NVIDIA GeForce GTX 680 vs NVIDIA Tesla K20Xm Comparison

Head-to-Head Benchmarks

The recorded data shows a narrow but consistent advantage for the NVIDIA Tesla K20Xm across the two shared benchmark workloads. In the Geekbench Metal test, the Tesla K20Xm scores 8035 against 7897 for the GeForce GTX 680, a delta of 1.7 percent in favor of the Tesla. The OpenCL results follow the same pattern: the Tesla K20Xm posts 17215, while the GeForce GTX 680 reaches 16906, a 1.8 percent gap. Across both head-to-head comparisons, the Tesla K20Xm wins two out of two tests, though the margins are modest enough that the two cards should be considered close competitors in raw compute throughput.

The average benchmark score tells a slightly different story. The GeForce GTX 680 averages 14150 across all recorded workloads, while the Tesla K20Xm averages 12625. This places the GeForce GTX 680 at the 55th percentile of all GPUs in the database, compared to the Tesla K20Xm at the 52nd percentile. The discrepancy between the head-to-head results and the overall averages suggests that the GeForce GTX 680 benefits from a broader set of workloads in which it performs relatively better, while the Tesla K20Xm's two head-to-head wins are confined to the specific Metal and OpenCL tests recorded.

Looking at the nearest rivals for each card provides additional context. The GeForce GTX 680's closest competitor is the NVIDIA Tesla K10 with an average score of 14029, a 0.9 percent advantage for the GTX 680. The GTX 680 also sits 0.9 percent behind the GeForce GTX 1070 Ti (14277) and 1.2 percent behind the Intel Iris Xe MAX Graphics (14315), with the AMD Radeon Vega 11 (14352) trailing the GTX 680 by 1.4 percent. The Tesla K20Xm, by contrast, sits 0.7 percent behind the AMD Radeon RX 7600M XT (12710), 1.2 percent behind the NVIDIA GeForce GTX 670 (12773), 1.6 percent behind the GTX 590 (12830), and 1.6 percent behind the AMD Radeon Pro 455 (12831). These figures indicate that while the Tesla K20Xm wins the direct head-to-head against the GTX 680, its overall standing among all GPUs is lower, reflecting a narrower set of strengths.

Where Each One Wins

The GeForce GTX 680 demonstrates its advantage in scenarios that rely on the full breadth of recorded benchmark data. Its average benchmark score of 14150 exceeds the Tesla K20Xm's 12625 by a substantial margin, and its 55th percentile placement versus the Tesla's 52nd percentile indicates that the GTX 680 is more competitive across a wider range of workloads. For users whose primary concern is general-purpose compute performance as measured by the database's aggregate scores, the GeForce GTX 680 is the stronger choice despite losing the two direct comparisons.

The Tesla K20Xm wins specifically in the Metal and OpenCL workloads recorded in the head-to-head benchmarks. Its 8035 Metal score and 17215 OpenCL score both surpass the GTX 680's corresponding results, making it the preferred card for applications that rely heavily on these APIs. The margin is small, but it is consistent across both tests. The Tesla K20Xm also brings a much larger memory configuration, with 6 GB of GDDR5 on a 384-bit bus delivering 249.6 GB/s of bandwidth, compared to the GTX 680's 2 GB on a 256-bit bus at 192.3 GB/s. This memory advantage could prove decisive in workloads where capacity or bandwidth is the limiting factor, even if the raw compute scores are close.

The architecture data reinforces this split. The Tesla K20Xm has 2688 shading units, 224 texture mapping units, and 48 render output units, all of which exceed the GTX 680's 1536 shading units, 128 TMUs, and 32 ROPs. The Tesla K20Xm also reaches higher theoretical peak rates: 40.99 GPixel/s and 164.0 GTexel/s, versus 33.86 GPixel/s and 135.4 GTexel/s for the GTX 680. Its FP32 throughput is 3.935 TFLOPS against 3.250 TFLOPS. These figures suggest the Tesla K20Xm should be faster in compute-heavy tasks, and the head-to-head results confirm this, but the GTX 680's higher average score indicates that the database's full set of workloads includes scenarios where the GTX 680's architecture, clock speeds, or driver optimizations produce better outcomes.

Architecture Differences

Both cards are built on NVIDIA's Kepler architecture and manufactured at TSMC's 28 nm process, but they diverge significantly in die size and complexity. The GeForce GTX 680 uses the GK104 chip with 3,540 million transistors on a 294 mm² die, while the Tesla K20Xm uses the GK110 chip with 7,080 million transistors on a 561 mm² die. Transistor density is similar, at 12.0 million per mm² for the GTX 680 and 12.6 million per mm² for the Tesla K20Xm, but the Tesla packs roughly twice the transistor count into roughly twice the die area.

