NVIDIA GeForce RTX 3090 vs NVIDIA GeForce RTX 5070 Comparison

The NVIDIA GeForce RTX 5070 and the NVIDIA GeForce RTX 3090 occupy a fascinating intersection of generations and performance tiers. The RTX 5070, built on the new Blackwell architecture, faces the outgoing flagship-class RTX 3090 from the Ampere generation. The aggregated benchmark data shows a near-total deadlock, with the RTX 5070 holding a razor-thin 0.6% lead in average benchmark score (41687 vs 41441). Both cards sit at the 83rd percentile among all GPUs, and their nearest rivals are nearly identical, including the AMD Radeon Pro 5300 and the NVIDIA Tesla M40. This pairing produces a complex competitive landscape where the winner is determined less by raw aggregate power and more by the specific workload, driver optimization, and API in use.

Where Each One Wins

The head-to-head benchmark data reveals a clear split in strengths. The RTX 5070 takes the majority of wins, securing victory in 6 out of 10 benchmarks, while the RTX 3090 wins the remaining 4. The RTX 5070's dominance is most pronounced in legacy DirectX workloads and compute-oriented tasks, whereas the RTX 3090 shows its strength in newer DirectX 12 and OpenCL environments.

For the RTX 5070, the most significant victories are in Passmark DirectX 11 and DirectX 9, where it leads by 25.9% and 19.4%, respectively. It also wins the Passmark G3D test by 9.4% and the Passmark G2D test by 22.8%. These results suggest the newer architecture handles older API paths with considerably more efficiency.

The RTX 3090, conversely, wins in 3DMark Steel Nomad (DirectX 12), Geekbench OpenCL, Passmark DirectX 10, and Passmark DirectX 12. Its margins are much narrower, ranging from 0.8% to 3.1%. The RTX 3090's win in the 3DMark Steel Nomad test, which is a modern DirectX 12 benchmark, is particularly notable as it indicates the older card retains a slight edge in current-generation gaming workloads.

Architecture Differences

The architectural gap between these two cards is vast, representing two distinct design philosophies and process nodes. The RTX 5070 utilizes the GB205 chip built on the Blackwell 2.0 architecture, manufactured on a 5 nm process at TSMC. In contrast, the RTX 3090 uses the GA102 chip based on the Ampere architecture, produced on an 8 nm process at Samsung. This process node difference is a primary driver of the efficiency and density disparities between the two.

The transistor counts tell a story of density versus raw scale. The RTX 5070 packs 31,100 million transistors into a 263 mm² die, achieving a density of 118.3M / mm². The RTX 3090, despite having fewer transistors at 28,300 million, uses a much larger 628 mm² die, resulting in a density of just 45.1M / mm². This means the RTX 5070 achieves higher transistor density on a smaller chip, while the RTX 3090 relies on a massive, less dense die.

The memory subsystems are also fundamentally different. The RTX 5070 features 12 GB of GDDR7 memory on a 192-bit bus, providing 672.0 GB/s of bandwidth. The RTX 3090 counters with a massive 24 GB of GDDR6X on a 384-bit bus, delivering 936.2 GB/s of bandwidth. This gives the RTX 3090 a significant 39.3% advantage in memory bandwidth and double the capacity. The RTX 5070 compensates with higher effective memory clocks at 28 Gbps versus the RTX 3090's 19.5 Gbps.

Core configurations also diverge sharply. The RTX 3090 has more shading units (10496 vs 6144), more TMUs (328 vs 192), more ROPs (112 vs 80), more RT cores (82 vs 48), and more tensor cores (328 vs 192). Despite this, the RTX 5070 achieves a higher pixel rate (201.0 GPixel/s vs 189.8 GPixel/s), while the RTX 3090 maintains a lead in texture rate (556.0 GTexel/s vs 482.3 GTexel/s). The RTX 5070 also posts a higher FP32 performance of 30.87 TFLOPS compared to the RTX 3090's 35.58 TFLOPS, though the older card still leads in raw compute.

Head-to-Head Benchmarks

The benchmark results are close enough that small deltas define the narrative. In 3DMark Steel Nomad (DX12), the RTX 3090 wins by a hair, scoring 5118 against the RTX 5070's 5077, a delta of just -0.8%. This suggests that for the latest DirectX 12 titles, the two cards are virtually interchangeable, with the RTX 3090 holding a marginal edge.

The Geekbench tests show a similar pattern. In OpenCL, the RTX 3090 wins with 191142 points versus 185269 for the RTX 5070, a 3.1% difference. However, in Vulkan, the tables turn, with the RTX 5070 winning 179413 to 174310, a 2.9% lead. This indicates that the RTX 5070 has better optimizations for the Vulkan API, while the RTX 3090 retains an advantage in OpenCL compute.

