NVIDIA GeForce GTX 680 vs NVIDIA RTX PRO 6000 Blackwell Comparison

NVIDIA’s RTX PRO 6000 Blackwell and GeForce GTX 680 sit at opposite ends of the GPU timeline, separated by over a decade of architectural evolution. The recorded data shows a generational chasm: the RTX PRO 6000 Blackwell posts an average benchmark score of 16,408, while the GTX 680 averages 14,150 across its three recorded tests. The newer card’s raw compute resources dwarf the older model, yet the GTX 680 still holds its own in specific legacy workloads. This analysis examines the head-to-head results, architectural differences, and use-case implications strictly from the database.

Head-to-Head Benchmarks

The database contains no direct head-to-head benchmark entries between these two cards, so the comparison relies on their respective average benchmark scores and nearest-rival deltas. The RTX PRO 6000 Blackwell’s sole recorded benchmark, 3DMark Steel Nomad DX12, yields a score of 16,408. This places it at the 59th percentile among all GPUs in the database. Its nearest rivals include the AMD Radeon PRO W7500 at 16,415 (a 0% delta), the AMD Radeon RX 5700 XT at 16,361 (a 0.3% delta), and the AMD Radeon Pro 5600M at 16,351 (a 0.4% delta). The RTX PRO 6000 Blackwell trails the NVIDIA GeForce RTX 5090 D V2 by 0.6%, which scores 16,504. These margins are remarkably tight, indicating that the RTX PRO 6000 Blackwell’s performance in this DX12 workload is competitive with a cluster of mid-range and professional cards, rather than a runaway leader.

The GTX 680’s benchmark profile is broader, with three recorded tests: Geekbench Metal at 7,897, Geekbench OpenCL at 16,906, and Geekbench Vulkan at 17,646. Its average benchmark score is 14,150, placing it at the 55th percentile among all GPUs. The GTX 680’s nearest rivals include the NVIDIA Tesla K10 at 14,029 (a 0.9% delta), the NVIDIA GeForce GTX 1070 Ti at 14,277 (a -0.9% delta), the Intel Iris Xe MAX Graphics at 14,315 (a -1.2% delta), and the AMD Radeon Vega 11 at 14,352 (a -1.4% delta). The GTX 680 is thus sandwiched between the Tesla K10 and the GTX 1070 Ti, with performance that is within 1.4% of all four rivals. This suggests that the GTX 680, despite its age, remains a capable performer in compute-oriented workloads like OpenCL and Vulkan.

Direct comparison of the two cards’ recorded scores is complicated by the lack of overlapping tests. The RTX PRO 6000 Blackwell’s 16,408 in 3DMark Steel Nomad DX12 is a pure DirectX 12 workload, while the GTX 680’s highest score, 17,646 in Geekbench Vulkan, reflects a different API and workload type. The RTX PRO 6000 Blackwell’s FP32 compute of 126.0 TFLOPS is approximately 38.8 times higher than the GTX 680’s 3.250 TFLOPS, but benchmark results do not scale linearly with raw TFLOPS. The GTX 680’s Vulkan and OpenCL scores suggest that its Kepler architecture, while ancient, still executes certain compute kernels efficiently enough to remain within striking distance of much newer hardware.

Architecture Differences

The architectural gap between these two GPUs is profound. The RTX PRO 6000 Blackwell uses the GB202 chip built on a 5 nm process at TSMC, packing 92,200 million transistors into a 750 mm² die. The transistor density is 122.9 million transistors per mm². The GTX 680 uses the GK104 chip on a 28 nm process, also at TSMC, with 3,540 million transistors on a 294 mm² die, yielding a density of 12.0 million transistors per mm². The RTX PRO 6000 Blackwell has roughly 26 times more transistors, but its die is only 2.55 times larger, reflecting the density advantage of the newer process.

The RTX PRO 6000 Blackwell’s Blackwell 2.0 architecture includes 24,064 shading units, 752 texture mapping units, 192 ROPs, 188 RT cores, and 752 tensor cores. The GTX 680’s Kepler architecture has 1,536 shading units, 128 TMUs, and 32 ROPs, with no RT cores and no tensor cores. The RTX PRO 6000 Blackwell’s shading unit count is 15.7 times higher, its TMU count is 5.9 times higher, and its ROP count is 6 times higher. The pixel rate difference is even starker: 502.5 GPixel/s versus 33.86 GPixel/s, a 14.8-fold gap. Texture rate is 1,968.0 GTexel/s versus 135.4 GTexel/s, a 14.5-fold gap.

Memory architecture also diverges completely. The RTX PRO 6000 Blackwell has 96 GB of GDDR7 memory on a 512-bit bus, delivering 1.79 TB/s of bandwidth. The GTX 680 has 2 GB of GDDR5 memory on a 256-bit bus, with 192.3 GB/s of bandwidth. The newer card has 48 times more memory capacity and 9.3 times more bandwidth. Clock speeds differ as well: the RTX PRO 6000 Blackwell runs at a base of 1590 MHz and a boost of 2617 MHz, while the GTX 680 runs at 1006 MHz base and 1058 MHz boost. The memory clock differs from 1750 MHz (28 Gbps effective) on the newer card versus 1502 MHz (6 Gbps effective) on the older.

API support reflects the generational shift. The RTX PRO 6000 Blackwell supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The GTX 680 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The newer card’s DirectX 12 Ultimate feature set includes hardware ray tracing and mesh shaders, which the GTX 680 lacks entirely due to its absence of RT cores.