Clock speeds differ notably. The GTX 680 has a base clock of 1006 MHz and a boost clock of 1058 MHz, while the Tesla K20Xm has no recorded base or boost clock values in the database. Memory clocks also differ: the GTX 680 runs at 1502 MHz with an effective data rate of 6 Gbps, while the Tesla K20Xm runs at 1300 MHz with an effective rate of 5.2 Gbps. Despite the lower memory clock, the Tesla K20Xm achieves higher bandwidth (249.6 GB/s versus 192.3 GB/s) because of its wider 384-bit memory bus versus the GTX 680's 256-bit bus.

The memory capacity difference is substantial: 2 GB on the GTX 680 versus 6 GB on the Tesla K20Xm. This has direct implications for workloads that require large datasets or high-resolution textures. The Tesla K20Xm also has no display outputs, whereas the GTX 680 provides 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. This positions the GTX 680 as a consumer graphics card with multi-monitor support, while the Tesla K20Xm is a compute-focused accelerator with no video output capability.

Power requirements scale with the larger chip. The Tesla K20Xm has a TDP of 235 W and a suggested PSU of 550 W, while the GTX 680 draws 195 W and requires a 450 W PSU. The GTX 680 uses 2x 6-pin power connectors, while the Tesla K20Xm has no recorded power connector information. Both cards are dual-slot designs, with the GTX 680 measuring 256 mm in length and the Tesla K20Xm measuring 267 mm.

The API support is identical across both cards: DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Neither card has ray tracing or tensor cores. The GTX 680 belongs to the GeForce 600 generation with a release date of 2012-03-21, while the Tesla K20Xm is part of the Tesla Kepler (Kxx) generation with a release date of 2012-11-11. The GTX 680's predecessor is the GeForce 500 series and its successor is the GeForce 700 series, while the Tesla K20Xm follows Tesla Fermi and precedes Tesla Maxwell.

FAQ

Q: Which card has a higher average benchmark score?

A: The NVIDIA GeForce GTX 680 has an average benchmark score of 14150, compared to 12625 for the NVIDIA Tesla K20Xm.

Q: Does the Tesla K20Xm win any benchmark tests against the GTX 680?

A: Yes, the Tesla K20Xm wins both recorded head-to-head tests: Geekbench Metal (8035 versus 7897) and Geekbench OpenCL (17215 versus 16906).

Q: How much memory bandwidth does each card provide?

A: The Tesla K20Xm provides 249.6 GB/s through a 384-bit bus, while the GTX 680 provides 192.3 GB/s through a 256-bit bus.

Q: Which GPU has more shading units?

A: The Tesla K20Xm has 2688 shading units, while the GTX 680 has 1536 shading units.

Q: What are the transistor counts for each card?

A: The GTX 680 has 3,540 million transistors on a 294 mm² die, while the Tesla K20Xm has 7,080 million transistors on a 561 mm² die.

Q: Do both cards support the same graphics APIs?

A: Yes, both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175.

The Verdict

The data presents a clear split between these two Kepler-based NVIDIA cards. The Tesla K20Xm is the superior compute accelerator in direct comparisons: it wins both head-to-head tests, offers more than four times the memory capacity, delivers higher memory bandwidth, and has a larger pool of shading units, TMUs, and ROPs. Its FP32 throughput of 3.935 TFLOPS and pixel rate of 40.99 GPixel/s exceed the GTX 680's 3.250 TFLOPS and 33.86 GPixel/s. For workloads that are compute-bound and memory-hungry, the Tesla K20Xm is the better choice.

However, the GeForce GTX 680 is the stronger all-round performer according to the database's aggregate metrics. Its average score of 14150 places it 12 percent ahead of the Tesla K20Xm's 12625, and its 55th percentile ranking versus the Tesla's 52nd percentile indicates broader competitiveness across the full set of recorded workloads. The GTX 680 also offers display outputs, which the Tesla K20Xm lacks entirely, making the GTX 680 the only option for users who need video output from the card itself.

The choice ultimately depends on the workload profile. Users who prioritize compute density, memory capacity, and peak throughput should select the Tesla K20Xm, provided they can accommodate its higher TDP of 235 W and lack of display outputs. Users who need a general-purpose GPU with display support, lower power draw of 195 W, and better aggregate benchmark performance should choose the GeForce GTX 680. The launch MSRP for the GTX 680 was 499 USD, while the Tesla K20Xm launched at 7,699 USD, reflecting their different market positions.