NVIDIA GeForce RTX 3070 vs NVIDIA GeForce RTX 5090 D V2 Comparison

# NVIDIA GeForce RTX 3070 vs NVIDIA GeForce RTX 5090 D V2

The NVIDIA GeForce RTX 3070 and NVIDIA GeForce RTX 5090 D V2 represent two distinct generations of NVIDIA hardware, separated by a massive architectural leap. The RTX 3070, an 8 nm Ampere part from 2020, sits at the 61st percentile of all GPUs, while the RTX 5090 D V2, a 5 nm Blackwell 2.0 flagship from 2025, holds the 59th percentile. Despite the newer card's lower percentile ranking, the head-to-head benchmark data tells a far more dramatic story: in 3DMark Steel Nomad DX12, the RTX 5090 D V2 scores 16,504 versus the RTX 3070's 3,162, a staggering 80.8% delta in favor of the newer card. The RTX 5090 D V2's average benchmark score of 16,504 stands against the RTX 3070's 17,208 average, though this comparison is skewed by the RTX 3070 having ten benchmark entries versus only one for its rival. This analysis digs into the architectural, specification, and benchmark data to understand what separates these two very different products.

FAQ

Q: Which card is faster in 3DMark Steel Nomad DX12?

A: The NVIDIA GeForce RTX 5090 D V2 wins decisively, scoring 16,504 compared to the RTX 3070's 3,162, a delta of -80.8% from the RTX 3070's perspective.

Q: How do their average benchmark scores compare?

A: The RTX 3070 has a higher average benchmark score of 17,208 across ten tests, while the RTX 5090 D V2 averages 16,504 from a single test. However, this is not an apples-to-apples comparison given the different test suites.

Q: What are the transistor counts and process nodes?

A: The RTX 3070 uses 17,400 million transistors on Samsung's 8 nm process, while the RTX 5090 D V2 packs 92,200 million transistors on TSMC's 5 nm node. The newer card's transistor density is 122.9M per mm² versus 44.4M per mm² for the older card.

Q: How much memory does each card have?

A: The RTX 3070 comes with 8 GB of GDDR6 on a 256-bit bus, delivering 448.0 GB/s bandwidth. The RTX 5090 D V2 features 24 GB of GDDR7 on a 384-bit bus, providing 1.34 TB/s bandwidth.

Q: What are the TDP and PSU requirements?

A: The RTX 3070 has a 220 W TDP and requires a 550 W suggested PSU. The RTX 5090 D V2 has a 575 W TDP and demands a 950 W suggested PSU.

Q: Which card is still in production?

A: The RTX 5090 D V2 is marked as "Active" in production status, while the RTX 3070 is "End-of-life." The RTX 5090 D V2 launched on 2025-08-14, whereas the RTX 3070 launched on 2020-08-31.

Architecture Differences

The architectural gap between these two GPUs is immense. The RTX 3070 is built on the GA104 chip using NVIDIA's Ampere architecture, fabricated on Samsung's 8 nm process. It houses 17,400 million transistors on a 392 mm² die, yielding a density of 44.4M transistors per mm². In contrast, the RTX 5090 D V2 is based on the GB202 chip with Blackwell 2.0 architecture, produced on TSMC's 5 nm node. It contains 92,200 million transistors across a 750 mm² die, achieving 122.9M transistors per mm². This represents a 5.3x increase in raw transistor count and a 2.77x improvement in transistor density.

The compute resources differ dramatically. The RTX 3070 features 5,888 shading units, 184 TMUs, 96 ROPs, 46 RT cores, and 184 tensor cores. The RTX 5090 D V2 scales this up to 21,760 shading units, 680 TMUs, 176 ROPs, 170 RT cores, and 680 tensor cores. That is 3.7x more shading units, 3.7x more TMUs, 1.83x more ROPs, 3.7x more RT cores, and 3.7x more tensor cores. The FP32 compute throughput jumps from 20.31 TFLOPS to 104.8 TFLOPS, a 5.16x increase, with FP16 also at 1:1 ratios for both cards.

