AMD Radeon RX 6800M vs NVIDIA GeForce RTX 3090 Comparison

Head-to-Head Benchmarks

The benchmark data presents a lopsided contest, with the NVIDIA GeForce RTX 3090 winning nine of the ten head-to-head comparisons. The most decisive margin comes in Passmark GPU Compute, where the RTX 3090 scores 15,356 against the RX 6800M's 5,032 — a 67.2% advantage. This reflects the substantial gap in raw compute throughput between the two designs. The RTX 3090 also dominates the modern DirectX 12 workload in 3DMark Steel Nomad, posting 5,118 points versus 2,238 for the AMD part, a 56.3% deficit for the RX 6800M. Across the legacy DirectX tests, the NVIDIA card maintains consistent leads: 44.5% in DirectX 10 (182 vs 101), 42.3% in DirectX 11 (220 vs 127), 40.9% in DirectX 12 (110 vs 65), and 45.1% in DirectX 9 (268 vs 147). The 2D performance gap is nearly identical at 49.4%, with the RTX 3090 scoring 1,063 in Passmark G2D versus 538.

The single bright spot for the AMD Radeon RX 6800M is Geekbench Vulkan, where it wins decisively by 75.7% — scoring 94,766 against the RTX 3090's 53,927. This is a remarkable outlier, suggesting the RDNA 2 architecture's Vulkan implementation is exceptionally well-optimized in this particular workload, or that the test environment favors the mobile part's driver stack. The OpenCL result tells a different story: the RTX 3090 leads 172,758 to 87,621, a 49.3% margin that aligns more closely with the overall pattern of NVIDIA dominance.

The aggregate averages reflect this split. The RTX 3090's average benchmark score of 27,565 places it near its nearest rivals — the RTX 4070 Mobile at 27,435 (0.5% ahead) and the RX 6700 XT at 27,425 (also 0.5% behind). The RX 6800M's average of 28,874 puts it in a similar percentile band, just 0.4% ahead of the RX 570 and 1% ahead of the R9 M295X, while trailing the Intel Arc A370M by 1%. Despite the RTX 3090 winning nine of ten direct comparisons, both parts sit at nearly identical percentiles versus all GPUs — 74th for AMD and 73rd for NVIDIA — indicating that the benchmark suite's weighting masks the head-to-head differential.

Where Each One Wins

The RTX 3090 wins across virtually every traditional GPU workload category: rasterization, compute, and legacy API compatibility. The Passmark DirectX suite shows the NVIDIA card is faster in every generation of DirectX, from 9 through 12, with margins ranging from 40.9% to 45.1%. The 3DMark Steel Nomad result confirms that advantage extends to modern DX12 titles as well. For compute-heavy applications — rendering, simulation, or machine learning tasks that leverage OpenCL — the RTX 3090's 67.2% lead in Passmark GPU Compute and 49.3% lead in Geekbench OpenCL make it the clear choice. The 2D performance advantage (49.4%) also suggests better desktop composition and video playback performance.

The RX 6800M's victory in Geekbench Vulkan is the sole exception, but it is a significant one. Vulkan is the API of choice for many modern game engines and cross-platform applications, and a 75.7% lead in this specific test indicates that AMD's RDNA 2 architecture has a substantial optimization advantage in Vulkan-based workloads. This could translate to real-world wins in Vulkan-native games or applications, even if the overall benchmark suite favors NVIDIA. For users prioritizing Vulkan performance specifically, the RX 6800M offers a compelling alternative despite losing the broader benchmark war.

Architecture Differences

The two GPUs represent fundamentally different design philosophies. The AMD Radeon RX 6800M uses the Navi 22 chip built on RDNA 2.0 architecture, fabricated on TSMC's 7 nm process. It contains 17,200 million transistors on a 335 mm² die, yielding a transistor density of 51.3 million per square millimeter. The NVIDIA GeForce RTX 3090 uses the GA102 chip on Ampere architecture, built on Samsung's 8 nm process. It packs 28,300 million transistors into a much larger 628 mm² die, with a lower density of 45.1 million per square millimeter. The RTX 3090's larger die and higher transistor count reflect its desktop-class ambitions, while the RX 6800M's smaller, denser design suits its IGP (integrated graphics processor) form factor.