The Passmark suite provides the clearest separation. The RTX 5070 dominates in DirectX 11 with a score of 277 versus 220, a massive 25.9% advantage. It also leads in DirectX 9 (320 vs 268, +19.4%) and G3D (29137 vs 26645, +9.4%). The G2D test also goes to the RTX 5070 (1305 vs 1063, +22.8%). The RTX 3090 only manages narrow wins in DirectX 10 (182 vs 180, +1.1%) and DirectX 12 (110 vs 108, +1.8%). In GPU Compute, the RTX 5070 wins again, scoring 15787 against 15356, a 2.8% lead.

These results paint a picture of the RTX 5070 being significantly more efficient in older APIs and general compute, while the RTX 3090 holds a slight edge in the most modern DirectX 12 path and OpenCL. The overall win count of 6 for the RTX 5070 versus 4 for the RTX 3090 is driven by the sheer magnitude of the wins in legacy and compute workloads, which outweigh the narrow margins of the RTX 3090's victories.

FAQ

Q: Which card has a higher average benchmark score?

A: The NVIDIA GeForce RTX 5070 has a marginally higher average benchmark score of 41687, compared to the RTX 3090's 41441, representing a 0.6% advantage for the newer card.

Q: Is the RTX 3090 better for DirectX 12 gaming?

A: Data from the 3DMark Steel Nomad (DX12) and Passmark DirectX 12 tests indicates the RTX 3090 holds a slight lead. It scores 5118 in Steel Nomad versus 5077 for the RTX 5070, and 110 in Passmark DX12 versus 108.

Q: Where does the RTX 5070 show its biggest advantage?

A: The RTX 5070 demonstrates its largest wins in legacy API benchmarks. It leads by 25.9% in Passmark DirectX 11 and by 19.4% in Passmark DirectX 9, indicating significantly better performance in older titles.

Q: How do the memory configurations compare?

A: The RTX 3090 has a substantial advantage with 24 GB of GDDR6X memory on a 384-bit bus, offering 936.2 GB/s of bandwidth. The RTX 5070 has 12 GB of GDDR7 on a 192-bit bus, providing 672.0 GB/s.

Q: What are the power requirements for each card?

A: The RTX 5070 has a TDP of 250 W and a suggested PSU of 600 W, while the RTX 3090 has a higher TDP of 350 W and requires a suggested 750 W PSU.

Q: Which card is more efficient in compute workloads?

A: The RTX 5070 leads in the Passmark GPU Compute test, scoring 15787 versus the RTX 3090's 15356, a 2.8% advantage. However, the RTX 3090 wins in Geekbench OpenCL by 3.1%.

Specification Differences

The specification sheets reveal stark contrasts across nearly every major component. The process node is a fundamental difference: the RTX 5070 uses a 5 nm process at TSMC, while the RTX 3090 uses an 8 nm process at Samsung. This leads to the RTX 5070's smaller die size of 263 mm² versus the RTX 3090's 628 mm², despite the newer card having more transistors (31,100 million vs 28,300 million).

Memory is another major divider. The RTX 5070 offers 12 GB of GDDR7 on a 192-bit bus with 672.0 GB/s bandwidth. The RTX 3090 offers double the capacity at 24 GB of GDDR6X on a 384-bit bus, delivering 936.2 GB/s. The clock speeds also differ, with the RTX 5070 running at a base of 2325 MHz and boost of 2512 MHz, versus the RTX 3090's much lower base of 1395 MHz and boost of 1695 MHz.

The core counts are heavily in favor of the RTX 3090. It has 10496 shading units, 328 TMUs, 112 ROPs, 82 RT cores, and 328 tensor cores. The RTX 5070 has 6144 shading units, 192 TMUs, 80 ROPs, 48 RT cores, and 192 tensor cores. Despite this, the RTX 5070 achieves a higher pixel rate (201.0 GPixel/s vs 189.8 GPixel/s), while the RTX 3090 leads in texture rate (556.0 GTexel/s vs 482.3 GTexel/s).

Power and physical dimensions also differ substantially. The RTX 5070 has a TDP of 250 W, is Dual-slot, uses a 1x 16-pin connector, and requires a 600 W PSU. The RTX 3090 has a TDP of 350 W, is Triple-slot, uses a 1x 12-pin connector, and requires a 750 W PSU. The RTX 3090 is also significantly larger at 336 mm in length versus 245 mm for the RTX 5070. The bus interface differs as well, with the RTX 5070 using PCIe 5.0 x16 and the RTX 3090 using PCIe 4.0 x16. Display outputs also vary, with the RTX 5070 offering 1x HDMI 2.1b and 3x DisplayPort 2.1b, while the RTX 3090 offers 1x HDMI 2.1 and 3x DisplayPort 1.4a.