NVIDIA GeForce GTX 680 vs NVIDIA GeForce RTX 5090 D V2 Comparison

The Verdict

The RTX 5090 D V2 is a modern flagship built for extreme workloads and future-proofing. The GTX 680 is a legacy card from a different era, now best suited for basic display output or retro compatibility. The recorded data shows the RTX 5090 D V2 delivers 104.8 TFLOPS of FP32 compute versus 3.250 TFLOPS for the GTX 680, a 32-fold advantage in raw shader throughput. In the database’s average benchmark score, the RTX 5090 D V2 sits at 16,504, while the GTX 680 averages 14,150. However, the GTX 680’s average is pulled from multiple API tests (Metal, OpenCL, Vulkan), while the RTX 5090 D V2 has a single DirectX 12 Steel Nomad score. The percentile ranks are close: the RTX 5090 D V2 is in the 59th percentile of all GPUs, the GTX 680 in the 55th. That narrow gap reflects the benchmark pool, not real-world capability.

Pick the RTX 5090 D V2 if you need current API support (DirectX 12 Ultimate, Vulkan 1.4), 24 GB of GDDR7 memory, and PCIe 5.0 connectivity. Pick the GTX 680 only if you have a legacy system that requires its 2x DVI outputs or if you are collecting vintage hardware. For any modern gaming, rendering, or compute task, the RTX 5090 D V2 is the only functional choice; the GTX 680 lacks ray tracing cores, tensor cores, and modern feature sets entirely.

Where Each One Wins

The RTX 5090 D V2 wins every meaningful category in the data. It has 21,760 shading units versus 1,536, 680 texture mapping units versus 128, and 176 ROPs versus 32. Its pixel rate is 423.6 GPixel/s against 33.86 GPixel/s, and its texture rate is 1,636.8 GTexel/s against 135.4 GTexel/s. Memory bandwidth is 1.34 TB/s versus 192.3 GB/s. The GTX 680 has no ray tracing cores, no tensor cores, and no FP16 path, while the RTX 5090 D V2 carries 170 RT cores, 680 tensor cores, and FP16 at 104.8 TFLOPS (1:1 with FP32). The GTX 680’s only advantage is its lower power draw: 195 W TDP versus 575 W, and a smaller 450 W suggested PSU versus 950 W. It also uses two 6-pin power connectors instead of a single 16-pin, which may be easier to accommodate in older systems.

In benchmark scores, the RTX 5090 D V2’s Steel Nomad result of 16,504 falls just 4 points short of the NVIDIA T400’s 16,508 (0% delta), and it is 0.5% ahead of the AMD Radeon PRO W7500 (16,415), 0.6% ahead of the NVIDIA RTX PRO 6000 Blackwell (16,408), and 0.9% ahead of the AMD Radeon RX 5700 XT (16,361). The GTX 680’s nearest rivals show a different picture: the NVIDIA Tesla K10 scores 14,029 (0.9% below), the GTX 1070 Ti scores 14,277 (0.9% above), the Intel Iris Xe MAX Graphics scores 14,315 (1.2% above), and the AMD Radeon Vega 11 scores 14,352 (1.4% above). So the GTX 680 is bracketed by a Tesla compute card and a mainstream GTX 10-series card, while the RTX 5090 D V2 is bracketed by workstation and prosumer cards.

Architecture Differences

The RTX 5090 D V2 uses the GB202 chip on TSMC’s 5 nm process, featuring 92,200 million transistors on a 750 mm² die, giving a transistor density of 122.9 million per mm². The GTX 680 uses the GK104 chip on TSMC’s 28 nm process, with 3,540 million transistors on a 294 mm² die, density of 12.0 million per mm². The RTX 5090 D V2 is Blackwell 2.0 architecture, part of the GeForce 50 generation; the GTX 680 is Kepler architecture from the GeForce 600 generation. The process node difference is enormous: 5 nm versus 28 nm, which explains the 26-fold transistor count increase.