The compute resource allocation differs dramatically. The RX 6800M features 2,560 shading units, 160 texture mapping units, 64 ROPs, and 40 ray tracing cores. The RTX 3090 counters with 10,496 shading units, 328 TMUs, 112 ROPs, 82 ray tracing cores, and 328 tensor cores. The NVIDIA part also employs a 1:1 FP16 to FP32 ratio — both at 35.58 TFLOPS — while AMD uses a 2:1 ratio, delivering 24.47 TFLOPS FP16 versus 12.24 TFLOPS FP32. The RTX 3090's tensor cores are a notable architectural addition absent from the AMD part, enabling hardware-accelerated AI workloads.

Memory architecture further separates them. The RX 6800M uses 12 GB of GDDR6 on a 192-bit bus, achieving 384.0 GB/s bandwidth. The RTX 3090 uses 24 GB of GDDR6X on a 384-bit bus, more than doubling bandwidth to 936.2 GB/s. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so API feature parity is maintained. The AMD part clocks significantly higher — 2,116 MHz base and 2,390 MHz boost versus 1,395 MHz base and 1,695 MHz boost — but the NVIDIA card's massive core count and memory bandwidth compensate for the lower clocks.

Specification Differences

The power envelope is the most striking differentiator. The RX 6800M has a TDP of 145 W and uses no power connectors, reflecting its integrated mobile design. The RTX 3090 draws 350 W and requires a single 12-pin connector, with a suggested 750 W power supply. The physical form factors diverge equally: the RX 6800M is an IGP (integrated graphics package) with portable-device-dependent display outputs, while the RTX 3090 is a triple-slot card measuring 336 mm in length, 140 mm in height, and 61 mm in width, with 1x HDMI 2.1 and 3x DisplayPort 1.4a outputs.

Memory specifications differ in every measurable dimension: 12 GB vs 24 GB capacity, GDDR6 vs GDDR6X type, 192-bit vs 384-bit bus width, and 384.0 GB/s vs 936.2 GB/s bandwidth. The memory clock also differs — 2000 MHz (16 Gbps effective) for AMD versus 1219 MHz (19.5 Gbps effective) for NVIDIA. The process node differs (7 nm TSMC vs 8 nm Samsung), as do the die dimensions and transistor counts. The RX 6800M launched on 2021-05-30, while the RTX 3090 launched earlier on 2020-08-31. The RTX 3090 has a launch MSRP of 1,499 USD. The RX 6800M's predecessor is Polaris Mobile; the RTX 3090's predecessor is GeForce 20 and its successor is GeForce 40. The RX 6800M's production status is end-of-life, as is the RTX 3090's.

FAQ

Q: Which GPU wins the overall head-to-head benchmark comparison?

A: The NVIDIA GeForce RTX 3090 wins nine of the ten benchmark comparisons, including all Passmark tests and 3DMark Steel Nomad. The AMD Radeon RX 6800M wins only the Geekbench Vulkan test.

Q: How large is the RTX 3090's lead in compute performance?

A: In Passmark GPU Compute, the RTX 3090 scores 15,356 versus the RX 6800M's 5,032, a 67.2% advantage. In Geekbench OpenCL, the lead is 172,758 versus 87,621, a 49.3% margin.

Q: Is there any workload where the RX 6800M is faster?

A: Yes, in Geekbench Vulkan the RX 6800M scores 94,766 against the RTX 3090's 53,927, a 75.7% advantage. This is the only benchmark where AMD wins.

Q: How do the memory subsystems compare?

A: The RTX 3090 offers 24 GB of GDDR6X on a 384-bit bus with 936.2 GB/s bandwidth. The RX 6800M offers 12 GB of GDDR6 on a 192-bit bus with 384.0 GB/s bandwidth.

Q: What are the power consumption differences?

A: The RX 6800M has a 145 W TDP and requires no power connectors. The RTX 3090 has a 350 W TDP, uses a single 12-pin connector, and requires a suggested 750 W power supply.

Q: How do their benchmark averages compare to rival GPUs?

A: The RX 6800M's average score of 28,874 is 0.4% ahead of the RX 570 and 1% ahead of the R9 M295X. The RTX 3090's average of 27,565 is 0.5% ahead of both the RTX 4070 Mobile and RX 6700 XT.