NVIDIA GeForce GTX 580 vs NVIDIA P106-090 Comparison

Head-to-Head Benchmarks

The only direct comparison available in the database is the Geekbench OpenCL test, and it delivers a decisive result. The NVIDIA P106-090 scores 21,304, while the GeForce GTX 580 records 15,283. That translates to a 28.3% margin in favor of the P106-090, a substantial gap that underscores the generational leap between the two architectures.

The GTX 580’s nearest rivals in the database include the GeForce RTX 2060 at 15,290 (a 0% delta), the AMD Radeon 680M at 15,270 (0.1% ahead), the RTX 3050 OEM at 15,199 (0.6% ahead), and the Radeon RX 7600 at 15,171 (0.7% ahead). These figures show that the GTX 580, despite its age, sits in a competitive cluster of modern entry-level and mid-range parts. It is not dramatically outpaced by any of them, with the largest delta being just 0.7%. This suggests that in raw compute workloads, the Fermi-based card still holds its own against much newer silicon.

For the P106-090, its nearest rivals paint a different picture. The GeForce GTX 570 averages 13,515, which is 0.3% ahead of the P106-090. The AMD Radeon Pro 555 scores 13,407, 0.5% behind. The Radeon HD 7770M hits 13,536, 0.5% ahead, and the Radeon RX 9070 XT lands at 13,543, also 0.5% ahead. Interestingly, the P106-090’s average benchmark score is 13,470, which is lower than its Geekbench OpenCL result alone. This is because the database includes additional tests for the P106-090: a 3DMark Steel Nomad DX12 score of 509 and a Geekbench Vulkan score of 18,596, both of which pull the average down. The GTX 580 has only the single OpenCL result, so its average equals 15,283.

The head-to-head delta of -28.3% (with the negative sign indicating the GTX 580 is behind) is the single most informative number in this matchup. It tells us that when both cards are pushed through the same OpenCL compute workload, the Pascal-based P106-090 delivers roughly a third more performance. This is not a marginal win; it is a clear generational advantage in raw throughput.

FAQ

Q: Which card has the higher average benchmark score?

A: The GeForce GTX 580 has a higher average benchmark score of 15,283 compared to the P106-090’s 13,470. However, this is partly because the GTX 580 has only one recorded benchmark, while the P106-090’s average includes lower scores from 3DMark and Vulkan tests.

Q: How much faster is the P106-090 in Geekbench OpenCL?

A: The P106-090 scores 21,304 versus 15,283 for the GTX 580, a 28.3% advantage. This is the only head-to-head benchmark recorded in the database.

Q: Does the P106-090 support Vulkan?

A: Yes, the P106-090 supports Vulkan 1.4 and has a Geekbench Vulkan score of 18,596. The GTX 580 has no Vulkan support listed in the database.

Q: What is the memory capacity difference?

A: The P106-090 has 3 GB of GDDR5 memory, while the GTX 580 has 1536 MB (1.5 GB). The P106-090 offers double the capacity.

Q: Are both cards still in production?

A: No, both are marked as end-of-life in the database. The GTX 580 was released in November 2010, and the P106-090 followed in July 2017.

Q: Which card has more shading units?

A: The P106-090 has 768 shading units, while the GTX 580 has 512. Despite having fewer texture mapping units (48 versus 64), the P106-090 achieves higher pixel and texture rates.

Architecture Differences

The two cards come from entirely different architectural eras. The GTX 580 is built on Fermi 2.0, using the GF110 chip, while the P106-090 uses the Pascal architecture with the GP106 chip. This is a multi-generational jump, and the data reflects it.

The manufacturing process is a major differentiator. The GTX 580 uses a 40 nm node at TSMC, while the P106-090 is fabricated on a 16 nm process, also at TSMC. The transistor counts tell a fascinating story: the GTX 580 packs 3,000 million transistors into a 520 mm² die, yielding a transistor density of 5.8 million per mm². The P106-090, by contrast, integrates 4,400 million transistors into a much smaller 200 mm² die, achieving a density of 22.0 million per mm². The P106-090 has 46.7% more transistors in a die that is 61.5% smaller. This is the clearest possible illustration of how process node shrinks enable denser, more efficient designs.

The memory subsystems differ in bus width and capacity. The GTX 580 uses a 384-bit bus with 1536 MB of GDDR5, while the P106-090 uses a 192-bit bus with 3 GB of GDDR5. Interestingly, the memory bandwidth is nearly identical: 192.4 GB/s for the GTX 580 and 192.2 GB/s for the P106-090. The P106-090 compensates for its narrower bus with faster effective memory speed (8 Gbps versus 4 Gbps).

