NVIDIA GeForce GTX 950 vs NVIDIA Tesla K20c Comparison

The NVIDIA GeForce GTX 950 and NVIDIA Tesla K20c occupy very different corners of the GPU landscape, yet both appear in the database with surprisingly close overall percentile rankings. The GTX 950 sits at the 53rd percentile among all GPUs, while the Tesla K20c lands at the 51st percentile. These two cards, separated by nearly three years in release timing, tell a story of architectural priorities: one built for consumer gaming and display output, the other engineered for compute throughput in professional environments. The recorded measurements show a single head-to-head benchmark, but the broader data reveals how each card approaches its work differently.

Head-to-Head Benchmarks

The only direct comparison available in the database is the Geekbench OpenCL test. Here, the GTX 950 delivers a score of 15,662, while the Tesla K20c manages 11,479. That is a 36.4% advantage for the GTX 950, a substantial margin that flips expectations given the Tesla’s larger silicon and higher raw compute specifications. The GTX 950 wins the sole head-to-head matchup, giving it one win against zero for the Tesla K20c.

What makes this result particularly interesting is the context of the average benchmark scores. The GTX 950’s average score across all recorded tests is 13,189, which is dragged down by its other benchmark results (7,172 in Metal and 16,734 in Vulkan). The Tesla K20c, having only the OpenCL test recorded, shows an average of 11,479, identical to its single score. When placed against their nearest rivals, the GTX 950’s OpenCL performance stands out even more: its nearest competitors in the database include the GeForce GTX 480 (average score 13,300, delta of -0.8%), the Tesla M2090 (13,075, +0.9%), the Radeon Pro 555X (13,321, -1%), and the FirePro M6100 (13,354, -1.2%). The GTX 950’s 15,662 OpenCL score exceeds all of these averages by a wide margin, indicating that this particular workload plays to its strengths.

For the Tesla K20c, the nearest rivals include the Radeon Pro 5500M (11,528, -0.4%), the Radeon RX 7800 XT (11,627, -1.3%), the GeForce GTX 1660 (11,680, -1.7%), and the GeForce GTX 780M (11,261, +1.9%). The Tesla’s 11,479 OpenCL score sits squarely within this cluster, suggesting that its compute performance is competitive with those cards in this specific test, even though those rivals are much newer consumer or workstation parts. The data implies that the Tesla K20c, despite its age, holds its own in OpenCL, but the GTX 950 simply outclasses it in this workload.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The GTX 950 records an average benchmark score of 13,189 across three tests (Metal, OpenCL, and Vulkan), while the Tesla K20c shows an average of 11,479 from its single OpenCL test. The GTX 950 leads by 1,710 points.

Q: What is the difference in memory bandwidth between the two cards?

A: The Tesla K20c offers 208.0 GB/s bandwidth from a 320-bit bus and 5 GB of GDDR5 memory, while the GTX 950 provides 105.8 GB/s from a 128-bit bus and 2 GB of GDDR5. The Tesla’s bandwidth is nearly double, which is typical for a compute-oriented card.

Q: Do both cards support DirectX 12?

A: Yes, but with different feature levels. The GTX 950 supports DirectX 12 (12_1), while the Tesla K20c supports DirectX 12 (11_0). This means the GTX 950 has access to higher-tier DirectX 12 features in the database’s recorded API support.

Q: Which card has more shading units?

A: The Tesla K20c has 2,496 shading units, far exceeding the GTX 950’s 768. However, the GTX 950 still wins the OpenCL benchmark, suggesting that shading unit count alone does not determine real-world performance in this test.

Q: What is the power consumption difference?

A: The GTX 950 is rated at 90 W TDP with a suggested PSU of 250 W, while the Tesla K20c is rated at 225 W TDP with a suggested PSU of 550 W. The Tesla requires substantially more power and a beefier power supply.

Q: Can the Tesla K20c output video to a display?

A: No. The Tesla K20c has no display outputs, while the GTX 950 includes 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.2. The Tesla is designed purely as a compute accelerator without any display connectivity.

Architecture Differences

The two cards stem from different NVIDIA architectures. The GTX 950 uses the GM206 chip based on Maxwell 2.0, while the Tesla K20c uses the GK110 chip based on Kepler. Both are fabricated on a 28 nm process at TSMC, but the silicon itself diverges sharply. The GM206 die measures 228 mm² and contains 2,940 million transistors, yielding a transistor density of 12.9 million per mm². The GK110 die is far larger at 561 mm² and packs 7,080 million transistors, with a slightly lower density of 12.6 million per mm². The Tesla’s die is more than twice the area, reflecting its compute-heavy design with more functional units.

The core configurations differ dramatically. The GTX 950 has 768 shading units, 48 texture mapping units, and 32 ROPs. The Tesla K20c features 2,496 shading units, 208 TMUs, and 40 ROPs. These numbers indicate that the Tesla is built for massive parallel throughput, while the GTX 950 is a more modest, balanced design. The pixel rate is similar: 38.02 GPixel/s for the GTX 950 versus 36.71 GPixel/s for the Tesla, but the texture rate tells a different story. The GTX 950 achieves 57.02 GTexel/s, while the Tesla reaches 146.8 GTexel/s, a 2.6x advantage in texture processing capability. In floating-point performance, the GTX 950 delivers 1.825 TFLOPS (FP32), whereas the Tesla K20c provides 3.524 TFLOPS, nearly double the raw compute throughput. Despite this, the GTX 950 wins the OpenCL test, which suggests that the benchmark favors the Maxwell architecture’s efficiency or the GTX 950’s higher clocks, not just raw specs.

