AMD Radeon RX 9070 XT vs NVIDIA GeForce GTX 680 Comparison

The Verdict

The data presents a generational chasm between these two cards. The AMD Radeon RX 9070 XT is the decisive winner in every shared benchmark, with the NVIDIA GeForce GTX 680 trailing by a substantial margin. For any modern gaming or compute workload, the RX 9070 XT is the only rational choice. The GTX 680, while historically significant, is end-of-life and its benchmark scores place it firmly in the past.

If you are building a new system or upgrading from a decade-old card, the RX 9070 XT is the clear pick. Its active production status, modern API support, and vastly superior performance make it the only viable option for current software. The GTX 680 should only be considered by someone collecting vintage hardware, as its performance is not competitive with even the RX 9070 XT’s closest rivals. The data is unambiguous: the RX 9070 XT wins 2 out of 2 head-to-head benchmarks, leaving no scenario where the GTX 680 is preferable for performance.

FAQ

Q: Which card is faster in the shared benchmarks?

A: The AMD Radeon RX 9070 XT wins both head-to-head tests. It scores 17,428 in Geekbench OpenCL versus 16,906 for the GTX 680, a 3% lead. In Geekbench Vulkan, the RX 9070 XT is dramatically faster with a score of 65,879 versus 17,646, a 73.2% advantage.

Q: How do their average benchmark scores compare?

A: The GTX 680 has a higher average benchmark score of 14,150 across all its tests, compared to the RX 9070 XT’s 13,543. However, this is misleading because the two cards run different test suites. The RX 9070 XT’s average is dragged down by low Passmark scores, while the GTX 680’s tests are more limited.

Q: What are the architecture differences?

A: The GTX 680 uses the Kepler architecture on a 28 nm process, while the RX 9070 XT uses RDNA 4.0 on a 4 nm process. The RX 9070 XT has dedicated ray tracing cores (64), which the GTX 680 lacks entirely. The RX 9070 XT also supports DirectX 12 Ultimate (12_2), while the GTX 680 only supports DirectX 12 (11_0).

Q: How much memory do they have?

A: The GTX 680 has 2 GB of GDDR5 memory, while the RX 9070 XT has 16 GB of GDDR6 memory. The RX 9070 XT also has a much higher memory bandwidth: 644.6 GB/s compared to 192.3 GB/s.

Q: Which card has a higher power draw?

A: The RX 9070 XT has a higher TDP of 304 W, compared to the GTX 680’s 195 W. The RX 9070 XT also requires a 700 W suggested PSU and two 8-pin power connectors, whereas the GTX 680 needs a 450 W PSU and two 6-pin connectors.

Q: What is the release date of each card?

A: The GTX 680 was released on 2012-03-21, and the RX 9070 XT was released on 2025-03-05. The GTX 680 is end-of-life, while the RX 9070 XT is still in active production.

Architecture Differences

The two GPUs represent a fundamental shift in graphics architecture. The GTX 680 is built on the Kepler architecture, a design from 2012 that focused on efficiency and balanced compute. It uses a 28 nm process node from TSMC, with a die size of 294 mm² and 3,540 million transistors. The RX 9070 XT, by contrast, uses the modern RDNA 4.0 architecture on a 4 nm process node. Its die is larger at 357 mm², but it packs an enormous 53,900 million transistors, resulting in a transistor density of 151.0M per mm² versus the GTX 680’s 12.0M per mm².

The most significant architectural addition to the RX 9070 XT is the inclusion of 64 ray tracing cores. This hardware support for ray tracing is completely absent on the GTX 680, which relies on the traditional rasterization pipeline. The RX 9070 XT also supports a more advanced feature set, including DirectX 12 Ultimate (12_2) and Vulkan 1.4, whereas the GTX 680 tops out at DirectX 12 (11_0) and Vulkan 1.2.175. Both cards support OpenGL 4.6.

The memory architecture also differs. While both use a 256-bit bus, the RX 9070 XT uses faster GDDR6 memory running at 20.1 Gbps effective, yielding 644.6 GB/s of bandwidth. The GTX 680 uses GDDR5 at 6 Gbps effective, for 192.3 GB/s. The RX 9070 XT also has a much larger frame buffer: 16 GB versus 2 GB. This difference alone makes the RX 9070 XT suitable for modern high-resolution textures, while the GTX 680 would be severely limited.

