AMD Radeon RX 6900 XT vs NVIDIA GeForce RTX 5080 Mobile Comparison

The recorded data frames this as an unusual pairing: a triple-slot desktop flagship against a laptop-integrated part from a much later generation. The AMD Radeon RX 6900 XT, an end-of-life desktop card released at a launch MSRP of 999 USD, meets the actively produced NVIDIA GeForce RTX 5080 Mobile. Both sit in the same overall percentile band in the database, 86 for the AMD card and 81 for the NVIDIA part, yet the head-to-head record splits decisively toward NVIDIA with 7 wins out of 10 shared tests, while AMD takes 3. The interesting story is not who wins more often, but where and why each one wins.

Head-to-Head Benchmarks

The single largest gap in the shared test set belongs to AMD. In Passmark GPU Compute, the RX 6900 XT scores 14547 against 12134 for the RTX 5080 Mobile, a 19.9 percent lead for the Radeon. That is a raw compute result, and it lines up with the Radeon's much higher throughput figures elsewhere in the database: a texture rate of 720.0 GTexel/s versus 360.0 GTexel/s, and a pixel rate of 288.0 GPixel/s versus 144.0 GPixel/s. Both GPUs post an identical 23.04 TFLOPS FP32 rating, but the AMD part doubles its FP16 output to 46.08 TFLOPS on a 2:1 ratio while the NVIDIA part runs FP16 at 1:1 for 23.04 TFLOPS.

AMD also wins the two API tests that favor its strengths. Geekbench OpenCL goes to the RX 6900 XT at 187673 versus 166986, a 12.4 percent margin, and Passmark DirectX 11 lands at 279 versus 253, a 10.3 percent advantage for the Radeon.

Everything else goes to the RTX 5080 Mobile, and some of it by wide margins. The biggest is 3DMark Steel Nomad (DX12), where the mobile Blackwell part scores 4952 against 4079, a 17.6 percent win for NVIDIA. Geekbench Vulkan follows the same pattern: 169754 versus 148525, a 12.5 percent NVIDIA lead. Passmark DirectX 9 is a 14.6 percent win for the RTX 5080 Mobile at 314 versus 268. The remaining NVIDIA wins are narrow: DirectX 10 by 2.3 percent (171 versus 167), DirectX 12 by 2.6 percent (116 versus 113), G2D by 3.7 percent (1095 versus 1055), and the headline Passmark G3D score by 3.5 percent (27711 versus 26732).

The pattern is clear. The RTX 5080 Mobile dominates modern graphics workloads and synthetic gaming benchmarks, while the RX 6900 XT retains an edge in legacy DirectCompute-style APIs and raw compute throughput.

Architecture Differences

These two parts come from different eras and different segments, and the silicon reflects it. The RX 6900 XT is the Navi 21 die on AMD's RDNA 2.0 architecture, built by TSMC on a 7 nm process. It packs 26,800 million transistors into a 520 mm² die, working out to a density of 51.5M per mm². The RTX 5080 Mobile uses the GB203 chip on NVIDIA's Blackwell 2.0 architecture, manufactured on TSMC's 5 nm node. It fits 45,600 million transistors into just 378 mm², a density of 120.6M per mm², more than double the Radeon's transistor density.

The shader configuration is inverted relative to their era. The Radeon carries 5120 shading units, 320 TMUs, and 128 ROPs, while the NVIDIA part fields 7680 shading units but only 240 TMUs and 96 ROPs. AMD lists 80 RT cores and no tensor cores; NVIDIA lists 60 RT cores and 240 tensor cores, the hardware behind its AI acceleration features and a key structural difference between the two designs.

Memory is another generational split. Both use a 256-bit bus and 16 GB of VRAM, but the Radeon pairs GDDR6 at an effective 16 Gbps for 512.0 GB/s of bandwidth, while the mobile Blackwell part runs GDDR7 at 28 Gbps effective for 896.0 GB/s, a 75 percent bandwidth advantage for NVIDIA.

