AMD Radeon RX 470 vs NVIDIA GeForce GTX 980 Ti Comparison

Head-to-Head Benchmarks

The benchmark data presents a split result that is unusual in its magnitude and direction. The NVIDIA GeForce GTX 980 Ti takes three of the four head-to-head tests, but the single AMD win is so decisive that it reshapes the overall narrative. Starting with the most striking outlier: in Geekbench Metal, the AMD Radeon RX 470 scores 41,690 points against the GTX 980 Ti's 19,520. That is a 113.6% advantage for the AMD card — more than double the NVIDIA card's output. This is not a marginal victory; it is a categorical one, and it reflects fundamental differences in how each architecture handles the Metal API workload.

The NVIDIA card dominates the remaining tests, though by smaller margins. In 3DMark Steel Nomad DX12, the GTX 980 Ti posts 1,321 points versus the RX 470's 842, a 36.3% lead. This is the largest NVIDIA win and likely reflects the raw compute and rasterization throughput advantages baked into the GM200 chip. In Geekbench OpenCL, the GTX 980 Ti scores 43,513 against 33,568 for the RX 470, a 22.9% edge. The Vulkan test shows a closer contest: 47,724 for NVIDIA versus 39,884 for AMD, a 16.4% difference. Across these three tests, the pattern is consistent — NVIDIA leads by a moderate to substantial margin, but the AMD card is never embarrassed, staying within 37% in its worst result.

The average benchmark scores tell a similar story. The RX 470 averages 28,996 across all tests, while the GTX 980 Ti averages 28,020. This means the AMD card actually holds a slight overall edge of roughly 3.4% in average score, despite losing three of four individual tests. That counterintuitive outcome is explained entirely by the Metal benchmark: the RX 470's 41,690 score is so far above the GTX 980 Ti's 19,520 that it pulls the AMD average above NVIDIA's, even though NVIDIA wins the other three tests by margins of 16.4% to 36.3%. The data thus shows two very different performance profiles — AMD winning decisively in one specific API, NVIDIA winning consistently but less dramatically in three others.

Architecture Differences

The two cards are built on fundamentally different architectures and process technologies. The AMD Radeon RX 470 uses the Ellesmere chip, built on GCN 4.0 architecture with a 14 nm process from GlobalFoundries. The NVIDIA GeForce GTX 980 Ti uses the GM200 chip, built on Maxwell 2.0 architecture with a 28 nm process from TSMC. This process gap — 14 nm versus 28 nm — explains why the AMD chip packs 5,700 million transistors into a 232 mm² die, while the NVIDIA chip needs 8,000 million transistors across a much larger 601 mm² die. The transistor density figures underscore this: AMD achieves 24.6 million transistors per mm², while NVIDIA manages only 13.3 million per mm².

The memory subsystems differ substantially. The RX 470 ships with 4 GB of GDDR5 on a 256-bit bus, delivering 211.2 GB/s of bandwidth. The GTX 980 Ti carries 6 GB of GDDR5 on a 384-bit bus, yielding 336.6 GB/s — roughly 59% more bandwidth. This explains the NVIDIA card's advantage in bandwidth-hungry workloads. The RX 470's memory runs at 1650 MHz (6.6 Gbps effective), while the GTX 980 Ti's memory runs at 1753 MHz (7 Gbps effective). Clock speeds also favor NVIDIA on the core side: the GTX 980 Ti boosts to 1076 MHz versus the RX 470's 1206 MHz boost, but the NVIDIA card starts from a higher base of 1000 MHz versus 926 MHz.

Compute resources tell a more nuanced story. The GTX 980 Ti has 2,816 shading units, 176 texture mapping units, and 96 ROPs. The RX 470 has 2,048 shading units, 128 TMUs, and just 32 ROPs. Despite having fewer shading units, the RX 470's higher clock speeds and newer architecture allow it to reach 4.940 TFLOPS of FP32 performance, while the GTX 980 Ti achieves 6.060 TFLOPS. The pixel rate difference is stark: 103.3 GPixel/s for NVIDIA versus 38.59 GPixel/s for AMD, a 2.7x advantage that reflects the 3x ROP count difference. Texture rates are closer, with NVIDIA at 189.4 GTexel/s and AMD at 154.4 GTexel/s.

Power and physical characteristics diverge sharply. The RX 470 is rated at 120 W TDP with a single 6-pin power connector and a 300 W suggested PSU. The GTX 980 Ti draws 250 W, needs both a 6-pin and an 8-pin connector, and requires a 600 W PSU. The NVIDIA card is also physically larger: 267 mm long, 111 mm tall, and 40 mm wide, versus the RX 470's 240 mm length, 95 mm height, and 35 mm width. Both are dual-slot cards with PCIe 3.0 x16 interfaces. Display outputs differ: AMD offers 1x HDMI 2.0b and 3x DisplayPort 1.4a, while NVIDIA offers 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.2. API support shows AMD with DirectX 12 (12_0) and NVIDIA with DirectX 12 (12_1), while both support OpenGL 4.6 and Vulkan — NVIDIA at 1.4 versus AMD at 1.3.

