NVIDIA GeForce GTX 590 vs NVIDIA Tesla K20c Comparison

# NVIDIA Tesla K20c vs NVIDIA GeForce GTX 590

The NVIDIA Tesla K20c and NVIDIA GeForce GTX 590 represent two very different philosophies from the same company, separated by roughly 20 months of GPU evolution. The K20c is a compute-focused accelerator built on the Kepler architecture, while the GTX 590 is a dual-GPU Fermi flagship aimed at raw graphics performance. The single benchmark available — Geekbench OpenCL — shows the GTX 590 scoring 12,830 against the K20c's 11,479, a 10.5% delta in favor of the older card. Yet the data reveals a more nuanced story about specialization, memory capacity, and architectural direction than that single score suggests.

The Verdict

The data points to a clear but narrow victory for the GTX 590 in the OpenCL benchmark. With a score of 12,830, it sits at the 52nd percentile among all GPUs, while the K20c's 11,479 places it at the 51st percentile. The delta of 10.5% is meaningful, but the two cards occupy nearly identical positions in the overall performance distribution. For users whose workload is dominated by OpenCL compute tasks, the GTX 590 offers a measurable advantage based on this metric alone. However, the K20c brings 5 GB of GDDR5 memory versus the GTX 590's 1,536 MB — a 3.25x capacity advantage that the benchmark does not capture. The K20c also consumes 225 W compared to the GTX 590's 365 W, making it the more power-efficient choice on paper. The verdict from the data: the GTX 590 wins the single measured performance test, but the K20c is the more balanced accelerator for memory-intensive or power-constrained deployments.

Where Each One Wins

The GTX 590 wins the only head-to-head benchmark provided, the Geekbench OpenCL test, with a 10.5% margin. Its launch MSRP of 699 USD contrasts sharply with the K20c's 3,199 USD launch MSRP — though price comparisons are not the focus here. The GTX 590 also offers display outputs (3x DVI and 1x mini-DisplayPort), making it usable as a conventional graphics card, whereas the K20c has no display outputs at all. This makes the GTX 590 the practical choice for a desktop workstation that needs both compute and visual output.

The K20c, despite losing the compute benchmark, wins on memory capacity (5 GB vs 1,536 MB), memory bandwidth (208.0 GB/s vs 164.0 GB/s), and thermal design power (225 W vs 365 W). Its texture rate of 146.8 GTexel/s dwarfs the GTX 590's 38.91 GTexel/s, and its pixel rate of 36.71 GPixel/s nearly doubles the GTX 590's 19.46 GPixel/s. For workloads that stress memory capacity or texture throughput, the K20c's architecture provides clear advantages that a single OpenCL score cannot represent.

Architecture Differences

The two GPUs come from different architectural generations. The K20c uses the GK110 chip built on Kepler architecture at a 28 nm process node, while the GTX 590 uses the GF110 chip on Fermi 2.0 architecture at a 40 nm node. Both are manufactured by TSMC, but the process shrink is substantial. The K20c packs 7,080 million transistors into a 561 mm² die, yielding a transistor density of 12.6M per mm². The GTX 590 contains 3,000 million transistors across a 520 mm² die, with a density of just 5.8M per mm². The K20c's transistor count is more than double the GTX 590's, crammed into only slightly more silicon area.

Shading unit counts differ dramatically: the K20c has 2,496 shading units and 208 TMUs, while the GTX 590 has 512 shading units and 64 TMUs. The K20c's ROP count of 40 is actually lower than the GTX 590's 48, which explains why pixel rate does not scale proportionally with shader count. The K20c supports Vulkan 1.2.175, while the GTX 590 has no Vulkan support listed. Both support DirectX 12 (11_0) and OpenGL 4.6. Memory clocks differ: the K20c runs at 1300 MHz (5.2 Gbps effective) versus the GTX 590's 854 MHz (3.4 Gbps effective), though the GTX 590's wider 384-bit bus partially compensates for its lower clock speed.

FAQ

Q: Which GPU has more memory?

A: The Tesla K20c has 5 GB of GDDR5 memory on a 320-bit bus, while the GTX 590 has 1,536 MB on a 384-bit bus. The K20c's capacity is over three times larger.

Q: What is the performance difference in the Geekbench OpenCL test?

A: The GTX 590 scores 12,830 versus the K20c's 11,479, giving the GTX 590 a 10.5% lead. This is the only head-to-head benchmark provided.

Q: Which card has higher memory bandwidth?