Memory architecture is another fundamental divergence. The RTX 3070 uses 8 GB of GDDR6 with a 256-bit bus and 448.0 GB/s bandwidth. The RTX 5090 D V2 moves to 24 GB of GDDR7 on a 384-bit bus, delivering 1.34 TB/s — exactly 3x the memory capacity and roughly 2.99x the bandwidth. The pixel rate improves from 165.6 GPixel/s to 423.6 GPixel/s, and the texture rate jumps from 317.4 GTexel/s to 1,636.8 GTexel/s.

The process node transition from 8 nm to 5 nm is critical, but the architecture shift from Ampere to Blackwell 2.0 brings its own changes. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so API coverage is identical. However, the bus interface moves from PCIe 4.0 x16 to PCIe 5.0 x16, and display outputs upgrade from 1x HDMI 2.1 plus 3x DisplayPort 1.4a to 1x HDMI 2.1b plus 3x DisplayPort 2.1b.

Head-to-Head Benchmarks

The only shared benchmark between these two cards is 3DMark Steel Nomad DX12, and the results are lopsided. The RTX 5090 D V2 scores 16,504, while the RTX 3070 manages 3,162. The delta is -80.8% from the RTX 3070's perspective, meaning the RTX 5090 D V2 outperforms it by roughly 5.22x. This is a decisive victory for the newer card, and the single head-to-head result gives the RTX 5090 D V2 a clean 1-0 win tally.

Looking at the RTX 3070's broader benchmark portfolio provides context. Its best scores include 112,821 in Geekbench OpenCL, 22,214 in Passmark G3D, and 21,022 in Geekbench Vulkan. The RTX 5090 D V2 has no comparable results in these tests, so a wider performance picture cannot be drawn from direct data. The RTX 3070's 3DMark Steel Nomad score of 3,162 is its lowest benchmark result by a wide margin, which suggests this particular test may be particularly demanding or that the card's architecture is not optimized for it.

The RTX 5090 D V2's nearest rivals in the database are curious: NVIDIA T400 (16,508 avg, 0% delta), AMD Radeon PRO W7500 (16,415 avg, 0.5% delta), NVIDIA RTX PRO 6000 Blackwell (16,408 avg, 0.6% delta), and AMD Radeon RX 5700 XT (16,361 avg, 0.9% delta). These are all within 0.9% of the RTX 5090 D V2's average score, which is surprising given the massive hardware differences. The RTX 3070's nearest rivals include AMD Radeon RX 7600 XT (17,083 avg, 0.7% delta), NVIDIA GeForce GTX 690 (17,037 avg, 1% delta), AMD Radeon HD 7970M (17,019 avg, 1.1% delta), and NVIDIA Tesla K40c (17,468 avg, -1.5% delta). The RTX 3070's rivals are all older or lower-tier cards, while the RTX 5090 D V2's rivals include a mix of professional and consumer parts.

Specification Differences

The specification tables reveal stark contrasts in nearly every measurable category:

• Process Node: 8 nm (Samsung) vs 5 nm (TSMC)
• Transistors: 17,400 million vs 92,200 million
• Die Size: 392 mm² vs 750 mm²
• Transistor Density: 44.4M / mm² vs 122.9M / mm²
• Base Clock: 1500 MHz vs 2017 MHz
• Boost Clock: 1725 MHz vs 2407 MHz
• Memory Speed: 1750 MHz (14 Gbps effective) vs 1750 MHz (28 Gbps effective)
• Memory Size: 8 GB vs 24 GB
• Memory Type: GDDR6 vs GDDR7
• Memory Bus: 256 bit vs 384 bit
• Memory Bandwidth: 448.0 GB/s vs 1.34 TB/s
• Shading Units: 5,888 vs 21,760
• TMUs: 184 vs 680
• ROPs: 96 vs 176
• RT Cores: 46 vs 170
• Tensor Cores: 184 vs 680
• Pixel Rate: 165.6 GPixel/s vs 423.6 GPixel/s
• Texture Rate: 317.4 GTexel/s vs 1,636.8 GTexel/s
• FP32: 20.31 TFLOPS vs 104.8 TFLOPS
• FP16: 20.31 TFLOPS vs 104.8 TFLOPS
• TDP: 220 W vs 575 W
• Power Connectors: 1x 12-pin vs 1x 16-pin
• Suggested PSU: 550 W vs 950 W
• Bus Interface: PCIe 4.0 x16 vs PCIe 5.0 x16
• Display Outputs: 1x HDMI 2.1, 3x DisplayPort 1.4a vs 1x HDMI 2.1b, 3x DisplayPort 2.1b
• Length: 242 mm (9.5 inches) vs 304 mm (12 inches)
• Height: 112 mm (4.4 inches) vs 137 mm (5.4 inches)
• Width: Not specified vs 48 mm (1.9 inches)
• Release Date: 2020-08-31 vs 2025-08-14
• Production Status: End-of-life vs Active

