NVIDIA GeForce GTX 1660 vs NVIDIA Tesla K20Xm Comparison

The NVIDIA Tesla K20Xm and the NVIDIA GeForce GTX 1660 represent two very different eras of GPU design, and the benchmark data reflects that divide clearly. The K20Xm is a Kepler-era compute monster from 2012, while the GTX 1660 is a Turing-based consumer card from 2019. In the only head-to-head benchmark available, the GeForce GTX 1660 dominates, scoring 47,850 in Geekbench OpenCL against the Tesla K20Xm’s 17,215. That is a massive 64% delta, meaning the GTX 1660 delivers nearly three times the raw compute performance in that specific test. The K20Xm’s average benchmark score of 12,625 puts it slightly ahead of the GTX 1660’s 11,680, but that is skewed by the limited test sets each card was subjected to. The GTX 1660’s nearest rival data shows it trading blows with modern cards like the AMD Radeon RX 7800 XT (0.5% delta) and the RX 6500 XT (-1.4%), while the K20Xm sits near older cards like the GTX 670 (-1.2%) and GTX 590 (-1.6%). The data is clear: in modern compute workloads, the GTX 1660 is in a different league.

Head-to-Head Benchmarks

The sole direct comparison in the data is the Geekbench OpenCL test, and it is not close. The GTX 1660 scores 47,850, while the K20Xm scores 17,215. The deltaPct of -64% indicates the K20Xm is 64% behind the GTX 1660. This is a decisive victory for the newer card. It is worth remembering the GTX 1660 also has Vulkan and DirectX 12 benchmark scores (50,137 and 1,065 respectively) that the K20Xm simply cannot match, as the older card has no corresponding entries. The K20Xm’s only other benchmark is Geekbench Metal at 8,035, which is not a cross-compatible test. Looking at the average benchmark scores, the K20Xm’s 12,625 actually edges out the GTX 1660’s 11,680. However, this is misleading because the averages are computed from different benchmark suites. The GTX 1660’s Passmark G3D score of 11,646 is far more representative of real-world graphics performance than the K20Xm’s compute-focused results. The K20Xm’s percentile rank of 52 versus the GTX 1660’s 51 shows they sit in a similar overall position in the database, but the GTX 1660 achieves this with a fraction of the power draw and far more modern feature support.

Where Each One Wins

The K20Xm wins in the niche of legacy compute workloads that favor its massive 2,688 shading units and 224 texture mapping units. Its 3.935 TFLOPS of FP32 performance is substantial, and its 249.6 GB/s of memory bandwidth over a 384-bit bus is higher than the GTX 1660’s 192.1 GB/s over a 192-bit bus. For tasks that are bandwidth-bound or that scale with raw shader count, the K20Xm can still hold its own. However, the GTX 1660 wins in nearly every other measurable category. Its 5.027 TFLOPS of FP32 performance is 28% higher than the K20Xm’s, and its FP16 performance of 10.05 TFLOPS (2:1) shows support for half-precision workloads. The GTX 1660 also has modern API support, including DirectX 12 (12_1) and Vulkan 1.4, whereas the K20Xm is limited to DirectX 12 (11_0) and Vulkan 1.2.175. In practical terms, the GTX 1660 is the clear winner for any gaming, modern rendering, or general-purpose compute task. The K20Xm’s lack of display outputs makes it useless for any interactive workload, while the GTX 1660 offers 1x DVI, 1x HDMI 2.0, and 1x DisplayPort 1.4a.

Architecture Differences

The architectural gap between these two cards is generational. The K20Xm uses the GK110 chip on the Kepler architecture, built on a 28 nm process at TSMC. It packs 7,080 million transistors on a massive 561 mm² die, resulting in a transistor density of 12.6M per mm². The GTX 1660 uses the TU116 chip on the Turing architecture, built on a 12 nm process, also at TSMC. It has 6,600 million transistors on a much smaller 284 mm² die, giving it a transistor density of 23.2M per mm². That is nearly double the density, which explains how the GTX 1660 achieves higher performance with fewer transistors. The K20Xm has 2,688 shading units, 224 TMUs, and 48 ROPs, while the GTX 1660 has 1,408 shading units, 88 TMUs, and 48 ROPs. Despite having half the shading units, the GTX 1660’s higher clock speeds (1,530 MHz base, 1,785 MHz boost) versus the K20Xm’s unspecified clocks (only memory at 1,300 MHz) allow it to outpace the older card. The GTX 1660 also has a lower TDP of 120 W versus the K20Xm’s 235 W, making it far more efficient. The K20Xm has no display outputs, while the GTX 1660 has a full set. Both cards have 6 GB of GDDR5 memory, but the K20Xm’s wider 384-bit bus gives it more bandwidth. Neither card has ray tracing or tensor cores.