Memory architecture differs completely. The RTX 5090 D V2 has 24 GB of GDDR7 on a 384-bit bus, running at 1750 MHz (28 Gbps effective), yielding 1.34 TB/s bandwidth. The GTX 680 has 2 GB of GDDR5 on a 256-bit bus, at 1502 MHz (6 Gbps effective), yielding 192.3 GB/s. The RTX 5090 D V2 also has four display outputs (1x HDMI 2.1b, 3x DisplayPort 2.1b) versus the GTX 680’s four outputs (2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2). API support is modern on the RTX 5090 D V2: DirectX 12 Ultimate (12_2), OpenGL 4.6, Vulkan 1.4. The GTX 680 only supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175.

The RTX 5090 D V2 has a base clock of 2017 MHz and boost of 2407 MHz; the GTX 680 runs at 1006 MHz base and 1058 MHz boost. That clock advantage, combined with far more execution units, produces the massive throughput gap. The GTX 680 has no tensor cores, which means no DLSS or AI-accelerated features, while the RTX 5090 D V2’s 680 tensor cores enable modern workloads.

FAQ

Q: Which card has higher raw compute performance?

A: The RTX 5090 D V2 delivers 104.8 TFLOPS FP32, while the GTX 680 delivers 3.250 TFLOPS FP32. That is roughly 32 times higher.

Q: Can the GTX 680 run modern DirectX 12 Ultimate games?

A: No. The GTX 680 only supports DirectX 12 (11_0), while the RTX 5090 D V2 supports DirectX 12 Ultimate (12_2). The GTX 680 also misses Vulkan 1.4 (it has 1.2.175) and has no ray tracing cores.

Q: How does memory capacity affect real-world use?

A: The RTX 5090 D V2 has 24 GB of GDDR7 on a 384-bit bus with 1.34 TB/s bandwidth. The GTX 680 has 2 GB of GDDR5 on a 256-bit bus with 192.3 GB/s. For modern textures and multi-tasking, 2 GB is severely limiting; the RTX 5090 D V2 can hold large datasets or high-resolution assets.

Q: What is the power requirement difference?

A: The RTX 5090 D V2 has a 575 W TDP and suggests a 950 W PSU, using a single 16-pin connector. The GTX 680 has a 195 W TDP, suggests a 450 W PSU, and uses two 6-pin connectors. The RTX 5090 D V2 needs substantially more power delivery.

Q: Which card has better benchmark positioning relative to its peers?

A: The RTX 5090 D V2 scores 16,504 in Steel Nomad, sitting within 0.9% of its nearest rivals (T400, PRO W7500, RTX PRO 6000 Blackwell, RX 5700 XT). The GTX 680 averages 14,150 across Metal, OpenCL, and Vulkan tests, sitting within 1.4% of its nearest rivals (Tesla K10, GTX 1070 Ti, Iris Xe MAX, Vega 11). The RTX 5090 D V2’s percentile is 59 versus 55 for the GTX 680.

Q: Does the GTX 680 support modern display interfaces?

A: No. It has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The RTX 5090 D V2 has 1x HDMI 2.1b and 3x DisplayPort 2.1b, which support higher resolutions and refresh rates.

Head-to-Head Benchmarks

There are no direct head-to-head benchmark results in the database, so the comparison relies on each card’s individual recorded scores. The RTX 5090 D V2 has a single benchmark: 3DMark Steel Nomad (DirectX 12) at 16,504. The GTX 680 has three benchmarks: Geekbench Metal at 7,897, Geekbench OpenCL at 16,906, and Geekbench Vulkan at 17,646. Its average is 14,150.