Where Each One Wins

The RTX PRO 6000 Blackwell wins decisively in any workload that leverages its modern feature set. The presence of 188 RT cores and 752 tensor cores means ray-traced rendering and AI-accelerated tasks are possible, while the GTX 680 has no such hardware. The 96 GB memory pool is suited for massive datasets, such as large language models or high-resolution 3D scenes, far beyond the GTX 680’s 2 GB limit. The 1.79 TB/s bandwidth supports high-throughput compute, and the 126.0 TFLOPS FP32 performance handles dense math operations. The 3DMark Steel Nomad DX12 score of 16,408, while not dominant over its nearest rivals, demonstrates that the card performs competitively in modern DX12 gaming and professional workloads.

The GTX 680 wins in legacy compatibility and low-power scenarios. Its Geekbench Vulkan score of 17,646 and OpenCL score of 16,906 are higher than the RTX PRO 6000 Blackwell’s single 3DMark score, though the tests are not directly comparable. The GTX 680’s 195 W TDP versus the RTX PRO 6000 Blackwell’s 600 W means it draws far less power, and its suggested PSU of 450 W versus 1000 W reflects a much lighter system requirement. For older applications built for Kepler-era hardware, the GTX 680 remains functional. Its 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2 outputs support legacy monitors, while the RTX PRO 6000 Blackwell offers 4x DisplayPort 2.1b only, which may not connect to older displays without adapters.

The GTX 680’s percentile ranking of 55 versus the RTX PRO 6000 Blackwell’s 59 shows that, in the database’s overall distribution, the newer card sits higher, but the gap is not enormous. The GTX 680’s average score of 14,150 is 13.7% lower than the RTX PRO 6000 Blackwell’s 16,408, but the older card’s three-test average includes two high compute scores that boost its standing.

Specification Differences

The two cards differ in nearly every measurable 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 per mm² versus 12.0M per mm². Base clock: 1590 MHz versus 1006 MHz. Boost clock: 2617 MHz versus 1058 MHz. Memory size: 96 GB versus 2 GB. Memory type: GDDR7 versus GDDR5. Memory bus width: 512-bit versus 256-bit. Memory bandwidth: 1.79 TB/s versus 192.3 GB/s. Shading units: 24,064 versus 1,536. TMUs: 752 versus 128. ROPs: 192 versus 32. RT cores: 188 versus null. Tensor cores: 752 versus null. Pixel rate: 502.5 GPixel/s versus 33.86 GPixel/s. Texture rate: 1,968.0 GTexel/s versus 135.4 GTexel/s. FP32: 126.0 TFLOPS versus 3.250 TFLOPS. FP16: 126.0 TFLOPS (1:1) versus null. TDP: 600 W versus 195 W. Power connectors: 1x 16-pin versus 2x 6-pin. Suggested PSU: 1000 W versus 450 W. Bus interface: PCIe 5.0 x16 versus PCIe 3.0 x16. Display outputs: 4x DisplayPort 2.1b versus 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2. DirectX support: 12 Ultimate (12_2) versus 12 (11_0). Vulkan support: 1.4 versus 1.2.175. Dimensions: 304 mm length, 137 mm height, 40 mm width versus 256 mm length, 111 mm height, 38 mm width. Release date: 2025-03-17 versus 2012-03-21. Production status: Active versus End-of-life. Predecessor: Workstation Ada versus GeForce 500. Successor: null versus GeForce 700. Launch MSRP: 8,565 USD versus 499 USD.

FAQ

Q: Which card has more memory bandwidth?

A: The RTX PRO 6000 Blackwell has 1.79 TB/s, compared to the GTX 680’s 192.3 GB/s.

Q: Does the GTX 680 support hardware ray tracing?

A: No, the GTX 680 has no RT cores, while the RTX PRO 6000 Blackwell has 188 RT cores.

Q: What is the FP32 compute difference?

A: The RTX PRO 6000 Blackwell delivers 126.0 TFLOPS, while the GTX 680 delivers 3.250 TFLOPS, a roughly 38.8-fold difference.

Q: Which card has a higher average benchmark score?

A: The RTX PRO 6000 Blackwell has an average benchmark score of 16,408, while the GTX 680 averages 14,150, a 13.7% difference.

Q: What process nodes are used?

A: The RTX PRO 6000 Blackwell uses a 5 nm process, and the GTX 680 uses a 28 nm process, both at TSMC.

Q: Which card has more shading units?

A: The RTX PRO 6000 Blackwell has 24,064 shading units, compared to the GTX 680’s 1,536.

The Verdict

The data is unambiguous. The RTX PRO 6000 Blackwell is the superior card for any modern workload that requires high compute throughput, large memory capacity, or advanced features like ray tracing and tensor operations. Its 126.0 TFLOPS FP32, 96 GB GDDR7 memory, and 1.79 TB/s bandwidth place it in a different class from the GTX 680. The GTX 680, with its 3.250 TFLOPS FP32, 2 GB GDDR5 memory, and no RT or tensor cores, is limited to legacy applications and low-power scenarios. Its 195 W TDP and 450 W suggested PSU make it an easy fit for older systems, but its 55th percentile ranking versus the newer card’s 59th shows that even in the database’s overall distribution, the RTX PRO 6000 Blackwell holds the edge.

Buyers who need modern API support, such as DirectX 12 Ultimate or Vulkan 1.4, should choose the RTX PRO 6000 Blackwell. Those running older software that relies on Kepler-era compatibility and who prioritize low power consumption might find the GTX 680 sufficient, but the performance gap is too large for any demanding application. The RTX PRO 6000 Blackwell’s launch MSRP of 8,565 USD reflects its workstation positioning, while the GTX 680’s 499 USD launch MSRP was that of an enthusiast gaming card in 2012. The verdict is clear: for any serious compute, rendering, or modern gaming task, the RTX PRO 6000 Blackwell is the only viable choice based on recorded data.