NVIDIA GeForce GTX 1660 SUPER vs NVIDIA Tesla K20c Comparison

The Verdict

The data in this comparison is unusually one-sided. The NVIDIA GeForce GTX 1660 SUPER holds a massive advantage in the only benchmark both cards share, and it does so while drawing less power, using a modern architecture, and offering standard display outputs. The Tesla K20c, however, is a compute-oriented card from a different era, and its strengths lie in raw shading unit count and memory bus width, not in directly measured gaming or general GPU performance.

For a builder choosing between these two today, the GTX 1660 SUPER is the practical pick. It wins the head-to-head OpenCL benchmark by 357.3%, which is not a marginal gap but a generational chasm. The card is also end-of-life, but so is the Tesla K20c, so neither offers future platform longevity. The GTX 1660 SUPER is the one you can actually plug into a monitor, as it has DVI, HDMI 2.0, and DisplayPort 1.4a outputs; the Tesla K20c has no display outputs at all, making it useless as a primary graphics card.

The Tesla K20c is only interesting if you are building a compute-only workstation and need many shading units per card. It has 2496 shading units versus 1408 on the GTX 1660 SUPER, and 208 texture mapping units versus 88, which suggests it could handle certain parallel workloads efficiently. But the recorded OpenCL score does not support that theory in practice, as the GTX 1660 SUPER scores 52,490 versus 11,479. The Tesla K20c also requires a 550 W suggested power supply versus 300 W for the GTX 1660 SUPER, and it uses two power connectors (1x 6-pin plus 1x 8-pin) versus a single 8-pin. That is a meaningful installation constraint.

The verdict is straightforward: unless you have a specific compute workload that historically favors Kepler-based Tesla cards, and you can tolerate no display output and higher power draw, the GTX 1660 SUPER is the superior choice. The benchmark data does not leave room for interpretation.

Architecture Differences

The two cards come from different architectural generations. The GTX 1660 SUPER uses the TU116 chip based on the Turing architecture, built on a 12 nm process at TSMC. It packs 6,600 million transistors into a 284 mm² die, giving a transistor density of 23.2 million per square millimeter. The Tesla K20c uses the GK110 chip based on the older Kepler architecture, also from TSMC but on a 28 nm process. That chip holds 7,080 million transistors across a much larger 561 mm² die, resulting in a lower density of 12.6 million per square millimeter.

The transistor counts are close, but the die size difference is substantial. The Kepler chip is nearly double the physical area, which tells you the 28 nm node could not pack transistors as tightly. The Turing chip achieves higher performance with fewer transistors and a smaller die, which is a direct benefit of the newer process node.

Memory technology also differs. The GTX 1660 SUPER uses 6 GB of GDDR6 memory with a 192-bit bus and a bandwidth of 336.0 GB/s. The Tesla K20c has 5 GB of GDDR5 memory with a wider 320-bit bus but a lower bandwidth of 208.0 GB/s. The effective memory clock tells the story: the GTX 1660 SUPER runs at 14 Gbps effective, while the Tesla K20c runs at 5.2 Gbps effective. Even with a narrower bus, the GTX 1660 SUPER pushes significantly more data per second.

In terms of compute features, the GTX 1660 SUPER supports FP16 at 10.05 TFLOPS with a 2:1 ratio, while the Tesla K20c has no recorded FP16 capability. FP32 performance favors the GTX 1660 SUPER as well, at 5.027 TFLOPS versus 3.524 TFLOPS. The Tesla K20c has higher texture rate in theory? No, the GTX 1660 SUPER wins there too at 157.1 GTexel/s versus 146.8 GTexel/s. Pixel rate also favors the GTX 1660 SUPER at 85.68 GPixel/s versus 36.71 GPixel/s.

API support is another generational gap. The GTX 1660 SUPER supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The Tesla K20c supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The lower DirectX feature level and older Vulkan version are meaningful for modern games.

The bus interface also differs: PCIe 3.0 x16 on the GTX 1660 SUPER versus PCIe 2.0 x16 on the Tesla K20c. That matters for data transfer between CPU and GPU, especially in systems with older motherboards.

FAQ

Q: Which card is faster in OpenCL compute performance?

A: The NVIDIA GeForce GTX 1660 SUPER scores 52,490 in Geekbench OpenCL, while the Tesla K20c scores 11,479. That is a 357.3% advantage for the GTX 1660 SUPER.

Q: Can I use the Tesla K20c for display output?

A: No. The Tesla K20c has no display outputs. The GTX 1660 SUPER has 1x DVI, 1x HDMI 2.0, and 1x DisplayPort 1.4a.

Q: Which card has more shading units?

A: The Tesla K20c has 2496 shading units, while the GTX 1660 SUPER has 1408. However, the GTX 1660 SUPER still delivers higher FP32 performance at 5.027 TFLOPS versus 3.524 TFLOPS.

Q: What are the power requirements for each card?

A: The GTX 1660 SUPER has a TDP of 125 W and a suggested power supply of 300 W, using a single 8-pin connector. The Tesla K20c has a TDP of 225 W and a suggested power supply of 550 W, using a 6-pin plus an 8-pin connector.

Q: Which card has higher memory bandwidth?

A: The GTX 1660 SUPER has a bandwidth of 336.0 GB/s with 6 GB of GDDR6 on a 192-bit bus. The Tesla K20c has 208.0 GB/s with 5 GB of GDDR5 on a 320-bit bus.

Q: Are both cards end-of-life?

A: Yes, both are marked as end-of-life in the database. The GTX 1660 SUPER was released on October 28, 2019, and the Tesla K20c on November 11, 2012.