Interpreting these requires care. The GTX 680’s Vulkan and OpenCL scores are actually higher than the RTX 5090 D V2’s Steel Nomad score, but that is because Steel Nomad is a demanding modern DirectX 12 workload, while Geekbench compute tests are synthetic and less representative of gaming. The RTX 5090 D V2’s FP32 rate of 104.8 TFLOPS dwarfs the GTX 680’s 3.250 TFLOPS, so in any compute-heavy task the RTX 5090 D V2 is overwhelmingly faster. The GTX 680’s higher Geekbench scores likely reflect the lower complexity of those older API tests, not actual capability.

The nearest rival data clarifies positioning. The RTX 5090 D V2 at 16,504 is essentially tied with the NVIDIA T400 (16,508, 0% delta), and slightly ahead of the AMD Radeon PRO W7500 (16,415, 0.5% delta), NVIDIA RTX PRO 6000 Blackwell (16,408, 0.6% delta), and AMD Radeon RX 5700 XT (16,361, 0.9% delta). That means its Steel Nomad score is within 1% of four very different cards, from a low-end T400 to a workstation Blackwell card. The GTX 680 at 14,150 average is 0.9% above the Tesla K10 (14,029), 0.9% below the GTX 1070 Ti (14,277), 1.2% below the Intel Iris Xe MAX (14,315), and 1.4% below the AMD Vega 11 (14,352). So the GTX 680 is sandwiched between a dual-GPU Tesla and a mainstream GTX 1070 Ti, which shows how far behind modern cards it sits.

The biggest win for the RTX 5090 D V2 is in every architectural metric: 21,760 shading units versus 1,536, 680 TMUs versus 128, 176 ROPs versus 32. The biggest win for the GTX 680 is power efficiency in the data: 195 W TDP versus 575 W, and a 450 W suggested PSU versus 950 W. The GTX 680 also uses two 6-pin connectors, which are more common in older power supplies, versus the RTX 5090 D V2’s single 16-pin, which requires a modern PSU.

Specification Differences

The two cards differ in nearly every specification. Process node: 5 nm versus 28 nm. Transistors: 92,200 million versus 3,540 million. Die size: 750 mm² versus 294 mm². Transistor density: 122.9M/mm² versus 12.0M/mm². Base clock: 2017 MHz versus 1006 MHz. Boost clock: 2407 MHz versus 1058 MHz. Memory clock: 1750 MHz (28 Gbps effective) versus 1502 MHz (6 Gbps effective). Memory size: 24 GB GDDR7 versus 2 GB GDDR5. Bus width: 384-bit versus 256-bit. Bandwidth: 1.34 TB/s versus 192.3 GB/s. Shading units: 21,760 versus 1,536. TMUs: 680 versus 128. ROPs: 176 versus 32. RT cores: 170 versus none. Tensor cores: 680 versus none. Pixel rate: 423.6 GPixel/s versus 33.86 GPixel/s. Texture rate: 1,636.8 GTexel/s versus 135.4 GTexel/s. FP32: 104.8 TFLOPS versus 3.250 TFLOPS. FP16: 104.8 TFLOPS (1:1) versus absent. TDP: 575 W versus 195 W. Power connectors: 1x 16-pin versus 2x 6-pin. Suggested PSU: 950 W versus 450 W. Bus interface: PCIe 5.0 x16 versus PCIe 3.0 x16. Display outputs: 1x HDMI 2.1b, 3x DisplayPort 2.1b versus 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2. DirectX: 12 Ultimate (12_2) versus 12 (11_0). Vulkan: 1.4 versus 1.2.175. Dimensions: 304 mm length, 137 mm height, 48 mm width versus 256 mm length, 111 mm height, 38 mm width. Release date: August 2025 versus March 2012. Production status: Active versus End-of-life. Launch MSRP: 2,299 USD versus 499 USD. The RTX 5090 D V2 is also larger in every dimension and belongs to the GeForce 50 generation with predecessor GeForce 40 and successor GeForce 60; the GTX 680 belongs to GeForce 600 with predecessor GeForce 500 and successor GeForce 700.