Clock speeds are another differentiator. The GTX 950 has a base clock of 1024 MHz and a boost clock of 1188 MHz, while the Tesla K20c has no recorded base or boost clocks in the database. Memory clocks also differ: the GTX 950 runs at 1653 MHz with 6.6 Gbps effective, while the Tesla K20c runs at 1300 MHz with 5.2 Gbps effective. The Tesla’s wider 320-bit bus compensates for its lower memory clock, resulting in higher overall bandwidth. The GTX 950 uses PCIe 3.0 x16, while the Tesla K20c uses the older PCIe 2.0 x16 interface, which could limit data transfer speeds in some workloads. The Tesla also supports Vulkan 1.2.175 compared to the GTX 950’s Vulkan 1.4, and its DirectX 12 support is capped at level 11_0 versus the GTX 950’s 12_1.

The Verdict

Based strictly on the recorded data, the GTX 950 is the clear winner in the one benchmark where both were tested. Its 36.4% lead in OpenCL, combined with a higher average benchmark score and better API support (DirectX 12_1 and Vulkan 1.4), makes it the more capable card for general-purpose compute as measured here. The Tesla K20c, despite having more shading units, more memory, higher bandwidth, and nearly double the FP32 throughput, falls behind in actual benchmark performance. This suggests that the GTX 950’s newer architecture and higher clocks deliver better real-world efficiency in the Geekbench OpenCL workload.

That said, the Tesla K20c should not be dismissed outright. Its 5 GB memory capacity and 208.0 GB/s bandwidth are far superior to the GTX 950’s 2 GB and 105.8 GB/s, which could matter for memory-bound applications not captured by the single benchmark. The Tesla also has no display outputs, positioning it as a dedicated compute board, whereas the GTX 950 is a consumer card with full display connectivity. The percentile rankings are close (53rd versus 51st), indicating that in the broader database, neither card is an outlier. For a buyer choosing between these two, the GTX 950 offers better measured performance and modern API features, while the Tesla K20c offers more memory and raw compute resources that may be relevant in specialized workloads.

Specification Differences

The two cards differ across nearly every major specification. The GTX 950 uses the GM206 chip on Maxwell 2.0 architecture, while the Tesla K20c uses GK110 on Kepler. The GTX 950 has a base clock of 1024 MHz and boost of 1188 MHz; the Tesla has no recorded clocks. Memory differs: the GTX 950 has 2 GB of GDDR5 on a 128-bit bus with 105.8 GB/s bandwidth, while the Tesla has 5 GB of GDDR5 on a 320-bit bus with 208.0 GB/s. Shading units are 768 versus 2,496, TMUs are 48 versus 208, and ROPs are 32 versus 40. The GTX 950 achieves 38.02 GPixel/s and 57.02 GTexel/s, while the Tesla achieves 36.71 GPixel/s and 146.8 GTexel/s. FP32 performance is 1.825 TFLOPS for the GTX 950 versus 3.524 TFLOPS for the Tesla.

Power requirements are starkly different: the GTX 950 has a 90 W TDP and needs a 250 W PSU with a single 6-pin connector, while the Tesla has a 225 W TDP and needs a 550 W PSU with one 6-pin and one 8-pin connector. The GTX 950 uses PCIe 3.0 x16, the Tesla uses PCIe 2.0 x16. Display outputs are present only on the GTX 950 (1x DVI, 1x HDMI 2.0, 3x DisplayPort 1.2), while the Tesla has none. The GTX 950 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4; the Tesla supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Physical dimensions also differ: the GTX 950 is 202 mm (8 inches) long, while the Tesla is 267 mm (10.5 inches) long. Both are dual-slot cards. The GTX 950 was released in August 2015 with a launch MSRP of 159 USD, while the Tesla K20c was released in November 2012 with a launch MSRP of 3,199 USD.

Where Each One Wins

The GTX 950 wins decisively in the only recorded head-to-head benchmark, the Geekbench OpenCL test, with a 36.4% margin. It also holds advantages in API support, offering DirectX 12 (12_1) versus the Tesla’s 11_0, and Vulkan 1.4 versus 1.2.175. Its higher clocks (1024 MHz base, 1188 MHz boost) and newer architecture contribute to its benchmark success. The GTX 950 is also the only one with display outputs, making it suitable for any task requiring visual output. Its lower TDP of 90 W and smaller physical footprint (202 mm versus 267 mm) mean it fits in more systems with less power overhead.

The Tesla K20c wins in raw specifications that the benchmark does not directly test. Its 5 GB memory capacity is 2.5x larger than the GTX 950’s 2 GB, which is critical for datasets that exceed 2 GB. Its 208.0 GB/s bandwidth is nearly double the GTX 950’s 105.8 GB/s, enabling faster data movement for memory-intensive workloads. The Tesla’s 2,496 shading units and 3.524 TFLOPS FP32 performance indicate a design optimized for massive parallel compute, even if the OpenCL test does not reflect that advantage. Its 146.8 GTexel/s texture rate is 2.6x higher than the GTX 950’s, which could benefit certain texture-heavy compute tasks. The Tesla also has a wider 320-bit memory bus, which typically improves memory efficiency in large data transfers, and its PCIe 2.0 interface, while older, still provides adequate bandwidth for many compute scenarios.

In practical terms, the GTX 950 is the better choice for anyone needing a GPU that can both compute and display, with modern API support and lower power demands. The Tesla K20c is the better choice for a headless compute server where memory capacity and bandwidth outweigh benchmark scores, and where the absence of display outputs is not a limitation. The recorded data shows the GTX 950 winning the only direct comparison, but the Tesla’s specifications suggest it may excel in workloads not represented by the Geekbench OpenCL test. The database records one win for the GTX 950 and zero for the Tesla, yet the Tesla’s higher raw specs cannot be ignored for specialized use cases.