Specification Differences

The two cards differ on nearly every measurable specification. The RX 9070 XT has significantly more processing units: 4,096 shading units, 256 TMUs, and 128 ROPs, compared to the GTX 680’s 1,536 shading units, 128 TMUs, and 32 ROPs. Clock speeds are also far apart, with the RX 9070 XT running at a 1660 MHz base and 2970 MHz boost, versus the GTX 680’s 1006 MHz base and 1058 MHz boost. The RX 9070 XT also has a game clock of 2400 MHz, a feature the GTX 680 lacks.

Compute throughput shows an enormous gap. The RX 9070 XT delivers 48.66 TFLOPS of FP32 performance, and it matches that with 48.66 TFLOPS of FP16 (1:1). The GTX 680 manages only 3.250 TFLOPS of FP32 and has no listed FP16 capability. Pixel and texture rates follow the same pattern: the RX 9070 XT achieves 380.2 GPixel/s and 760.3 GTexel/s, while the GTX 680 achieves 33.86 GPixel/s and 135.4 GTexel/s.

Power and connectivity also differ. The RX 9070 XT has a TDP of 304 W and requires a 700 W PSU with two 8-pin connectors. The GTX 680 has a 195 W TDP and needs a 450 W PSU with two 6-pin connectors. The RX 9070 XT uses PCIe 5.0 x16, while the GTX 680 uses PCIe 3.0 x16. Display outputs are more modern on the RX 9070 XT, with 1x HDMI 2.1b and 3x DisplayPort 2.1a, versus the GTX 680’s 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The RX 9070 XT is also a dual-slot card like the GTX 680, but its dimensions are not listed. The launch MSRP for the GTX 680 was 499 USD, and for the RX 9070 XT it was 599 USD.

Head-to-Head Benchmarks

The only two benchmarks that both cards ran are Geekbench OpenCL and Geekbench Vulkan. The results show a clear hierarchy. In Geekbench OpenCL, the RX 9070 XT scores 17,428, which is 3% higher than the GTX 680’s 16,906. This is a modest win, suggesting that in raw compute tasks that are not heavily optimized for modern hardware, the older card can still hold its own to some degree. However, the delta is small and does not reflect real-world gaming performance.

The Geekbench Vulkan test tells a completely different story. The RX 9070 XT crushes the GTX 680 with a score of 65,879 versus 17,646. That is a 73.2% advantage for the RX 9070 XT, and it highlights the massive generational leap in graphics API efficiency. The GTX 680’s Vulkan driver is limited to version 1.2.175, while the RX 9070 XT supports Vulkan 1.4, which likely explains part of the gap.

Looking at the broader benchmark suites, the GTX 680’s best scores are 17,646 in Vulkan and 16,906 in OpenCL, with a Geekbench Metal score of 7,897. Its average across all tests is 14,150. The RX 9070 XT’s suite includes many more tests, with its highest being 65,879 in Vulkan and 26,795 in Passmark G3D. Its average is 13,543, which is lower than the GTX 680’s average, but this is due to the inclusion of low-scoring Passmark DirectX tests (e.g., 78 in DirectX 12 and 154 in DirectX 10). These low scores are likely artifacts of the test settings or driver compatibility, not an indication of real performance.

Where Each One Wins

The RX 9070 XT wins in every meaningful category. It is the clear choice for any modern gaming workload, particularly at higher resolutions where its 16 GB of memory and 644.6 GB/s bandwidth become essential. The Vulkan benchmark advantage of 73.2% indicates that the RX 9070 XT is far better optimized for current graphics APIs, making it the superior card for DirectX 12 and Vulkan titles. Its 64 ray tracing cores also enable hardware-accelerated ray tracing, a feature that is simply unavailable on the GTX 680. For compute-heavy tasks like GPU rendering or machine learning inference, the RX 9070 XT’s 48.66 TFLOPS of FP32 performance is an order of magnitude higher than the GTX 680’s 3.250 TFLOPS.

The GTX 680 has only one area of marginal advantage: its average benchmark score of 14,150 is 4.5% higher than the RX 9070 XT’s 13,543. However, this is misleading because the test suites are different, and the GTX 680 does not face the same Passmark DirectX tests that drag down the RX 9070 XT’s average. If you look at the shared Geekbench tests, the RX 9070 XT leads in both.

For a PC builder today, the choice is straightforward. The RX 9070 XT is an active product with modern features, far superior performance, and a higher power envelope. The GTX 680 is a historical artifact. Its only potential use case is for retro gaming or testing legacy software that requires a 2012-era GPU. Even then, its 2 GB of memory and lack of modern API support will limit it. The data shows no scenario where the GTX 680 is the better purchase for a new build.