NVIDIA GeForce RTX 3090 vs NVIDIA GeForce RTX 5070 Mobile Comparison

FAQ

Q: Which GPU wins in raw compute performance?

A: The RTX 3090 dominates compute workloads. It leads in Passmark GPU compute by 46.1% (15,356 vs 8,279), and its FP32 throughput of 35.58 TFLOPS is 2.7 times the RTX 5070 Mobile's 13.13 TFLOPS. The 3090 also has 328 tensor cores versus 144, and 82 RT cores versus 36.

Q: How does the RTX 5070 Mobile compare in Vulkan performance?

A: The 5070 Mobile wins decisively in Geekbench Vulkan with 116,960 points versus 53,927 for the 3090 — a 116.9% advantage. This is the single largest margin in either direction across all head-to-head tests.

Q: What is the memory configuration difference?

A: The RTX 3090 has 24 GB of GDDR6X on a 384-bit bus with 936.2 GB/s bandwidth. The 5070 Mobile has 8 GB of GDDR7 on a 128-bit bus with 384.0 GB/s. The 3090's memory bandwidth is 2.4 times higher.

Q: Which card is more power-efficient?

A: The RTX 5070 Mobile is dramatically more efficient on paper. Its TDP is 50 W versus 350 W for the 3090 — a seven-fold difference. The 5070 Mobile also requires no power connectors and is an IGP-class slot width, while the 3090 needs a triple-slot cooler and a 1x 12-pin connector.

Q: How do their overall benchmark averages compare?

A: The RTX 5070 Mobile has a higher average benchmark score of 29,928 against the 3090's 27,565. However, the 3090 sits at the 73rd percentile of all GPUs, while the 5070 Mobile is at the 75th percentile — a narrow overall gap.

Q: Which card wins in legacy DirectX tests?

A: The RTX 3090 wins every DirectX test: DX9 by 20.1% (268 vs 214), DX10 by 29.1% (182 vs 129), DX11 by 12.7% (220 vs 192), and DX12 by 15.5% (110 vs 93). The 5070 Mobile only wins one head-to-head test total.

The Verdict

The data splits this comparison cleanly by use case. The RTX 3090 is the pick for anyone needing maximum compute throughput, large memory capacity, or high-resolution texture work. It wins 8 of 9 head-to-head benchmarks, leads in Passmark G3D by 23.6% (26,645 vs 20,355), and has 24 GB of VRAM versus 8 GB. Its 350 W TDP and triple-slot cooler make it a desktop-only proposition.

The RTX 5070 Mobile is the choice for portable systems or anyone who prioritizes modern API performance and efficiency. It wins the only Vulkan test by a massive 116.9% margin, posts a higher average benchmark score (29,928 vs 27,565), and uses 50 W of power versus 350 W. Its 5 nm TSMC process node (vs 8 nm Samsung for the 3090) and Blackwell 2.0 architecture deliver that efficiency.

For a desktop workstation or high-end gaming rig where power and space are not constraints, the RTX 3090 is the better performer in most measured workloads. For a laptop or any power-constrained build, the RTX 5070 Mobile is the only feasible option — and its Vulkan advantage shows it can still compete in modern titles.

Head-to-Head Benchmarks

The RTX 3090's wins are broad but not uniform. Its largest victory is in Passmark GPU compute, where it scores 15,356 versus 8,279 — a 46.1% lead. That gap aligns with the 3090's 35.58 TFLOPS FP32 throughput versus 13.13 TFLOPS for the 5070 Mobile. In Geekbench OpenCL, the 3090 also wins by 29.2% (172,758 vs 122,238), confirming a strong compute advantage.

In graphics workloads, the 3090 leads in Passmark G3D by 23.6% (26,645 vs 20,355). It also wins every legacy DirectX test: DX9 by 20.1%, DX10 by 29.1%, DX11 by 12.7%, and DX12 by 15.5%. The 2D test (Passmark G2D) also goes to the 3090 by 15.7% (1,063 vs 896).

The RTX 5070 Mobile's single win is in Geekbench Vulkan: 116,960 versus 53,927, a 116.9% lead. This is not a marginal victory — it is more than double the 3090's score. This suggests the Blackwell 2.0 architecture has substantially better Vulkan driver and hardware support, likely from its newer generation.

The average benchmark scores tell a nuanced story. The 5070 Mobile averages 29,928, which is 8.6% higher than the 3090's 27,565. This happens because the 3090's Vulkan score is anomalously low relative to its other results. The 3090's nearest rivals include the RTX 4070 Mobile (27,435) and RX 6700 XT (27,425), both within 0.5%. The 5070 Mobile's nearest rival is the RTX 3070 Ti at 29,945 (0.1% delta) and RTX 2080 Ti at 29,783 (0.5% delta).

