AMD Radeon RX 5700 XT vs NVIDIA GeForce GTX 680 Comparison

Where Each One Wins

The recorded benchmark data splits cleanly by API and workload type. The AMD Radeon RX 5700 XT wins 2 out of 3 head-to-head tests, taking Geekbench Metal and Geekbench Vulkan decisively. The NVIDIA GeForce GTX 680 claims only Geekbench OpenCL, but the margin there is substantial.

For modern graphics workloads, the RX 5700 XT is the clear pick. Its Metal score of 74647 versus 7897 for the GTX 680 represents a 845.3% advantage in Geekbench Metal, a gap so large that no driver tweak or clock adjustment could close it. Vulkan tells a similar story: 68310 for AMD against 17646 for NVIDIA, a 287.1% lead. If the target application stack uses Metal or Vulkan, the data says there is no contest in practice.

The GTX 680 wins in compute-oriented OpenCL testing. Its Geekbench OpenCL score of 16906 beats the RX 5700 XT's 9524, a 43.7% deficit for the newer card. The pattern suggests the GTX 680's Kepler architecture retains a relative strength in OpenCL compute tasks, but this is a single win against two much larger losses.

The broader database picture reinforces this split. The RX 5700 XT sits at the 59th percentile among all GPUs, while the GTX 680 sits at the 55th. Their average benchmark scores are 16361 and 14150 respectively. The RX 5700 XT lands between the AMD Radeon Pro 5600M (16351, 0.1% delta) and the NVIDIA RTX PRO 6000 Blackwell class of cards, while the GTX 680 sits just above the NVIDIA Tesla K10 (14029, 0.9% delta) in average score.

Architecture Differences

The underlying silicon could not be more different in era. The RX 5700 XT uses the Navi 10 chip built on TSMC's 7 nm process with 10,300 million transistors on a 251 mm² die. The GTX 680 uses GK104 on TSMC's 28 nm process with 3,540 million transistors on a 294 mm² die. The density difference is stark: 41.0M transistors per mm² for Navi 10 versus 12.0M per mm² for GK104. The RX 5700 XT crams nearly three and a half times as many transistors per area into a smaller chip, and this density shift is the root of most performance differences.

The RX 5700 XT belongs to the RDNA 1.0 architecture, part of the Radeon RX 5000 series generation. The GTX 680 runs Kepler, part of the GeForce 600 generation. Both are end-of-life products for their respective manufacturers, and both use TSMC as foundry, so the comparison isolates architecture and process rather than fab differences.

Shader resources differ broadly. The RX 5700 XT carries 2560 shading units, 160 texture mapping units, and 64 ROPs. The GTX 680 has 1536 shading units, 128 TMUs, and 32 ROPs. The AMD part has 67% more shading units, 25% more TMUs, and double the ROP count. Neither card has ray tracing cores or tensor cores, so both rely on traditional rasterization and compute paths.

Clock behavior also differs. The RX 5700 XT lists a 1605 MHz base, 1755 MHz game clock, and 1905 MHz boost. The GTX 680 lists a 1006 MHz base and 1058 MHz boost, with no game clock field. The AMD part runs at a substantially higher frequency, which compounds its shader count advantage. The GTX 680's boost clock of 1058 MHz is lower than the RX 5700 XT's base clock of 1605 MHz.

Head-to-Head Benchmarks

Three benchmark comparisons exist in the database, and each tells a distinct story about API-level strengths.

Geekbench Metal is the largest single victory for the RX 5700 XT. The score of 74647 versus 7897 gives AMD a 845.3% lead. This is not a marginal edge; the AMD card is roughly nine and a half times faster in this test. For any Metal-based workload, the GTX 680 is effectively outclassed. The nearest rivals for the RX 5700 XT, the AMD Radeon Pro 5600M and NVIDIA RTX PRO 6000 Blackwell, sit within 0.3% of its average score, so its Metal result is consistent with its peer group placement.

Geekbench Vulkan shows a similar shape. The RX 5700 XT scores 68310 against 17646 for the GTX 680, a 287.1% lead. While not as extreme as the Metal gap, this still represents a near-tripling of performance. Vulkan is the modern cross-platform graphics API, and the RX 5700 XT handles it with much more headroom. The GTX 680's Vulkan support stops at version 1.2.175, while the RX 5700 XT supports Vulkan 1.4, which likely contributes to the gap.

Geekbench OpenCL reverses the trend. The GTX 680 scores 16906, beating the RX 5700 XT's 9524 by 43.7%. This is the only test the NVIDIA card wins, and it wins by a meaningful margin. The RX 5700 XT actually loses ground to the older card in this compute API. The GTX 680's Kepler architecture, despite its age and smaller transistor count, posts a higher OpenCL result. This is the single counterpoint to the RX 5700 XT's dominance elsewhere.

It is worth remembering the RX 5700 XT has a much larger set of recorded benchmarks than the GTX 680. The AMD card appears in 3DMark Steel Nomad DX12, Passmark DirectX 9/10/11/12, Passmark G2D, G3D, and GPU compute tests, plus the three Geekbench tests. The GTX 680 only has the three Geekbench entries. The average benchmark score of 16361 for the RX 5700 XT versus 14150 for the GTX 680 reflects that broader test coverage.