The RTX 5090 D V2 is physically larger in every dimension, consumes 2.61x more power, and requires a substantially more robust PSU. Both cards are dual-slot designs, but the newer card is 62 mm longer and 25 mm taller.

Where Each One Wins

The RTX 5090 D V2 wins the only direct head-to-head benchmark, 3DMark Steel Nomad DX12, by an 80.8% margin. It also wins on every raw specification: compute throughput, memory capacity and bandwidth, pixel and texture rates, and core counts. The RTX 5090 D V2 is clearly the card for scenarios demanding maximum performance in modern DX12 workloads, as evidenced by its Steel Nomad result.

The RTX 3070 wins in the broader benchmark average comparison, posting 17,208 versus 16,504 for the newer card. This is partly because the RTX 3070 has ten benchmark entries, including strong showings in Geekbench OpenCL (112,821) and Passmark G3D (22,214), while the RTX 5090 D V2 has only the single Steel Nomad result. The RTX 3070 also wins on efficiency in a relative sense: its 220 W TDP is far lower than the 575 W TDP of the RTX 5090 D V2, and its smaller physical footprint (242 mm versus 304 mm length) makes it easier to fit in compact cases.

For users with existing 550 W PSUs or smaller cases, the RTX 3070 is the more practical choice from a system integration standpoint. The RTX 3070's percentile rank of 61 versus the RTX 5090 D V2's 59 also suggests the older card holds up well across its benchmark suite, though this is again influenced by test count disparity.

The Verdict

The data is unambiguous about raw performance: the NVIDIA GeForce RTX 5090 D V2 is in a completely different performance class. Its 5.22x advantage in 3DMark Steel Nomad DX12, combined with 5.16x higher FP32 throughput, 3x more memory, and 2.99x more bandwidth, makes it the clear choice for anyone prioritizing maximum frame rates in modern DX12 titles. The 170 RT cores versus 46 and 680 tensor cores versus 184 suggest substantially stronger ray tracing and AI-assisted rendering capabilities.

However, the RTX 3070 is not without merit in this comparison. Its higher average benchmark score of 17,208, achieved across a diverse ten-test suite, indicates strong all-around performance in legacy and compute workloads. Its 220 W TDP and 550 W PSU requirement make it far easier to integrate into existing systems, and its smaller physical size expands case compatibility. The RTX 3070's 61st percentile versus the RTX 5090 D V2's 59th percentile is a statistical curiosity that likely reflects the newer card's limited benchmark data rather than a true performance reversal.

The choice between these two cards hinges on what the data prioritizes. For a user building a new high-end system with the PSU headroom and case space to accommodate a 575 W, 304 mm-long card, the RTX 5090 D V2 is the superior performer in the only test where they directly compete. For a user upgrading an existing system or working within tighter power and space constraints, the RTX 3070 offers respectable performance with far lower system requirements. The RTX 3070's end-of-life status and the RTX 5090 D V2's active production status further tilt the long-term availability argument toward the newer card. The benchmark results indicate the RTX 5090 D V2 is the performance king, but the RTX 3070 remains a viable option for constrained builds.