AMD Radeon RX 590 vs NVIDIA GeForce GTX 980 Ti Comparison

The NVIDIA GeForce GTX 980 Ti and the AMD Radeon RX 590 are two end-of-life cards from different eras that end up surprisingly close in the database rankings, separated by just a few points of overall standing and one win apiece in direct benchmark comparisons. The GTX 980 Ti, a Maxwell 2.0 flagship built on TSMC's 28 nm node, holds the edge in classic rasterization throughput and DirectX 12 gaming tests. The RX 590, a Polaris 30 refresh on GlobalFoundries' 12 nm process, fights back hard in compute workloads and runs cooler on paper thanks to a lower thermal design envelope. Which one deserves a slot in an older build depends almost entirely on what the machine is asked to do: raw graphics performance favors the NVIDIA card, general GPU compute favors the AMD one.

The Verdict

The recorded data paints a split decision. In the 3DMark Steel Nomad DirectX 12 test, the GTX 980 Ti scores 1321 against the RX 590's 1081, a 22.2 percent advantage for the Maxwell flagship. In Geekbench Metal, the situation reverses dramatically: the RX 590 posts 48407 against 19520 for the GTX 980 Ti, a 59.7 percent landslide for AMD. Each card wins one of the two head-to-head contests, so the tiebreaker comes down to use case.

For gaming, the GTX 980 Ti is the stronger pick according to these measurements. Its database percentile against all GPUs is 73, three points above the RX 590's 70, and its average benchmark score of 28020 exceeds the RX 590's 24744. It also brings a wider 384-bit memory bus, 336.6 GB/s of bandwidth against 256.0 GB/s, and a much higher pixel fill rate at 103.3 GPixel/s versus 49.44 GPixel/s. Cards that depend on pushing large numbers of pixels per frame benefit directly from that kind of disparity.

For compute, machine learning-adjacent hobbyist work, or anything that leans on Metal performance, the RX 590 is clearly the better tool. Its Metal score is nearly two and a half times higher, and its FP16 throughput matches its FP32 rate at 7.119 TFLOPS with a 1:1 ratio, a feature the Maxwell card lacks entirely (its FP16 figure is not recorded). It also carries 8 GB of GDDR5 rather than 6 GB, which matters for workloads with large working sets.

Power and platform considerations also differ. The GTX 980 Ti is rated at 250 W with one 6-pin and one 8-pin connector and a suggested 600 W power supply. The RX 590 draws 175 W from a single 8-pin connector and calls for a 450 W supply. Builders with modest power delivery should note that gap. Both cards are dual-slot, PCIe 3.0 x16 parts, and both are end-of-life, so availability is limited to the second-hand channel for either.

FAQ

Q: Which card is faster overall in the database?

A: The GTX 980 Ti holds the higher average benchmark score, 28020 versus 24744, and ranks in the 73rd percentile against all GPUs compared to the RX 590's 70th percentile. The margin is real but modest.

Q: Which card wins in gaming benchmarks?

A: The GTX 980 Ti wins the 3DMark Steel Nomad DirectX 12 test by 22.2 percent, 1321 to 1081. This is the only head-to-head gaming result recorded between the two.

Q: Which card is better for GPU compute work?

A: The RX 590, decisively. Its Geekbench Metal score of 48407 beats the GTX 980 Ti's 19520 by 59.7 percent. The AMD card also matches its FP16 throughput to its FP32 rate at 7.119 TFLOPS, while the GTX 980 Ti records no FP16 figure.

Q: How do their power requirements differ?

A: The GTX 980 Ti is rated for 250 W and requires one 6-pin plus one 8-pin connector, with a suggested 600 W power supply. The RX 590 is rated for 175 W, needs a single 8-pin connector, and pairs with a suggested 450 W supply.

Q: Which card has more memory?

A: The RX 590, with 8 GB of GDDR5 on a 256-bit bus delivering 256.0 GB/s. The GTX 980 Ti has 6 GB of GDDR5, but its wider 384-bit bus delivers more bandwidth at 336.6 GB/s.

Q: How do these cards compare to other GPUs in the database?

A: The GTX 980 Ti's average score sits within a fraction of a percent of the AMD Radeon Pro W5500X (27973, 0.2 percent delta), the AMD FirePro S7150 (28117, -0.3 percent), the AMD Radeon RX 7800M (27883, 0.5 percent) and the AMD Radeon Pro Vega 20 (27839, 0.7 percent). The RX 590 clusters near the NVIDIA RTX A5000 Mobile (24763, -0.1 percent), Intel Arc A350M (24647, 0.4 percent), AMD Radeon RX 6600 XT (24442, 1.2 percent) and NVIDIA GeForce GTX 1630 (24277, 1.9 percent).

Architecture Differences

These two cards come from fundamentally different design philosophies separated by three years. The GTX 980 Ti arrived in June 2015 as part of the GeForce 900 generation, built on the Maxwell 2.0 architecture and the GM200 chip. It was fabricated at TSMC on a 28 nm process, packing 8,000 million transistors into a very large 601 mm² die. That works out to a transistor density of 13.3M per mm², a figure that reflects the older node it was built on.