Specification Differences

The two cards diverge on nearly every measurable specification. Memory capacity: the RX 5700 XT has 8 GB of GDDR6, the GTX 680 has 2 GB of GDDR5. Memory bandwidth: 448.0 GB/s versus 192.3 GB/s, a 2.3x advantage for AMD. Both use a 256-bit bus, so the bandwidth gap comes entirely from memory type and clock speed. The RX 5700 XT runs memory at 1750 MHz (14 Gbps effective), while the GTX 680 runs at 1502 MHz (6 Gbps effective).

Compute rates follow the same pattern. The RX 5700 XT delivers 9.754 TFLOPS FP32, versus 3.250 TFLOPS for the GTX 680. Pixel rate: 121.9 GPixel/s versus 33.86 GPixel/s. Texture rate: 304.8 GTexel/s versus 135.4 GTexel/s. The RX 5700 XT also supports FP16 at 19.51 TFLOPS with a 2:1 ratio, while the GTX 680 has no recorded FP16 capability.

Power and connectivity differ. The RX 5700 XT has a 225 W TDP and requires a 550 W suggested PSU, with one 6-pin and one 8-pin connector. The GTX 680 has a 195 W TDP, a 450 W suggested PSU, and two 6-pin connectors. Both are dual-slot cards, but the RX 5700 XT is longer at 272 mm versus 256 mm, and slightly narrower at 36 mm width versus 38 mm. Both are 111 mm tall.

Interface and outputs differ. The RX 5700 XT uses PCIe 4.0 x16, while the GTX 680 uses PCIe 3.0 x16. Display outputs: the RX 5700 XT has 1x HDMI 2.0b and 3x DisplayPort 1.4a. The GTX 680 has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The AMD card drops DVI entirely and offers more modern display connectors.

API support differs in two important ways. DirectX: the RX 5700 XT supports DirectX 12 (12_1), while the GTX 680 supports DirectX 12 (11_0), a feature level difference. OpenGL is 4.6 for both. Vulkan: the RX 5700 XT supports 1.4, the GTX 680 supports 1.2.175. The RX 5700 XT launched in 2019, the GTX 680 in 2012. The Radeon's predecessor is Vega and successor is Navi II; the GeForce's predecessor is GeForce 500 and successor is GeForce 700.

FAQ

Q: Which card is faster in modern gaming workloads?

A: The RX 5700 XT wins both Geekbench Metal (74647 versus 7897, a 845.3% lead) and Geekbench Vulkan (68310 versus 17646, a 287.1% lead). These are the APIs used by current games, so the data favors AMD heavily.

Q: Does the GTX 680 win any benchmark?

A: Yes, it wins Geekbench OpenCL with a score of 16906 versus 9524 for the RX 5700 XT, a 43.7% margin. This is the only head-to-head test the NVIDIA card wins.

Q: How do their average benchmark scores compare?

A: The RX 5700 XT has an average benchmark score of 16361, while the GTX 680 has 14150. The RX 5700 XT sits at the 59th percentile among all GPUs, the GTX 680 at the 55th.

Q: What are the memory differences?

A: The RX 5700 XT has 8 GB of GDDR6 with 448.0 GB/s bandwidth. The GTX 680 has 2 GB of GDDR5 with 192.3 GB/s bandwidth. Both use a 256-bit bus.

Q: Which card has better API support?

A: The RX 5700 XT supports DirectX 12 (12_1), Vulkan 1.4, and OpenGL 4.6. The GTX 680 supports DirectX 12 (11_0), Vulkan 1.2.175, and OpenGL 4.6. The AMD card has the higher DirectX feature level and newer Vulkan version.

Q: Are both cards still in production?

A: No, both are end-of-life products. The RX 5700 XT was released in 2019, the GTX 680 in 2012.

The Verdict

The data supports a clear choice for most use cases. The RX 5700 XT dominates in Metal and Vulkan, holds a massive lead in memory capacity and bandwidth, offers more than triple the FP32 compute throughput, and carries a much higher average benchmark score. Its 59th percentile placement versus 55th for the GTX 680 confirms the hierarchy. The 8 GB memory buffer alone makes it the practical choice for any modern workload, since 2 GB is insufficient for current game textures and compute buffers.

The GTX 680 has one genuine strength: OpenCL compute. Its 43.7% lead in that test suggests Kepler still has some utility in OpenCL-specific applications. But that single win does not offset losses of 845.3% and 287.1% in the other two tests. The GTX 680 also requires less power (195 W versus 225 W TDP) and a smaller PSU (450 W versus 550 W), which matters in legacy system builds.

Choose the RX 5700 XT if the workload uses Metal, Vulkan, DirectX 12, or any modern graphics API. Choose the GTX 680 only if OpenCL compute is the sole requirement and the power envelope is a hard constraint. For everything else, the RX 5700 XT is the recorded data's unambiguous winner. The launch MSRP for the RX 5700 XT was 399 USD, and 499 USD for the GTX 680 at launch.