GPU Comparison

The NVIDIA P106-090 and AMD Radeon RX 580 occupy nearly the same position in the overall performance distribution, with the P106-090 sitting at the 54th percentile and the RX 580 at the 53rd. Their average benchmark scores are close, 13,470 for the NVIDIA part versus 12,928 for the AMD part, but that aggregate similarity masks a decisive head-to-head sweep. The RX 580 wins all three directly comparable tests, and in each case the margin is substantial. The data indicates these are not equivalent performers; the RX 580 is a meaningfully faster card, while the P106-090’s average score is propped up by its strong showing in compute-oriented workloads.

Head-to-Head Benchmarks

The most lopsided result comes from the Geekbench Vulkan test, where the AMD Radeon RX 580 scores 45,173 against the NVIDIA P106-090’s 18,596. That is a delta of -58.8% from the AMD card’s perspective, meaning the RX 580 delivers more than double the Vulkan performance. This is a massive gap, not a marginal one, and it reflects fundamental differences in how each architecture handles modern API workloads. The P106-090 is a mining-focused card with no display outputs, and its Vulkan path appears severely constrained relative to the Polaris 20 die.

The 3DMark Steel Nomad DX12 test tells a similar story, though the absolute numbers are smaller. The RX 580 scores 1,005, while the P106-090 manages only 509. The delta here is -49.4%, again favoring the AMD part by a wide margin. In a demanding DX12 scenario, the RX 580 is roughly twice as fast. The P106-090’s score of 509 places it in a range where even older or lower-tier parts can compete, whereas the RX 580’s 1,005 is a solid mid-range result for its generation.

The closest of the three head-to-head tests is Geekbench OpenCL, but it is still far from competitive. The RX 580 posts 37,453, and the P106-090 trails at 21,304, a delta of -43.1%. OpenCL is often a strong suite for NVIDIA hardware, and the P106-090’s Pascal architecture does perform respectably in absolute terms, its score is higher than its DX12 and Vulkan results would suggest. However, the RX 580’s 2,304 shading units and 6.175 TFLOPS of FP32 throughput simply overwhelm the P106-090’s 768 shading units and 2.352 TFLOPS.

The wins are unanimous. The AMD Radeon RX 580 takes all three benchmarks, with deltas ranging from -43.1% to -58.8%. The P106-090 does not win a single head-to-head test. This is not a case where one card wins on some workloads and loses on others; the RX 580 is consistently and overwhelmingly faster across every measured API and workload type.

FAQ

Q: Which card has the higher average benchmark score?

A: The NVIDIA P106-090 has a slightly higher average benchmark score at 13,470, compared to the AMD Radeon RX 580’s 12,928. However, this aggregate figure is misleading, as the RX 580 wins every head-to-head test by margins of at least 43%.

Q: How much faster is the RX 580 in Vulkan?

A: In the Geekbench Vulkan test, the RX 580 scores 45,173 versus the P106-090’s 18,596. This represents a delta of -58.8%, meaning the RX 580 is more than twice as fast in this specific workload.

Q: Does the P106-090 win any benchmark against the RX 580?

A: No. The head-to-head data shows the P106-090 with zero wins and the RX 580 with three wins. The closest result is in OpenCL, where the RX 580 still leads by 43.1%.

Q: What does the percentile ranking tell us?

A: The P106-090 sits at the 54th percentile of all GPUs, and the RX 580 sits at the 53rd. Despite the RX 580’s dominant head-to-head performance, the P106-090’s average score is higher because its compute-focused benchmarks (Geekbench OpenCL at 21,304 and Vulkan at 18,596) are relatively strong compared to its gaming-oriented scores.

Q: How do the nearest rivals compare for each card?

A: The P106-090’s closest rival is the NVIDIA GeForce GTX 570, which scores 13,515, a delta of -0.3%. The RX 580’s closest rival is the NVIDIA GeForce RTX 3050 Ti Mobile, which scores 12,940, a delta of -0.1%. Both cards are tightly clustered with their immediate competitors in the aggregate rankings.

Q: What is the difference in FP32 compute throughput?

A: The RX 580 delivers 6.175 TFLOPS of FP32 performance, while the P106-090 delivers 2.352 TFLOPS. The AMD card has roughly 2.6 times the raw single-precision compute throughput, which explains its large lead in compute-heavy benchmarks.

The Verdict

The data is unambiguous: the AMD Radeon RX 580 is the faster card in every measured scenario. The head-to-head results show a minimum lead of 43.1% in OpenCL, expanding to 49.4% in DX12 and 58.8% in Vulkan. For any workload that relies on modern graphics APIs or general-purpose compute, the RX 580 is the correct choice. Its 6.175 TFLOPS of FP32 throughput, 193.0 GTexel/s texture rate, and 256.0 GB/s memory bandwidth provide a substantial hardware foundation that the P106-090 cannot match with its 2.352 TFLOPS, 73.49 GTexel/s, and 192.2 GB/s.

