrow 2.2% margin. That is a close result, essentially a statistical tie when considering run-to-run variance, but it does put the older Kepler card ahead in raw compute throughput under OpenCL.

The Vulkan test tells a very different story. The AMD Radeon RX 560X posts 20231, while the GTX 690 manages only 16675. That is a 21.3% advantage for the RX 560X, a substantial gap that flips the overall comparison. The RX 560X's average benchmark score across both tests is 18626, versus 17037 for the GTX 690, which works out to roughly a 9.3% lead in aggregate performance. The percentile data reinforces this: the RX 560X sits at the 62nd percentile among all GPUs, while the GTX 690 sits at the 60th.

Looking at the nearest rivals in the database provides context for these numbers. The RX 560X's average score of 18626 places it within 0.3% of the AMD Radeon Pro 5700 XT (18685) and 0.5% ahead of the AMD FirePro D500 (18533). It trails the NVIDIA GeForce RTX 2070 (18789) by 0.9% and the NVIDIA Tesla K80 (18866) by 1.3%. The GTX 690's 17037 average is nearly identical to the AMD Radeon HD 7970M (17019, a 0.1% delta), sits 0.3% below the AMD Radeon RX 7600 XT (17083), 0.6% above the NVIDIA Tesla M4 (16932), and 1% behind the NVIDIA GeForce RTX 3070 (17208). These deltas show both cards are firmly mid-pack in the database, but the RX 560X occupies a slightly higher tier.

The Vulkan result deserves extra scrutiny. A 21.3% win is not a marginal edge; it suggests architectural factors at play. The RX 560X's GCN 4.0 design with explicit Vulkan 1.3 support appears to translate workloads more efficiently through that API than the GTX 690's Kepler architecture, which lists Vulkan 1.2.175. Meanwhile, the OpenCL result, despite being close, favors the card with higher raw FP32 throughput: the GTX 690's 3.130 TFLOPS versus the RX 560X's 2.611 TFLOPS. That 0.519 TFLOPS difference (about 19.9% more compute) barely shows up in the actual OpenCL score, indicating the RX 560X extracts more useful work per FLOP.

Where Each One Wins

The AMD Radeon RX 560X wins in modern API workloads. Its Vulkan score of 20231 is not just higher, it is dramatically higher, suggesting better driver optimization and hardware support for current graphics APIs. The card also carries the higher percentile ranking (62nd versus 60th) and the higher average benchmark score (18626 versus 17037). For any workload that leverages Vulkan, the data points unambiguously to the RX 560X.

The NVIDIA GeForce GTX 690 wins in legacy compute scenarios. Its OpenCL score of 17399 edges out the RX 560X, and its hardware specs support that result: 1536 shading units, 128 texture mapping units, 32 ROPs, and a 256-bit memory bus delivering 192.3 GB/s of bandwidth. The RX 560X counters with 1024 shading units, 64 TMUs, 16 ROPs, and a 128-bit bus at 112.0 GB/s. The GTX 690 also has a higher pixel rate (32.61 GPixel/s versus 20.40 GPixel/s) and texture rate (130.4 GTexel/s versus 81.60 GTexel/s). These are all classic rasterization and fill-rate metrics, so the GTX 690 would likely hold up better in older DirectX 11 titles, consistent with its DirectX 12 (11_0) API rating.

The dual-GPU nature of the GTX 690 deserves mention. It is a 300 W card with two 8-pin power connectors and a suggested 700 W power supply, versus the RX 560X's 75 W TDP, no power connectors, and 250 W suggested PSU. That power draw explains the performance in raw compute: two GK104 dies working in parallel. But the Vulkan deficit suggests scaling inefficiencies or driver limitations in newer APIs, where the single-die RX 560X with better API support overtakes it.

FAQ

Q: Which card has the higher average benchmark score?

A: The AMD Radeon RX 560X averages 18626 across its benchmark tests, compared to 17037 for the NVIDIA GeForce GTX 690, a difference of roughly 1589 points.

Q: How large is the Vulkan performance gap?

A: The RX 560X scores 20231 in Geekbench Vulkan, while the GTX 690 scores 16675. The RX 560X leads by 21.3%, a substantial margin.

Q: Does the GTX 690 win any benchmark test?

A: Yes, the GTX 690 wins the Geekbench OpenCL test with 17399 versus 17020 for the RX 560X, a 2.2% margin.

Q: What are the percentile rankings for each card?

A: The RX 560X ranks at the 62nd percentile among all GPUs, while the GTX 690 ranks at the 60th percentile.

Q: How does the RX 560X compare to the RTX 2070 in the database?

A: The RX 560X's average score of 18626 is 0.9% below the NVIDIA GeForce RTX 2070's 18789, placing them very close in overall performance.

Q: How does the GTX 690 compare to the RX 7600 XT?

A: The GTX 690's average score of 17037 is 0.3% below the AMD Radeon RX 7600 XT's 17083, a negligible difference.

Specification Differences

The two cards differ across nearly every major specification category. The RX 560X uses a 14 nm process from GlobalFoundries, while the GTX 690 uses TSMC's 28 nm node. Transistor counts are close: 3,000 million for the RX 560X versus 3,540 million for the GTX 690, but die size diverges sharply, 123 mm² versus 294 mm². This yields a transistor density of 24.4 million per mm² for the AMD chip, more than double the GTX 690's 12.0 million per mm².