FAQ

Q: Which card is faster in OpenCL compute?

A: The GTX 1660 is dramatically faster, scoring 47,850 in Geekbench OpenCL versus the K20Xm’s 17,215, a 64% delta in favor of the GTX 1660.

Q: Do both cards have the same amount of memory?

A: Yes, both have 6 GB of GDDR5 memory, but the K20Xm has a 384-bit bus with 249.6 GB/s bandwidth, while the GTX 1660 has a 192-bit bus with 192.1 GB/s bandwidth.

Q: Which card has better API support?

A: The GTX 1660 supports DirectX 12 (12_1) and Vulkan 1.4, while the K20Xm is limited to DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6.

Q: Can the K20Xm be used for display output?

A: No, the K20Xm has no display outputs. It is a compute-only card, whereas the GTX 1660 has 1x DVI, 1x HDMI 2.0, and 1x DisplayPort 1.4a.

Q: What is the power consumption difference?

A: The K20Xm has a TDP of 235 W and requires a 550 W power supply, while the GTX 1660 has a TDP of 120 W and requires only a 300 W power supply.

Q: How do these cards rank against all other GPUs?

A: The K20Xm sits at the 52nd percentile, while the GTX 1660 is at the 51st percentile. The K20Xm’s average benchmark score is 12,625, slightly higher than the GTX 1660’s 11,680.

Specification Differences

The two cards differ in nearly every specification except memory size and ROP count. The K20Xm uses a 28 nm process, while the GTX 1660 uses 12 nm. The K20Xm has 7,080 million transistors on a 561 mm² die, while the GTX 1660 has 6,600 million on 284 mm². Transistor density is 12.6M/mm² for the K20Xm and 23.2M/mm² for the GTX 1660. The K20Xm has 2,688 shading units and 224 TMUs, versus the GTX 1660’s 1,408 shading units and 88 TMUs. The K20Xm’s memory clock is 1,300 MHz (5.2 Gbps effective), while the GTX 1660’s is 2,001 MHz (8 Gbps effective). The K20Xm’s bandwidth is 249.6 GB/s, and the GTX 1660’s is 192.1 GB/s. The K20Xm’s pixel rate is 40.99 GPixel/s, and its texture rate is 164.0 GTexel/s, while the GTX 1660 achieves 85.68 GPixel/s and 157.1 GTexel/s. FP32 performance is 3.935 TFLOPS for the K20Xm and 5.027 TFLOPS for the GTX 1660. The K20Xm has a TDP of 235 W and a suggested PSU of 550 W, while the GTX 1660 has a TDP of 120 W and a suggested PSU of 300 W. The K20Xm has no power connectors listed and no display outputs, while the GTX 1660 has 1x 8-pin power and three display outputs. The K20Xm is 267 mm long, while the GTX 1660 is 229 mm long. The K20Xm’s launch MSRP was 7,699 USD, and the GTX 1660’s was 219 USD.

The Verdict

The data points to one clear conclusion: the GTX 1660 is the superior card for almost any use case. It delivers 64% higher OpenCL performance, 28% higher FP32 throughput, and vastly better API support. It is also more efficient, with a 120 W TDP versus the K20Xm’s 235 W, and it has display outputs, making it usable for gaming and general graphics. The K20Xm’s only advantages are its higher memory bandwidth (249.6 GB/s vs. 192.1 GB/s) and its higher average benchmark score (12,625 vs. 11,680), but the latter is based on a limited and different set of tests. If you are choosing between these two for any modern workload, the GTX 1660 is the obvious pick. The K20Xm is a historical curiosity, a compute-focused relic from the Kepler era that has been thoroughly outclassed by a much cheaper, much newer Turing card. For anyone building a system today, the GTX 1660 is the only rational choice, and the benchmark data supports that without reservation.