Specification Differences

The two GPUs differ in nearly every measurable specification. The RTX 3090 has 10,496 shading units, 328 TMUs, and 112 ROPs. The RTX 5070 Mobile has 4,608 shading units, 144 TMUs, and 48 ROPs. The 3090's pixel rate is 189.8 GPixel/s versus 68.40 GPixel/s for the 5070 Mobile; texture rates are 556.0 GTexel/s versus 205.2 GTexel/s.

Memory is a major differentiator. The 3090 uses 24 GB of GDDR6X on a 384-bit bus with 936.2 GB/s bandwidth. The 5070 Mobile has 8 GB of GDDR7 on a 128-bit bus with 384.0 GB/s. The 3090's memory clock is 1219 MHz (19.5 Gbps effective), while the 5070 Mobile runs at 1500 MHz (24 Gbps effective). The newer GDDR7 is faster per pin, but the 3090's wider bus provides more total bandwidth.

Power and physical design are opposites. The 3090 has a 350 W TDP, is triple-slot, requires a 1x 12-pin power connector, and needs a 750 W suggested PSU. Its dimensions are 336 mm long, 140 mm high, and 61 mm wide. The 5070 Mobile is rated at 50 W TDP, is IGP-class (integrated into laptops), uses no power connectors, and has portable-device-dependent display outputs. The 3090 has 1x HDMI 2.1 and 3x DisplayPort 1.4a.

Bus interface also differs: the 5070 Mobile uses PCIe 5.0 x16, while the 3090 is PCIe 4.0 x16. Release dates are far apart — the 3090 launched on 2025-04-14? No, the 3090 launched 2020-08-31 and is end-of-life; the 5070 Mobile launched 2025-04-14 and is active. The 3090's launch MSRP was 1,499 USD.

Architecture Differences

The RTX 3090 uses the GA102 chip on Samsung's 8 nm process, with 28,300 million transistors on a 628 mm² die. Transistor density is 45.1 million per mm². The RTX 5070 Mobile uses the GB206 chip on TSMC's 5 nm process, with 21,900 million transistors on a 181 mm² die. Its transistor density is 121.0 million per mm² — 2.7 times denser.

Architecture generations are completely different. The 3090 is Ampere, the predecessor being GeForce 20 and successor GeForce 40. The 5070 Mobile is Blackwell 2.0, with predecessor GeForce 40 Mobile. This generation gap explains the Vulkan performance difference: Blackwell 2.0 is two generations newer.

Core counts reflect the different design philosophies. The 3090 has 82 RT cores and 328 tensor cores. The 5070 Mobile has 36 RT cores and 144 tensor cores. Despite fewer cores, the newer architecture delivers competitive modern-API performance. The 3090's FP16 performance is 35.58 TFLOPS (1:1 with FP32); the 5070 Mobile's is 13.13 TFLOPS (also 1:1).

Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The process node difference (5 nm vs 8 nm) and transistor density gap are the primary efficiency drivers. The 5070 Mobile delivers 13.13 TFLOPS at 50 W, while the 3090 delivers 35.58 TFLOPS at 350 W. Per watt, the 5070 Mobile is roughly 2.7 times more efficient in FP32 (0.26 TFLOPS/W vs 0.10 TFLOPS/W).

Where Each One Wins

RTX 3090 wins in: Compute-heavy workloads like GPU rendering, scientific simulation, and machine learning training. Its 46.1% lead in Passmark GPU compute and 29.2% lead in Geekbench OpenCL make it the clear choice for these tasks. The 24 GB VRAM (3x the 5070 Mobile's 8 GB) is essential for large models or high-resolution texture sets. It also wins every DirectX benchmark, including DX12 by 15.5%, making it better for traditional Windows gaming in most titles. The 3090's 936.2 GB/s bandwidth is 2.4 times higher, benefiting bandwidth-bound workloads.

RTX 5070 Mobile wins in: Modern Vulkan-based applications and any power-constrained environment. The 116.9% Vulkan lead (116,960 vs 53,927) is massive and suggests Blackwell 2.0's driver stack is far better optimized for current graphics APIs. It is the only choice for laptops due to its 50 W TDP and IGP form factor. Its higher average benchmark score (29,928 vs 27,565) and 75th percentile ranking (vs 73rd for the 3090) indicate it is the better overall performer in the aggregate. PCIe 5.0 support also provides future-proofing for data transfer.

Tie-breaker: If you need raw compute, VRAM capacity, and don't care about power consumption, the RTX 3090 is the data-supported winner. If you need portability, efficiency, or Vulkan performance, the RTX 5070 Mobile wins decisively. The 3090 is end-of-life with no successor listed; the 5070 Mobile is active and current. For most modern gaming and general use, the 5070 Mobile's higher average score and Vulkan advantage make it the better all-rounder despite losing 8 of 9 individual tests.