Specification Differences

The following fields differ between the two cards:

• Process node: 12 nm (GTX 1660 SUPER) versus 28 nm (Tesla K20c)
• Transistors: 6,600 million versus 7,080 million
• Die size: 284 mm² versus 561 mm²
• Transistor density: 23.2M / mm² versus 12.6M / mm²
• Memory size: 6 GB versus 5 GB
• Memory type: GDDR6 versus GDDR5
• Memory bus width: 192 bit versus 320 bit
• Memory bandwidth: 336.0 GB/s versus 208.0 GB/s
• Effective memory clock: 14 Gbps versus 5.2 Gbps
• Shading units: 1408 versus 2496
• TMUs: 88 versus 208
• ROPs: 48 versus 40
• Pixel rate: 85.68 GPixel/s versus 36.71 GPixel/s
• Texture rate: 157.1 GTexel/s versus 146.8 GTexel/s
• FP32 performance: 5.027 TFLOPS versus 3.524 TFLOPS
• FP16 performance: 10.05 TFLOPS (2:1) versus none recorded
• TDP: 125 W versus 225 W
• Power connectors: 1x 8-pin versus 1x 6-pin + 1x 8-pin
• Suggested PSU: 300 W versus 550 W
• Bus interface: PCIe 3.0 x16 versus PCIe 2.0 x16
• Display outputs: 1x DVI, 1x HDMI 2.0, 1x DisplayPort 1.4a versus no outputs
• DirectX version: 12 (12_1) versus 12 (11_0)
• Vulkan version: 1.4 versus 1.2.175
• Length: 229 mm (9 inches) versus 267 mm (10.5 inches)
• Release date: October 28, 2019 versus November 11, 2012
• Predecessor: GeForce 10 versus Tesla Fermi
• Successor: GeForce 20 versus Tesla Maxwell

Head-to-Head Benchmarks

The only shared benchmark in the database is Geekbench OpenCL. This is a compute test that measures general GPU throughput across a variety of workloads. The results are stark.

The NVIDIA GeForce GTX 1660 SUPER scores 52,490 points. The NVIDIA Tesla K20c scores 11,479 points. The delta is 357.3%, meaning the GTX 1660 SUPER is more than four and a half times faster in this test. That is not a small edge; it is a dominant result that effectively ends the comparison.

To put that in context, look at the nearest rivals for each card. The GTX 1660 SUPER has an average benchmark score of 12,986, and its closest rivals include the RTX 3050 Ti Mobile at 12,940 (0.4% slower), the RX 580 at 12,928 (0.4% slower), and the Tesla M2090 at 13,075 (0.7% faster). The Tesla K20c has an average score of 11,479, with rivals like the Radeon Pro 5500M at 11,528 (0.4% faster), the RX 7800 XT at 11,627 (1.3% faster), and the GTX 1660 at 11,680 (1.7% faster). Notice that the GTX 1660 SUPER's nearest rivals are all within 1% of its average score, indicating it sits right in a competitive cluster. The Tesla K20c's rivals are also close, but that cluster is at a much lower performance level.

The single OpenCL result shows a 357.3% gap between the two cards, which is far larger than the differences between either card and its own nearest rivals. The GTX 1660 SUPER is not just better; it is in a different performance class entirely, at least for this workload.

The GTX 1660 SUPER also has additional benchmark data in other tests: 3DMark Steel Nomad DX12 at 1,278, Geekbench Vulkan at 57,102, PassMark G3D at 12,699, PassMark G2D at 805, PassMark GPU Compute at 5,076, and various DirectX tests (DX9 at 189, DX10 at 65, DX11 at 104, DX12 at 50). The Tesla K20c has no recorded results for any of those tests. That means the GTX 1660 SUPER is the only one of the two with any evidence of gaming or general graphics performance.

Where Each One Wins

Based on the recorded data, the GTX 1660 SUPER wins in every measurable category. It has higher FP32 and FP16 compute, higher memory bandwidth, higher pixel and texture rates, better API support, lower power draw, and it is the only card with display outputs. The OpenCL benchmark confirms a 357.3% performance advantage.

The Tesla K20c does have structural advantages that do not translate into benchmark wins. It has more shading units (2496 versus 1408), more texture mapping units (208 versus 88), and a wider memory bus (320 bit versus 192 bit). In theory, those specs could favor certain compute kernels that are heavily parallel and memory-latency sensitive. However, the only recorded test shows the GTX 1660 SUPER crushing it, so there is no empirical support for that theory in this database.

For gaming, the choice is clear. The GTX 1660 SUPER supports DirectX 12 (12_1), Vulkan 1.4, and OpenGL 4.6, has 6 GB of GDDR6, and outputs to standard displays. The Tesla K20c cannot display anything at all. For general compute, the OpenCL score is decisive. For professional workloads that specifically require Kepler-era Tesla features, the K20c might have niche value, but the data does not show it.

The GTX 1660 SUPER also wins on practical grounds. It is 229 mm long versus 267 mm for the Tesla K20c, fits in a dual-slot width like the K20c, and requires only a 300 W power supply. The Tesla K20c needs 550 W and dual power connectors. The GTX 1660 SUPER is the card you can install in a normal desktop without upgrading your power supply or checking clearance.

In use-case terms, the GTX 1660 SUPER is the choice for anyone building a gaming PC, a general-purpose workstation, or a compute box that also needs display output. The Tesla K20c is only worth considering for a legacy compute node where Kepler-specific code is required and power draw is not a concern. Even then, the 357.3% OpenCL deficit makes that hard to justify. The data says: buy the GTX 1660 SUPER.