NVIDIA GeForce RTX 5070 vs NVIDIA GeForce RTX 5090 Mobile Comparison

The NVIDIA GeForce RTX 5090 Mobile and the NVIDIA GeForce RTX 5070 are both Blackwell 2.0 parts, but they target different corners of the market. Benchmark data shows the RTX 5090 Mobile wins 7 of 10 head-to-head tests, yet the RTX 5070 counters with significant victories in compute and 2D workloads. The RTX 5090 Mobile posts an average benchmark score of 45152, placing it in the 84th percentile of all GPUs, while the RTX 5070 scores 40377, landing in the 82nd percentile. The mobile part is 0.5% behind the AMD Radeon Pro 5500 XT and 0.8% ahead of the NVIDIA GeForce RTX 4070 Ti; the desktop RTX 5070 is 0.1% ahead of the AMD Radeon Pro 580 and 0.8% ahead of the AMD Radeon Pro WX 7100. Pick the RTX 5090 Mobile for maximum raster and ray-tracing performance in a laptop; pick the RTX 5070 if raw compute throughput and 2D responsiveness matter more.

The Verdict

The data splits these two cleanly. The RTX 5090 Mobile is the faster gaming and 3D rendering part. It leads in 3DMark Steel Nomad DX12 by 15.6% (5871 vs 5077), in Geekbench OpenCL by 16.9% (201834 vs 172660), and in Geekbench Vulkan by 10.9% (198405 vs 178923). It also wins PassMark G3D by 3.1% (30034 vs 29137) and PassMark DirectX 12 by a massive 27.8% (138 vs 108). For anyone prioritizing modern game workloads and DirectX 12 Ultimate titles, the RTX 5090 Mobile is the clear choice from this data.

The RTX 5070, however, is not a straightforward loser. It wins PassMark GPU Compute by 15.1% (15787 vs 13401) and PassMark G2D by 19% (1305 vs 1057). It also narrowly takes PassMark DirectX 11 by 2.9% (277 vs 269). This makes it the better option for compute-heavy tasks like GPU-accelerated rendering or machine learning inference, and for 2D desktop work. The RTX 5090 Mobile’s average score is 11.8% higher overall, but that advantage is not universal.

The RTX 5090 Mobile also carries a 95 W TDP versus the RTX 5070’s 250 W, despite the mobile part being an IGP with no power connectors. The RTX 5070 is a dual-slot desktop card requiring a 600 W PSU and a 16-pin connector. If power efficiency in a portable chassis is the priority, the RTX 5090 Mobile wins on paper. If absolute compute throughput in a fixed desktop rig is the goal, the RTX 5070’s wins in GPU Compute make it the data-backed pick.

Where Each One Wins

The RTX 5090 Mobile dominates in every 3D-focused benchmark category except one. Its largest victory is PassMark DirectX 12, where it scores 27.8% higher (138 vs 108). This suggests a strong advantage in modern API workloads that leverage async compute and ray tracing. The 15.6% lead in 3DMark Steel Nomad DX12 reinforces this, as that test stresses DX12 feature sets. Geekbench OpenCL and Vulkan wins of 16.9% and 10.9% respectively show broad compute API strength, though the OpenCL win is notable because it contradicts the PassMark GPU Compute result.

The RTX 5070 wins exactly where the RTX 5090 Mobile is weakest. Its 15.1% lead in PassMark GPU Compute (15787 vs 13401) is the largest margin for either card in any test. This is a pure shader-throughput metric, and the desktop part’s higher clocks (2512 MHz boost vs 1515 MHz boost) likely drive this. The 19% win in PassMark G2D (1305 vs 1057) points to better 2D acceleration and memory access patterns for desktop UI work. The small 2.9% win in PassMark DirectX 11 (277 vs 269) is a legacy API edge.

The remaining tests are close. PassMark DirectX 10 and DirectX 9 go to the RTX 5090 Mobile by 1.7% (183 vs 180) and 1.3% (324 vs 320) respectively — essentially ties. PassMark G3D is a 3.1% win for the RTX 5090 Mobile (30034 vs 29137), which is a narrower margin than the 3DMark and Geekbench gaps suggest. This indicates the RTX 5090 Mobile’s advantage grows with API complexity and resolution, while the RTX 5070 holds its ground in simpler or older workloads.

Architecture Differences

Both GPUs use TSMC’s 5 nm process and share the Blackwell 2.0 architecture. The RTX 5090 Mobile uses the GB203 chip, while the RTX 5070 uses GB205. The RTX 5090 Mobile is physically larger: 378 mm² die size versus 263 mm². Transistor counts differ substantially, with the RTX 5090 Mobile at 45,600 million and the RTX 5070 at 31,100 million. Transistor density is similar — 120.6M per mm² for the mobile part and 118.3M per mm² for the desktop part.

The RTX 5090 Mobile packs 10496 shading units, 328 TMUs, and 112 ROPs. The RTX 5070 has 6144 shading units, 192 TMUs, and 80 ROPs. Ray tracing cores also differ: 82 on the RTX 5090 Mobile versus 48 on the RTX 5070. Tensor core counts are 328 versus 192. This hardware advantage explains the RTX 5090 Mobile’s higher FP32 throughput of 31.80 TFLOPS versus 30.87 TFLOPS on the RTX 5070 — a narrow 3% lead despite a 71% advantage in shading units. The RTX 5070 compensates with much higher clocks: 2325 MHz base and 2512 MHz boost, versus 990 MHz base and 1515 MHz boost on the RTX 5090 Mobile.