The P106-090 is not without merit. Its average benchmark score of 13,470 is slightly higher than the RX 580’s 12,928, and its 54th percentile ranking is one point higher. This is driven by its respectable Geekbench OpenCL score of 21,304 and its Vulkan score of 18,596, which are strong for a card with only 768 shading units. In compute tasks that scale well with NVIDIA’s Pascal architecture, the P106-090 can hold its own. However, it is a mining GPU with no display outputs, a PCIe 1.0 x1 bus interface, and a 75 W TDP. Its design goals are power efficiency and compute throughput, not gaming or general-purpose graphics.

The RX 580, with its 8 GB of GDDR5 memory, 256-bit bus, and 1x HDMI 2.0b plus 3x DisplayPort 1.4a outputs, is a fully functional graphics card. It requires a 450 W suggested PSU and a single 8-pin power connector, but it delivers 185 W of performance that decisively beats the P106-090 across all benchmarks. The P106-090’s 75 W TDP and 250 W suggested PSU make it far more power-efficient, but that efficiency comes at the cost of roughly half the performance in every head-to-head test. For anyone choosing between these two, the RX 580 is the only reasonable pick unless the specific constraints of power draw and mining-focused operation are paramount.

Specification Differences

The two cards diverge sharply on almost every specification that matters. The NVIDIA P106-090 is built on a 16 nm process at TSMC, with 4,400 million transistors on a 200 mm² die. The AMD Radeon RX 580 uses a 14 nm process at GlobalFoundries, with 5,700 million transistors on a 232 mm² die. The RX 580 has a higher transistor density at 24.6M per mm² versus 22.0M per mm².

Clock speeds favor the NVIDIA part. The P106-090 runs at a base clock of 1354 MHz and a boost clock of 1531 MHz, compared to the RX 580’s 1257 MHz base and 1340 MHz boost. Memory clocks are nearly identical at 2002 MHz for the P106-090 and 2000 MHz for the RX 580, with both using 8 Gbps effective GDDR5.

Memory configuration is a major differentiator. The P106-090 has 3 GB of GDDR5 on a 192-bit bus, yielding 192.2 GB/s of bandwidth. The RX 580 has 8 GB of GDDR5 on a 256-bit bus, yielding 256.0 GB/s. The RX 580 also has far more shading units (2304 versus 768), texture mapping units (144 versus 48), and a higher texture rate (193.0 GTexel/s versus 73.49 GTexel/s). The P106-090 counters with more ROPs (48 versus 32) and a higher pixel rate (73.49 GPixel/s versus 42.88 GPixel/s).

FP32 throughput heavily favors the RX 580 at 6.175 TFLOPS versus 2.352 TFLOPS. FP16 performance is dramatically different: the RX 580 achieves 6.175 TFLOPS at a 1:1 ratio, while the P106-090 manages only 36.74 GFLOPS at a 1:64 ratio. Power requirements differ substantially, with the P106-090 at 75 W TDP and the RX 580 at 185 W TDP. The suggested PSU is 250 W for the NVIDIA card and 450 W for the AMD card. The P106-090 uses a 1x 6-pin power connector, while the RX 580 uses a 1x 8-pin.

The bus interface is another stark contrast. The P106-090 uses PCIe 1.0 x1, which is a severe bottleneck for a modern GPU. The RX 580 uses PCIe 3.0 x16. Display outputs are absent on the P106-090, while the RX 580 offers 1x HDMI 2.0b and 3x DisplayPort 1.4a. Physical dimensions are similar, with the P106-090 at 250 mm and the RX 580 at 241 mm. The RX 580 has a launch MSRP of 229 USD.

Architecture Differences

The NVIDIA P106-090 is based on the GP106 chip using the Pascal architecture. It is classified under the "Mining GPUs" generation, which explains its lack of display outputs and its PCIe 1.0 x1 interface, neither is needed for a dedicated mining card. Pascal is NVIDIA’s 16 nm architecture, and the P106-090 implements it with 768 shading units. Its FP16 performance of 36.74 GFLOPS at a 1:64 ratio indicates that half-precision compute is heavily de-emphasized, consistent with a card designed for FP32 mining workloads.

The AMD Radeon RX 580 uses the Polaris 20 chip and the GCN 4.0 architecture. It belongs to the Polaris (RX 500) generation, with its predecessor listed as Arctic Islands and its successor as Vega. GCN 4.0 is a mature architecture that emphasizes parallel compute, and the RX 580’s 2304 shading units reflect that design philosophy. Its FP16 performance is identical to its FP32 performance at 6.175 TFLOPS, indicating a 1:1 ratio that allows full-rate half-precision compute.

API support differs slightly. The P106-090 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The RX 580 supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.3. The NVIDIA card has a higher DirectX feature level and a newer Vulkan version, but this does not translate into benchmark wins. In fact, the RX 580’s Vulkan score is 58.8% higher, suggesting that architectural throughput matters more than API version numbers.

The foundry and process choices are notable. TSMC’s 16 nm process gives the P106-090 a smaller die (200 mm²) with fewer transistors (4,400 million). GlobalFoundries’ 14 nm process for the RX 580 results in a larger die (232 mm²) with more transistors (5,700 million). The RX 580’s higher transistor count and density (24.6M per mm²) provide the raw resources for its superior shading, texturing, and memory performance. The P106-090’s lower power envelope (75 W versus 185 W) is a direct consequence of its reduced compute resources and narrower memory bus.