Compute capabilities also diverge significantly. The GTX 580 delivers 1.581 TFLOPS of FP32 performance, while the P106-090 achieves 2.352 TFLOPS, a 48.8% advantage. The P106-090 also has a listed FP16 performance of 36.74 GFLOPS with a 1:64 ratio, a figure absent for the GTX 580. Pixel rates are starkly different: 24.70 GPixel/s for the GTX 580 versus 73.49 GPixel/s for the P106-090. Texture rates follow the same pattern: 49.41 GTexel/s versus 73.49 GTexel/s.

The API support reflects the newer architecture. The GTX 580 supports DirectX 12 (11_0) and OpenGL 4.6, with no Vulkan listed. The P106-090 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. This means the P106-090 is better equipped for modern graphics APIs.

Specification Differences

The recorded data shows several key specification gaps between the two cards. The most obvious is the process node: 40 nm for the GTX 580 versus 16 nm for the P106-090. This directly impacts die size (520 mm² versus 200 mm²) and transistor density (5.8M versus 22.0M per mm²).

Clock speeds are only listed for the P106-090, with a base clock of 1354 MHz and a boost clock of 1531 MHz. The GTX 580’s base and boost clocks are not recorded in the database, but its memory clock is 1002 MHz (4 Gbps effective), while the P106-090 runs at 2002 MHz (8 Gbps effective).

Memory configuration differs: 1536 MB versus 3 GB, and 384-bit versus 192-bit bus widths. Bandwidth is essentially a tie at 192.4 versus 192.2 GB/s. Shading units favor the P106-090 (768 versus 512), while TMUs favor the GTX 580 (64 versus 48). ROPs are equal at 48.

Power requirements are dramatically different. The GTX 580 has a TDP of 244 W and requires a 550 W suggested PSU, along with 1x 6-pin and 1x 8-pin connectors. The P106-090 sips power at 75 W, needs only a 250 W PSU, and uses a single 6-pin connector. This is a 69.3% reduction in TDP.

The bus interface is another major split: the GTX 580 uses PCIe 2.0 x16, while the P106-090 uses PCIe 1.0 x1. This is unusual, as the P106-090 is a mining-focused card with no display outputs. The GTX 580 has 2x DVI and 1x mini-HDMI 1.3a outputs, while the P106-090 has none.

Physical dimensions are similar in length: 267 mm for the GTX 580 versus 250 mm for the P106-090. Both are dual-slot cards. The GTX 580 has a recorded height of 111 mm, while the P106-090’s height is not listed.

Where Each One Wins

The P106-090 wins in every directly comparable compute metric. Its Geekbench OpenCL score is 28.3% higher. Its FP32 throughput is 48.8% higher. Its pixel rate is nearly triple (73.49 versus 24.70 GPixel/s). Its texture rate is 48.8% higher. It has more memory (3 GB versus 1.5 GB), more shading units (768 versus 512), and a much smaller power footprint (75 W versus 244 W). For any workload that stresses raw compute, memory capacity, or efficiency, the P106-090 is clearly superior.

The GTX 580, however, holds advantages in specific specification areas. It has a wider memory bus (384-bit versus 192-bit), more TMUs (64 versus 48), and a standard PCIe 2.0 x16 interface that allows it to function as a normal graphics card with display outputs. It also has a higher average benchmark score (15,283 versus 13,470), though this is influenced by the different benchmark suites recorded for each card.

The GTX 580’s real-world use case is as a traditional graphics card for gaming or general desktop use, thanks to its display outputs and wider bus. The P106-090, with no display outputs and a PCIe 1.0 x1 interface, is clearly intended for compute-only or mining workloads. Its 3 GB of memory and high compute throughput make it suitable for tasks that do not require a display connection.

The Verdict

The data points to a clear split based on intended use. For anyone needing a functional graphics card with display outputs, the GTX 580 is the only option of the two, as the P106-090 has no outputs at all. The GTX 580 also offers a wider memory bus and more texture units, which could benefit certain workloads.

For compute-heavy tasks, the P106-090 is the dominant choice. Its 28.3% lead in Geekbench OpenCL, 48.8% lead in FP32, and nearly triple pixel rate make it far more capable in raw number crunching. It also consumes 169 W less power, making it dramatically more efficient. The 3 GB memory capacity is double that of the GTX 580, which helps with larger datasets.

The percentile rankings place the GTX 580 at the 58th percentile of all GPUs, while the P106-090 sits at the 54th percentile. This seems counterintuitive given the P106-090’s compute wins, but the GTX 580’s higher average score reflects the fact that its only recorded benchmark is a strong OpenCL result, whereas the P106-090’s average is dragged down by its 3DMark and Vulkan scores.

The verdict, strictly from the data: pick the GTX 580 if you need a traditional GPU with display outputs and a standard PCIe interface. Pick the P106-090 if you prioritize compute performance, memory capacity, and power efficiency, and you can work around its lack of display outputs and limited bus interface. The P106-090 is the better compute engine; the GTX 580 is the better all-around graphics card.