Memory configurations diverge sharply. The RTX 5090 Mobile has 24 GB of GDDR7 on a 256-bit bus, yielding 896.0 GB/s bandwidth. The RTX 5070 has 12 GB of GDDR7 on a 192-bit bus, yielding 672.0 GB/s bandwidth. The RTX 5090 Mobile’s memory bandwidth advantage is 33.3%. Both run memory at 1750 MHz with 28 Gbps effective speed. The RTX 5090 Mobile also leads in pixel rate (169.7 GPixel/s vs 201.0 GPixel/s for the RTX 5070 — wait, the RTX 5070 is actually higher) and texture rate (496.9 GTexel/s vs 482.3 GTexel/s). The RTX 5070’s higher pixel rate of 201.0 GPixel/s suggests better fill-rate efficiency from its boost clocks.

Form factor and power delivery are the biggest differentiators. The RTX 5090 Mobile is an IGP with a 95 W TDP, no power connectors, and portable-device-dependent display outputs. The RTX 5070 is a dual-slot card measuring 245 mm by 115 mm by 40 mm, with a 250 W TDP, one 16-pin connector, and a 600 W suggested PSU. It outputs via 1x HDMI 2.1b and 3x DisplayPort 2.1b. Both use PCIe 5.0 x16 and support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The RTX 5090 Mobile released on 2025-03-26; the RTX 5070 on 2025-03-03.

FAQ

Q: Which GPU has higher average benchmark scores?

A: The RTX 5090 Mobile has an average benchmark score of 45152, which is 11.8% higher than the RTX 5070’s 40377.

Q: Does the RTX 5070 win any benchmark categories?

A: Yes, it wins three tests: PassMark GPU Compute (15787 vs 13401, a 15.1% lead), PassMark G2D (1305 vs 1057, a 19% lead), and PassMark DirectX 11 (277 vs 269, a 2.9% lead).

Q: What is the memory capacity difference?

A: The RTX 5090 Mobile has 24 GB of GDDR7 on a 256-bit bus, while the RTX 5070 has 12 GB of GDDR7 on a 192-bit bus. The RTX 5090 Mobile’s bandwidth is 896.0 GB/s versus 672.0 GB/s.

Q: How do power requirements compare?

A: The RTX 5090 Mobile has a 95 W TDP and is an IGP with no power connectors. The RTX 5070 has a 250 W TDP, requires a 16-pin connector, and suggests a 600 W PSU.

Q: Which GPU has better ray tracing hardware?

A: The RTX 5090 Mobile has 82 RT cores, compared to 48 on the RTX 5070. It also has 328 tensor cores versus 192.

Q: What is the FP32 performance difference?

A: The RTX 5090 Mobile delivers 31.80 TFLOPS FP32, slightly ahead of the RTX 5070’s 30.87 TFLOPS. Both achieve FP16 at 1:1 ratios with their respective FP32 figures.

Head-to-Head Benchmarks

The largest single margin in this comparison belongs to the RTX 5090 Mobile in PassMark DirectX 12, where it scores 138 versus 108 — a 27.8% advantage. This is a legacy API test, but the gap suggests the mobile part’s extra ROPs and memory bandwidth (896.0 GB/s vs 672.0 GB/s) pay off under heavy draw-call loads. The RTX 5070’s biggest win is PassMark GPU Compute, where it posts 15787 against 13401, a 15.1% lead. That result is surprising given the RTX 5090 Mobile’s higher FP32 rating, but the desktop part’s boost clock of 2512 MHz versus 1515 MHz appears to dominate raw shader execution.

In 3DMark Steel Nomad DX12, the RTX 5090 Mobile wins 5871 to 5077, a 15.6% margin. This is the most modern test in the suite and reflects the RTX 5090 Mobile’s superior RT core count (82 vs 48) and memory bandwidth. Geekbench OpenCL shows a 16.9% win (201834 vs 172660) for the RTX 5090 Mobile, while Geekbench Vulkan shows a 10.9% win (198405 vs 178923). These two results bracket the compute performance gap — OpenCL favors the mobile part more than Vulkan.

The RTX 5070’s 19% win in PassMark G2D (1305 vs 1057) is the second-largest margin overall. This 2D test is sensitive to memory latency and driver overhead, and the RTX 5070’s lower memory size (12 GB) with higher clocks may reduce access latency. PassMark DirectX 11 is a narrow 2.9% win for the RTX 5070 (277 vs 269), suggesting that older DX11 titles run nearly identically on both. PassMark DirectX 10 and DirectX 9 are essentially tied, with the RTX 5090 Mobile winning by 1.7% (183 vs 180) and 1.3% (324 vs 320) respectively.

PassMark G3D, the overall gaming proxy, goes to the RTX 5090 Mobile by just 3.1% (30034 vs 29137). This is the smallest 3D margin, indicating that in mixed workloads the two are close. The RTX 5090 Mobile’s wins in 3DMark and Geekbench are larger, which points to its advantage scaling with API modernity and memory bandwidth demands. The RTX 5070’s PassMark GPU Compute win, however, is a clear counter-signal: for non-graphics compute, the desktop part is the better tool. Overall, the RTX 5090 Mobile wins 7 benchmarks; the RTX 5070 wins 3.