Clock speeds favor the RX 560X: a base of 1175 MHz and boost of 1275 MHz, versus 915 MHz base and 1019 MHz boost for the GTX 690. Memory configurations differ in capacity and bus width: the RX 560X has 4 GB GDDR5 on a 128-bit bus delivering 112.0 GB/s, while the GTX 690 has 2 GB GDDR5 on a 256-bit bus delivering 192.3 GB/s. Effective memory speed is 7 Gbps for the RX 560X and 6 Gbps for the GTX 690.

Compute resources favor the GTX 690 in raw counts: 1536 shading units, 128 TMUs, and 32 ROPs, versus 1024 shading units, 64 TMUs, and 16 ROPs for the RX 560X. Pixel rate is 32.61 GPixel/s for the GTX 690 and 20.40 GPixel/s for the RX 560X; texture rate is 130.4 GTexel/s versus 81.60 GTexel/s. FP32 compute is 3.130 TFLOPS for the GTX 690 and 2.611 TFLOPS for the RX 560X. The RX 560X lists FP16 at 2.611 TFLOPS (1:1), while the GTX 690 has no FP16 entry.

Power and physical dimensions also differ. The RX 560X has a 75 W TDP, no power connectors, a 250 W suggested PSU, and 170 mm length (6.7 inches). The GTX 690 has a 300 W TDP, two 8-pin power connectors, a 700 W suggested PSU, and measures 279 mm (11 inches) long, 111 mm (4.4 inches) high, and 38 mm (1.5 inches) wide. Bus interface is PCIe 3.0 x8 for the RX 560X and PCIe 3.0 x16 for the GTX 690. Display outputs are 1x DVI, 1x HDMI 2.0b, 1x DisplayPort 1.4a for the RX 560X; the GTX 690 uses 3x DVI and 1x mini-DisplayPort 1.2.

Architecture Differences

The RX 560X is built on GCN 4.0 architecture using the Polaris 21 chip, part of the Polaris (RX 500X) generation. The GTX 690 uses the Kepler architecture with the GK104 chip, part of the GeForce 600 generation. The RX 560X supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.3. The GTX 690 lists DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. That DirectX level difference is notable: the RX 560X meets the full DirectX 12 feature set, while the GTX 690 only reaches the 11_0 feature level, which limits its modern API capabilities.

The GTX 690 is a dual-GPU card, which the specification data implies through its 300 W TDP, dual 8-pin connectors, and dual-slot width. The RX 560X is a single-GPU card with a 75 W TDP and no power connectors. The Kepler architecture's dual-die design likely explains the GTX 690's high fill rates and compute counts, but also its Vulkan deficit, as multi-GPU scaling often suffers in newer APIs.

The RX 560X's smaller die and higher transistor density indicate a more modern design, even though it has fewer compute units. Its memory capacity is double that of the GTX 690 (4 GB versus 2 GB), which matters for texture-heavy workloads and modern game assets. The GTX 690's wider memory bus and higher bandwidth, 192.3 GB/s versus 112.0 GB/s, partially compensates for the smaller capacity.

The Verdict

The data favors the AMD Radeon RX 560X for modern workloads. Its 21.3% Vulkan lead, higher average score (18626 versus 17037), higher percentile (62 versus 60), and full DirectX 12 (12_0) support make it the more future-proof choice. It also consumes far less power, 75 W versus 300 W, and fits in a smaller physical footprint, 170 mm versus 279 mm length. For anyone running Vulkan-based games or applications, the RX 560X is the clear pick.

The NVIDIA GeForce GTX 690 retains relevance in legacy compute. Its OpenCL advantage, albeit narrow at 2.2%, combined with higher raw fill rates (32.61 GPixel/s and 130.4 GTexel/s) and higher FP32 throughput (3.130 TFLOPS), makes it suited for older DirectX 11 workloads and compute tasks that scale well across dual dies. The 2 GB memory capacity is a limitation, but the 256-bit bus helps mitigate that in bandwidth-sensitive scenarios.

The database shows a 1-1 split in benchmark wins, but the magnitude of those wins is asymmetric. The RX 560X's Vulkan victory is nearly ten times larger, in percentage terms, than the GTX 690's OpenCL edge. The average scores confirm this: the RX 560X sits closer to an RTX 2070 (0.9% behind) than the GTX 690 sits to an RX 7600 XT (0.3% behind), but the GTX 690's rival set is lower-performing overall. The GTX 690's nearest rivals average between 16932 and 17208, while the RX 560X's nearest rivals average between 18533 and 18866. That is the real story: the RX 560X competes in a higher performance tier despite its lower power envelope.

For a buyer choosing between these two end-of-life cards, the RX 560X offers better aggregate performance, lower power requirements, and superior modern API support. The GTX 690 is only preferable for specific legacy compute scenarios where its dual-GPU raw throughput and higher memory bandwidth can be fully utilized. The recorded data does not support a general recommendation for the GTX 690; it wins one test by a hair, while the RX 560X wins the other by a mile.