The RX 590 followed in November 2018 as a Polaris 30 part of the GCN 4.0 architecture, the RX 500 generation. GlobalFoundries manufactured it on a 12 nm process, squeezing 5,700 million transistors into a far smaller 232 mm² die. Its density of 24.6M per mm² is nearly double the NVIDIA card's, illustrating how much the process gap mattered. Despite the smaller die and fewer transistors, the Polis chip actually posts the higher FP32 throughput at 7.119 TFLOPS against 6.060 TFLOPS, thanks largely to its much higher clock speeds.

Both cards predate hardware ray tracing and AI acceleration, so neither has RT cores or tensor cores of any kind. Their feature support diverges in the details: the GTX 980 Ti supports DirectX 12 at feature level 12_1 while the RX 590 stops at 12_0, but the AMD card counters with DisplayPort 1.4a outputs against the NVIDIA card's DisplayPort 1.2, plus a slightly newer Vulkan target at 1.3 versus 1.4 recorded on the NVIDIA side. Both support OpenGL 4.6 and HDMI 2.0-class output, though the AMD card specifies HDMI 2.0b.

Lineage also differs. The GTX 980 Ti succeeded the GeForce 700 series and was itself followed by GeForce 10. The RX 590's predecessor line is listed as Arctic Islands, with Vega as the successor generation.

Specification Differences

The specification sheet is where the philosophical split becomes concrete. The GTX 980 Ti runs at a 1000 MHz base and 1076 MHz boost clock; the RX 590 runs considerably faster at 1469 MHz base and 1545 MHz boost. Memory clocks follow the same pattern: 1753 MHz (7 Gbps effective) for the NVIDIA card, 2000 MHz (8 Gbps effective) for the AMD one.

The NVIDIA card compensates with sheer width. It fields 2816 shading units, 176 texture mapping units, and 96 render output units. The RX 590 counters with 2304 shading units, 144 TMUs, and just 32 ROPs, less than half the NVIDIA card's count in the last category. That ROP deficit is the clearest technical explanation for the pixel rate gap: 103.3 GPixel/s for the GTX 980 Ti versus 49.44 GPixel/s for the RX 590. Texture rate goes the other way, with the RX 590 at 222.5 GTexel/s ahead of the GTX 980 Ti's 189.4 GTexel/s, a direct consequence of its higher clocks across its TMU array.

Memory is a trade rather than a hierarchy: 6 GB on a 384-bit bus with 336.6 GB/s of bandwidth for NVIDIA, 8 GB on a 256-bit bus with 256.0 GB/s for AMD. Physical dimensions favor the smaller AMD card, which measures 241 mm long against the GTX 980 Ti's 267 mm, and its height and width are not recorded in the database. The NVIDIA card spans 111 mm tall and 40 mm wide. Display output counts also differ: the GTX 980 Ti offers one DVI, one HDMI 2.0 and three DisplayPort 1.2 connectors, while the RX 590 drops DVI entirely in favor of one HDMI 2.0b and three DisplayPort 1.4a connectors. Both had launch MSRPs recorded at 649 USD for the GTX 980 Ti and 279 USD for the RX 590.

Head-to-Head Benchmarks

Only two direct comparisons exist in the recorded data, and they point in opposite directions.

The GTX 980 Ti's marquee result is the 3DMark Steel Nomad DirectX 12 test, where it scores 1321 to the RX 590's 1081. That 22.2 percent margin is consistent with the card's hardware advantages in bandwidth and fill rate: a 384-bit bus feeding 336.6 GB/s, plus 96 ROPs generating 103.3 GPixel/s. Steel Nomad is a modern, demanding DirectX 12 workload, and the Maxwell flagship's raw rasterization resources carry it clearly. This single win, combined with the higher aggregate scores, is why the GTX 980 Ti sits at the 73rd percentile versus the RX 590's 70th.

The RX 590's answer is overwhelming. In Geekbench Metal it scores 48407 while the GTX 980 Ti manages only 19520, a 59.7 percent deficit for the NVIDIA card. The gap reflects both architecture and throughput: GCN 4.0's compute orientation, the Polaris chip's 7.119 TFLOPS of FP32 against 6.060 TFLOPS, and native 1:1 FP16 support that Maxwell 2.0 simply does not offer in this configuration. For any workload that resembles Metal's compute profile, the RX 590 is in a different class.

The broader database context tempers both conclusions. The GTX 980 Ti's average score of 28020 places it essentially dead even with the AMD Radeon Pro W5500X (27973, 0.2 percent), the AMD FirePro S7150 (28117, -0.3 percent), the AMD Radeon RX 7800M (27883, 0.5 percent), and the AMD Radeon Pro Vega 20 (27839, 0.7 percent). The RX 590's 24744 average puts it alongside the NVIDIA RTX A5000 Mobile (24763, -0.1 percent), the Intel Arc A350M (24647, 0.4 percent), the AMD Radeon RX 6600 XT (24442, 1.2 percent), and the NVIDIA GeForce GTX 1630 (24277, 1.9 percent).

The final tally: one win each. Gaming performance and overall standing favor the GTX 980 Ti; compute throughput, memory capacity, and efficiency of power delivery favor the RX 590. Both are end-of-life hardware whose launch MSRPs (649 USD and 279 USD respectively) framed very different market positions, and both remain capable of the workloads their benchmark profiles indicate.