The Verdict

The data paints a clear picture for different use cases. The NVIDIA GeForce GTX 980 Ti is the stronger card for general GPU compute and modern 3D rendering workloads. It wins the 3DMark Steel Nomad DX12 test by 36.3%, the Geekbench OpenCL test by 22.9%, and the Geekbench Vulkan test by 16.4%. These are not trivial margins — they represent real performance advantages that would manifest in gaming, rendering, and OpenCL-based applications. The GTX 980 Ti's higher FP32 throughput, larger memory pool (6 GB versus 4 GB), and substantially higher memory bandwidth (336.6 GB/s versus 211.2 GB/s) support its dominance in these tests.

The AMD Radeon RX 470 is the correct choice for Metal-based workloads. Its 113.6% lead in Geekbench Metal is the single largest margin in any test, and it is so dominant that it lifts the AMD card's average benchmark score above NVIDIA's. Users working within Apple's Metal ecosystem — whether for creative applications, machine learning, or other Metal-optimized tasks — would see dramatically better performance from the RX 470. The AMD card also offers other advantages: it consumes roughly half the power (120 W versus 250 W), requires a smaller PSU (300 W versus 600 W), and occupies less physical space in a chassis.

The percentile rankings reinforce the split nature of this comparison. The RX 470 sits at the 74th percentile of all GPUs, while the GTX 980 Ti sits at the 73rd percentile — essentially identical overall standings. The nearest rivals for each card confirm the performance tier: the RX 470's nearest neighbor is the AMD Radeon RX 6800M (0.4% delta), while the GTX 980 Ti's is the AMD Radeon Pro W5500X (0.2% delta). Neither card is dramatically better or worse than the other in aggregate; they simply excel in different domains.

FAQ

Q: Which card wins more benchmark tests?

A: The NVIDIA GeForce GTX 980 Ti wins three of the four head-to-head tests — 3DMark Steel Nomad DX12, Geekbench OpenCL, and Geekbench Vulkan. The AMD Radeon RX 470 wins only Geekbench Metal, but by a much larger margin of 113.6%.

Q: What is the largest performance gap between the two cards?

A: The largest gap is in Geekbench Metal, where the AMD Radeon RX 470 scores 41,690 versus the GTX 980 Ti's 19,520 — a 113.6% difference in AMD's favor. The next largest is 3DMark Steel Nomad DX12, where NVIDIA leads by 36.3%.

Q: How does memory capacity compare?

A: The NVIDIA GeForce GTX 980 Ti has 6 GB of GDDR5 memory on a 384-bit bus with 336.6 GB/s bandwidth. The AMD Radeon RX 470 has 4 GB of GDDR5 on a 256-bit bus with 211.2 GB/s bandwidth. NVIDIA's memory bandwidth is about 59% higher.

Q: Which card has better average benchmark scores?

A: The AMD Radeon RX 470 has a slightly higher average benchmark score of 28,996, compared to the GTX 980 Ti's 28,020. This is driven entirely by the RX 470's massive Metal test victory.

Q: What are the power requirements for each card?

A: The AMD Radeon RX 470 has a 120 W TDP and requires a 300 W PSU with a single 6-pin connector. The NVIDIA GeForce GTX 980 Ti has a 250 W TDP, needs a 600 W PSU, and uses one 6-pin plus one 8-pin connector.

Q: How do the architectures differ in process node?

A: The AMD Radeon RX 470 uses GCN 4.0 on a 14 nm process from GlobalFoundries, while the NVIDIA GeForce GTX 980 Ti uses Maxwell 2.0 on a 28 nm process from TSMC. AMD achieves 24.6M transistors per mm² versus NVIDIA's 13.3M per mm².

Where Each One Wins

AMD Radeon RX 470 wins in Metal-specific workloads. The 113.6% lead in Geekbench Metal is the defining result of this comparison. Any application that leverages Apple's Metal API — including many creative, productivity, and scientific computing tools on macOS — will run substantially faster on the RX 470. The card's GCN 4.0 architecture, despite having fewer shading units (2,048 versus 2,816), clearly handles Metal compute more efficiently than Maxwell 2.0. The RX 470 also wins on efficiency metrics: 120 W TDP, a single 6-pin connector, and a 300 W PSU requirement make it far easier to integrate into existing systems. Its smaller physical footprint (240 mm versus 267 mm length) and lower transistor density requirements also make it a more practical card for compact builds.

NVIDIA GeForce GTX 980 Ti wins in traditional 3D rendering and compute benchmarks. The 36.3% lead in 3DMark Steel Nomad DX12 indicates that the GTX 980 Ti is the stronger card for modern DirectX 12 gaming and rendering workloads. Its OpenCL lead of 22.9% and Vulkan lead of 16.4% show broad compute superiority outside of Metal. The GTX 980 Ti's raw resources — 2,816 shading units, 176 TMUs, 96 ROPs, 6 GB of memory, and 336.6 GB/s bandwidth — give it a hardware base that the RX 470 cannot match in most scenarios. Its 6.060 TFLOPS FP32 performance versus 4.940 TFLOPS for AMD, and its 103.3 GPixel/s pixel rate versus 38.59 GPixel/s, are decisive in rasterization-heavy tasks. The NVIDIA card also supports DirectX 12 (12_1) and Vulkan 1.4, while AMD supports DirectX 12 (12_0) and Vulkan 1.3, giving it a slight API feature advantage.

For users who need both Metal and general compute, the choice depends on which workload dominates. The RX 470's average score advantage (28,996 versus 28,020) is real but small, and it is entirely attributable to the Metal outlier. In any mixed workload where Metal is not the primary API, the GTX 980 Ti's consistent wins across three different test suites make it the more reliable performer.