A: The K20c achieves 208.0 GB/s, while the GTX 590 reaches 164.0 GB/s. The K20c's advantage comes from its faster memory clock of 1300 MHz versus 854 MHz.

Q: What are the power consumption differences?

A: The K20c has a TDP of 225 W with a suggested 550 W PSU, while the GTX 590 has a TDP of 365 W and requires a 750 W PSU. The K20c is the more power-efficient option.

Q: Can either card output video?

A: Only the GTX 590 has display outputs: 3x DVI and 1x mini-DisplayPort. The K20c has no display outputs, making it strictly a compute accelerator.

Q: Which card uses a smaller manufacturing process?

A: The K20c is built on TSMC's 28 nm process, while the GTX 590 uses 40 nm. The K20c also has a higher transistor density at 12.6M per mm² versus 5.8M per mm².

Head-to-Head Benchmarks

The single Geekbench OpenCL result provides the only direct comparison. The GTX 590's score of 12,830 beats the K20c's 11,479 by 10.5%. In the context of their nearest rivals, the K20c sits just 0.4% below the AMD Radeon Pro 5500M (11,528) and 1.3% below the AMD Radeon RX 7800 XT (11,627). It edges out the NVIDIA GeForce GTX 780M (11,261) by 1.9%. The GTX 590, meanwhile, matches the AMD Radeon Pro 455 (12,831) almost exactly with a 0% delta, sits 0.1% below the AMD FirePro W5100 (12,847), and runs 0.3% under the AMD Radeon 740M (12,870). It leads the NVIDIA GeForce GTX 670 (12,773) by 0.4%.

The 10.5% gap between the two cards is notable, but the surrounding rival data suggests both occupy similar performance tiers. The K20c's nearest rivals span from 11,261 to 11,680, while the GTX 590's span from 12,731 to 12,870. There is no overlap between these ranges, confirming the GTX 590's consistent advantage in this test. However, the K20c's massive lead in texture rate (146.8 vs 38.91 GTexel/s) and pixel rate (36.71 vs 19.46 GPixel/s) indicates that the Geekbench OpenCL test may not fully exercise the K20c's strengths. The benchmark winner also has the older architecture and nearly half the transistor count, which raises questions about what specific workloads the test favors.

Specification Differences

The two cards differ on nearly every measurable specification. Process node: 28 nm for the K20c versus 40 nm for the GTX 590. Transistors: 7,080 million versus 3,000 million. Die size: 561 mm² versus 520 mm². Transistor density: 12.6M per mm² versus 5.8M per mm². Memory: 5 GB versus 1,536 MB. Memory clock: 1300 MHz versus 854 MHz. Memory bandwidth: 208.0 GB/s versus 164.0 GB/s. Shading units: 2,496 versus 512. TMUs: 208 versus 64. ROPs: 40 versus 48. Pixel rate: 36.71 GPixel/s versus 19.46 GPixel/s. Texture rate: 146.8 GTexel/s versus 38.91 GTexel/s. FP32 performance: 3.524 TFLOPS versus 1,244.2 GFLOPS. TDP: 225 W versus 365 W. Power connectors: 1x 6-pin + 1x 8-pin versus 2x 8-pin. Suggested PSU: 550 W versus 750 W. Dimensions: 267 mm length versus 279 mm length (the GTX 590 also has height and width listed at 111 mm and 40 mm, respectively). Display outputs: none versus 3x DVI and 1x mini-DisplayPort. Architecture: Kepler versus Fermi 2.0. Generation: Tesla Kepler versus GeForce 500.

The FP32 gap is particularly striking — the K20c delivers 3.524 TFLOPS, nearly three times the GTX 590's 1,244.2 GFLOPS. Yet the GTX 590 still wins the OpenCL benchmark. This suggests that the K20c's compute capability is not being fully utilized by the test, or that other factors like driver optimization, memory latency, or scheduling overhead play a larger role than raw throughput. The GTX 590's higher ROP count (48 vs 40) may also contribute to its benchmark showing, as ROPs handle final pixel output and certain compute operations. The K20c's dual-slot cooler and single 6-pin plus 8-pin configuration contrast with the GTX 590's dual-slot design needing two 8-pin connectors, reflecting the latter's higher power draw. Both cards are end-of-life products, with the K20c released in late 2012 and the GTX 590 in early 2011. The K20c's predecessor is Tesla Fermi and its successor is Tesla Maxwell; the GTX 590's predecessor is GeForce 400 and its successor is GeForce 600.