AMD Radeon RX 580 2048SP vs NVIDIA P106-100 Comparison

The NVIDIA P106-100 and the AMD Radeon RX 580 2048SP are two end-of-life graphics cards that occupy a similar performance tier, yet they achieve their scores through fundamentally different design philosophies. The data reveals a fascinating split: the NVIDIA card dominates compute and DirectX 12 workloads, while the AMD card fights back decisively in Vulkan. With an average benchmark score of 23249 for the P106-100 and 23061 for the RX 580 2048SP, the overall performance delta is a razor-thin 0.8% in favor of the NVIDIA part, placing both cards at the 68th percentile of all GPUs. This is a matchup defined by specific workload advantages rather than a clear overall winner.

Where Each One Wins

The benchmark results paint a clear picture of two distinct personalities. The NVIDIA P106-100 is the champion of compute-heavy and DirectX 12 applications. Its score of 35951 in Geekbench OpenCL is a massive 21.2% higher than the AMD card’s 29668, indicating a substantial advantage in general-purpose compute tasks that leverage OpenCL. Similarly, in the 3DMark Steel Nomad DirectX 12 test, the P106-100 scores 899 against the RX 580 2048SP’s 849, a 5.9% lead. This suggests that for users running modern DirectX 12 games or compute-centric applications, the NVIDIA card is the stronger performer.

Conversely, the AMD Radeon RX 580 2048SP finds its winning edge in the Vulkan API. The data shows a decisive victory for AMD in the Geekbench Vulkan test, where it scores 38666 compared to the P106-100’s 32897. This is a 14.9% advantage for the AMD card, flipping the script on the NVIDIA card’s compute lead. This makes the RX 580 2048SP the better choice for software titles and games that are built on the Vulkan API, where its architecture appears to scale more efficiently. The use-case split is therefore simple: choose the NVIDIA P106-100 for OpenCL compute and DirectX 12, and choose the AMD Radeon RX 580 2048SP for Vulkan-centric workloads.

Architecture Differences

The two cards are built on entirely different architectures and manufacturing processes, which explains their divergent benchmark behavior. The NVIDIA P106-100 is based on the GP106 chip, utilizing the Pascal architecture, and is fabricated by TSMC on a 16 nm process. The chip contains 4,400 million transistors on a 200 mm² die, resulting in a transistor density of 22.0M / mm². In contrast, the AMD Radeon RX 580 2048SP uses the Polaris 20 chip, built on the older GCN 4.0 architecture, and is manufactured by GlobalFoundries on a 14 nm process. This chip is substantially larger, with 5,700 million transistors on a 232 mm² die, giving it a higher transistor density of 24.6M / mm².

These architectural differences lead to vastly different core configurations. The AMD card boasts 2048 shading units, 128 TMUs, and 32 ROPs, while the NVIDIA card has fewer resources with 1280 shading units, 80 TMUs, and 48 ROPs. This explains the AMD card’s higher theoretical texture fill rate of 164.4 GTexel/s versus the NVIDIA’s 136.7 GTexel/s, and its higher FP32 throughput of 5.259 TFLOPS versus 4.375 TFLOPS. However, the NVIDIA card’s higher ROP count contributes to its superior pixel rate of 82.03 GPixel/s, far exceeding the AMD card’s 41.09 GPixel/s. The memory subsystems also differ significantly, with the P106-100 using a 192-bit bus for 6 GB of GDDR5 (192.2 GB/s bandwidth) while the RX 580 2048SP uses a wider 256-bit bus but only 4 GB of GDDR5 (224.0 GB/s bandwidth). The AMD card supports full-rate FP16 at 5.259 TFLOPS (1:1), whereas the NVIDIA card has a severely gimped FP16 rate of 68.36 GFLOPS (1:64), a key architectural distinction.

Head-to-Head Benchmarks

Looking at the direct comparisons, the NVIDIA P106-100 wins two out of three benchmarks, but the magnitudes of the wins and losses tell a complex story. The largest victory for NVIDIA is in the Geekbench OpenCL test, where the P106-100 scores 35951 against the RX 580 2048SP’s 29668. This 21.2% delta is the single biggest performance gap in the entire comparison, highlighting a fundamental strength in NVIDIA’s compute stack. The P106-100 also wins the 3DMark Steel Nomad DX12 test with a score of 899 versus 849, a more modest 5.9% lead that still indicates better performance in this specific DirectX 12 workload.

The AMD Radeon RX 580 2048SP, however, delivers a powerful counterpunch in the Geekbench Vulkan test. Its score of 38666 dwarfs the NVIDIA card’s 32897, establishing a 14.9% advantage. This is a significant margin, and it demonstrates that the GCN architecture, despite being older, has strong Vulkan drivers and hardware utilization. While the NVIDIA card wins on volume (2 wins vs 1), the AMD card’s single victory is substantial. The overall average benchmark scores are close—23249 for NVIDIA and 23061 for AMD—but the path to those scores is completely different, with NVIDIA’s wins being distributed across compute and DX12, while AMD’s win is concentrated entirely in Vulkan.

The Verdict

From the data, the choice between these two cards is dictated entirely by the target application. If the workload is dominated by OpenCL compute or DirectX 12 gaming, the NVIDIA P106-100 is the clear choice. Its 21.2% lead in OpenCL and 5.9% lead in DX12 are significant enough to make it the superior performer in those areas. The card’s higher pixel rate of 82.03 GPixel/s also suggests it may handle certain rasterization tasks more efficiently, despite its lower raw shader count.