Clocks and power tell the desktop-versus-laptop story. The RX 6900 XT runs a 1825 MHz base, a 2015 MHz game clock, and a 2250 MHz boost, fed by 2x 8-pin connectors under a 300 W TDP with a suggested 700 W PSU. The RTX 5080 Mobile runs 975 MHz base and 1500 MHz boost under an 80 W TDP with no power connectors at all, occupying an IGP slot. The Radeon is a triple-slot card measuring 267 mm long, 120 mm tall, and 50 mm wide; the NVIDIA part has no fixed dimensions since it depends on the host device. Connectivity also differs: PCIe 4.0 x16 on AMD versus PCIe 5.0 x16 on NVIDIA, and the Radeon offers fixed outputs (1x HDMI 2.1, 2x DisplayPort 1.4a, 1x USB Type-C) where the mobile part's outputs are device dependent. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Where Each One Wins

For modern gaming and forward-looking graphics workloads, the recorded data points to the RTX 5080 Mobile. It wins the flagship shared gaming test, 3DMark Steel Nomad, by 17.6 percent, plus the composite Passmark G3D score by 3.5 percent and Geekbench Vulkan by 12.5 percent. Despite operating at an 80 W TDP, its memory subsystem and Blackwell architecture carry those results.

The RX 6900 XT is the pick for compute-heavy and legacy-API workloads. Its 19.9 percent win in Passmark GPU Compute and 12.4 percent win in Geekbench OpenCL show where RDNA 2.0's raw throughput still holds up, and its 10.3 percent lead in Passmark DirectX 11 favors older game engines and applications built on that API. Its doubled FP16 rate, 46.08 TFLOPS versus 23.04 TFLOPS, also matters for workloads that can exploit it.

Context matters too. The database places the RX 6900 XT in the 86th percentile against all GPUs with an average benchmark score of 50951, ahead of rivals such as the NVIDIA CMP 50HX (51790 average, within 1.6 percent), the AMD Radeon RX Vega 64 (50001, 1.9 percent behind), the NVIDIA GeForce RTX 5070 Ti (49957, 2 percent behind), and the Intel Arc A550M (49737, 2.4 percent behind). The RTX 5080 Mobile sits in the 81st percentile with an average of 38349, essentially tied with the NVIDIA GeForce MX570 (38299, 0.1 percent), the RTX 4080 Mobile (38135, 0.6 percent behind), the MX570 A (38691, 0.9 percent ahead), and the AMD Radeon Pro 580X (38706, 0.9 percent ahead). So while NVIDIA wins the direct head-to-head 7-3, the Radeon's broader database standing and average score are higher.

FAQ

Q: Which GPU wins more head-to-head benchmark tests?

A: The RTX 5080 Mobile wins 7 of the 10 shared tests. The RX 6900 XT takes 3, all in compute and legacy-API tests.

Q: What is the biggest win for each side?

A: The RX 6900 XT's largest margin is 19.9 percent in Passmark GPU Compute (14547 versus 12134). The RTX 5080 Mobile's largest margin is 17.6 percent in 3DMark Steel Nomad DX12 (4952 versus 4079).

Q: How do their memory subsystems compare?

A: Both have 16 GB on a 256-bit bus, but the Radeon uses GDDR6 for 512.0 GB/s while the NVIDIA part uses GDDR7 for 896.0 GB/s.

Q: How much power does each consume?

A: The RX 6900 XT has a 300 W TDP with 2x 8-pin connectors and a suggested 700 W PSU. The RTX 5080 Mobile has an 80 W TDP and no power connectors.

Q: Are both cards still in production?

A: No. The RX 6900 XT is listed as end-of-life, released on 2020-10-27. The RTX 5080 Mobile is active, released on 2025-04-01.

Q: Which ranks higher across the full GPU database?

A: The RX 6900 XT, at the 86th percentile with an average score of 50951, versus the 81st percentile and 38349 for the RTX 5080 Mobile.

The Verdict

The data supports two different recommendations depending on the buyer. For a desktop user who needs raw compute throughput, OpenCL performance, or strong DirectX 11 results, the RX 6900 XT is the stronger card: a 19.9 percent compute lead, a 12.4 percent OpenCL lead, a higher database percentile, and an average benchmark score of 50951 versus 38349. Its 16 GB of VRAM matches the newer part, and its launch MSRP of 999 USD reflects the flagship tier it occupied.

For anyone prioritizing modern gaming performance, Vulkan, or efficiency, the RTX 5080 Mobile is the better fit despite its 80 W envelope. It wins the flagship Steel Nomad test by 17.6 percent, the composite G3D score, and every DirectX generation tested, backed by 896.0 GB/s of GDDR7 bandwidth and 240 tensor cores the Radeon simply lacks. It is also the only one of the two still in active production.

In short: the RX 6900 XT wins on compute muscle and database standing, the RTX 5080 Mobile wins on modern graphics performance and efficiency, and the 7-3 head-to-head record correctly signals that the generational and architectural gap favors NVIDIA in the workloads most users run today.