If the priority is Vulkan performance, the AMD Radeon RX 580 2048SP is the only logical pick. Its 14.9% advantage in the Geekbench Vulkan test is a decisive margin that cannot be ignored. The card’s higher FP32 throughput and texture fill rate also give it a theoretical edge in traditional rendering workloads that scale well with those resources. The AMD card also supports full-rate FP16, which could be a boon for specific compute tasks that use half-precision arithmetic, even though the NVIDIA card’s FP16 performance is negligible. Ultimately, the data shows no universal winner; it shows two specialized tools for different jobs.

FAQ

Q: Which card has a higher average benchmark score?

A: The NVIDIA P106-100 has a slightly higher average benchmark score of 23249, compared to the AMD Radeon RX 580 2048SP’s 23061.

Q: How large is the NVIDIA card’s lead in the OpenCL benchmark?

A: The NVIDIA P106-100 scores 35951 in Geekbench OpenCL, which is 21.2% higher than the AMD card’s score of 29668.

Q: Does the AMD card win any benchmark?

A: Yes, the AMD Radeon RX 580 2048SP wins the Geekbench Vulkan test with a score of 38666, which is 14.9% higher than the NVIDIA P106-100’s score of 32897.

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

A: The AMD Radeon RX 580 2048SP has a higher memory bandwidth of 224.0 GB/s, while the NVIDIA P106-100 has a bandwidth of 192.2 GB/s.

Q: Which card has a higher transistor count?

A: The AMD Radeon RX 580 2048SP has a higher transistor count of 5,700 million, compared to the NVIDIA P106-100’s 4,400 million transistors.

Specification Differences

The following specifications differ between the two cards:

• Chip: NVIDIA P106-100 uses "GP106"; AMD Radeon RX 580 2048SP uses "Polaris 20".
• Architecture: NVIDIA P106-100 uses "Pascal"; AMD Radeon RX 580 2048SP uses "GCN 4.0".
• Process Node: NVIDIA P106-100 is on "16 nm"; AMD Radeon RX 580 2048SP is on "14 nm".
• Foundry: NVIDIA P106-100 uses "TSMC"; AMD Radeon RX 580 2048SP uses "GlobalFoundries".
• Transistors: NVIDIA P106-100 has "4,400 million"; AMD Radeon RX 580 2048SP has "5,700 million".
• Die Size: NVIDIA P106-100 is "200 mm²"; AMD Radeon RX 580 2048SP is "232 mm²".
• Transistor Density: NVIDIA P106-100 has "22.0M / mm²"; AMD Radeon RX 580 2048SP has "24.6M / mm²".
• Base Clock: NVIDIA P106-100 is "1506 MHz"; AMD Radeon RX 580 2048SP is "1168 MHz".
• Boost Clock: NVIDIA P106-100 is "1709 MHz"; AMD Radeon RX 580 2048SP is "1284 MHz".
• Memory Clock: NVIDIA P106-100 is "2002 MHz / 8 Gbps effective"; AMD Radeon RX 580 2048SP is "1750 MHz / 7 Gbps effective".
• Memory Size: NVIDIA P106-100 has "6 GB"; AMD Radeon RX 580 2048SP has "4 GB".
• Memory Bus Width: NVIDIA P106-100 is "192 bit"; AMD Radeon RX 580 2048SP is "256 bit".
• Memory Bandwidth: NVIDIA P106-100 is "192.2 GB/s"; AMD Radeon RX 580 2048SP is "224.0 GB/s".
• Shading Units: NVIDIA P106-100 has "1280"; AMD Radeon RX 580 2048SP has "2048".
• TMUs: NVIDIA P106-100 has "80"; AMD Radeon RX 580 2048SP has "128".
• ROPs: NVIDIA P106-100 has "48"; AMD Radeon RX 580 2048SP has "32".
• Pixel Rate: NVIDIA P106-100 is "82.03 GPixel/s"; AMD Radeon RX 580 2048SP is "41.09 GPixel/s".
• Texture Rate: NVIDIA P106-100 is "136.7 GTexel/s"; AMD Radeon RX 580 2048SP is "164.4 GTexel/s".
• FP32: NVIDIA P106-100 is "4.375 TFLOPS"; AMD Radeon RX 580 2048SP is "5.259 TFLOPS".
• FP16: NVIDIA P106-100 is "68.36 GFLOPS (1:64)"; AMD Radeon RX 580 2048SP is "5.259 TFLOPS (1:1)".
• TDP: NVIDIA P106-100 is "120 W"; AMD Radeon RX 580 2048SP is "150 W".
• Power Connectors: NVIDIA P106-100 uses "1x 6-pin"; AMD Radeon RX 580 2048SP uses "1x 8-pin".
• Suggested PSU: NVIDIA P106-100 is "300 W"; AMD Radeon RX 580 2048SP is "450 W".
• Bus Interface: NVIDIA P106-100 is "PCIe 1.0 x16"; AMD Radeon RX 580 2048SP is "PCIe 3.0 x16".
• Display Outputs: NVIDIA P106-100 has "No outputs"; AMD Radeon RX 580 2048SP has "1x DVI, 1x HDMI 2.0b, 3x DisplayPort 1.4a".
• DirectX Support: NVIDIA P106-100 is "12 (12_1)"; AMD Radeon RX 580 2048SP is "12 (12_0)".
• Vulkan Support: NVIDIA P106-100 is "1.4"; AMD Radeon RX 580 2048SP is "1.3".
• Length: NVIDIA P106-100 is "250 mm / 9.8 inches"; AMD Radeon RX 580 2048SP is "241 mm / 9.5 inches".
• Release Date: NVIDIA P106-100 is "2017-06-18"; AMD Radeon RX 580 2048SP is "2018-10-14".