NVIDIA GeForce RTX 5090 Mobile

NVIDIA’s GeForce RTX 5090 Mobile brings the Blackwell architecture to laptops, pairing a 5 nm GB203 chip with 24 GB of GDDR7 memory. The data positions this part as a high-end mobile GPU, landing in the 85th percentile of all GPUs tracked by the database. Its average benchmark score of 45,152 places it in a crowded field where the margins between competing parts are remarkably thin.

Benchmark Performance

The benchmark data shows a GPU that trades blows with desktop-class hardware from the previous generation, though the margins are surprisingly narrow. In the aggregate average benchmark score, the RTX 5090 Mobile sits at 45,152. The nearest rival, the NVIDIA GeForce RTX 4070 Ti, scores 44,900, putting the mobile part just 0.6% ahead. That is a statistical tie in practical terms, indicating that the laptop GPU delivers performance on par with a desktop RTX 4070 Ti across the full suite of tests.

The comparison becomes more complex when looking at other rivals. The Intel Arc A730M averages 45,592, which is 1% higher than the RTX 5090 Mobile. Similarly, the AMD Radeon Pro 5500 XT scores 45,642, also 1.1% ahead. These deltas are within normal run-to-run variance, so the data supports a three-way tie at the top of this particular comparison group. The only rival that trails meaningfully is the AMD Radeon Pro 580, which scores 44,195; the RTX 5090 Mobile leads that part by 2.2%.

Drilling into individual tests reveals where the RTX 5090 Mobile excels. In 3DMark Steel Nomad DX12, it posts a score of 5,871, which is a strong result for a modern DirectX 12 workload. The Geekbench OpenCL score of 201,834 and Vulkan score of 198,405 show compute-heavy performance that is nearly identical across the two APIs, differing by only 1.7%. This suggests the GPU maintains consistent throughput regardless of the compute interface.

The PassMark suite tells a more nuanced story. The G3D score of 30,034 is the headline number, reflecting strong overall 3D rendering capability. However, the DirectX 9 score of 324 and DirectX 10 score of 183 indicate that older API workloads do not scale as well. The DirectX 11 score of 269 and DirectX 12 score of 138 show a peculiar inversion where the newer API actually scores lower than DirectX 11. This pattern suggests the GPU is heavily optimized for modern rendering paths, with legacy APIs receiving less attention. The G2D score of 1,057 is respectable for a discrete GPU, and the GPU compute score of 13,401 confirms solid general-purpose compute performance.

When interpreting these numbers, the key takeaway is that the RTX 5090 Mobile is not a generational leap over its nearest competitors. It edges out the RTX 4070 Ti by a hair, ties with the Intel Arc A730M, and trails the Radeon Pro 5500 XT by a negligible margin. The 2.2% advantage over the Radeon Pro 580 is the only decisive win in the group. For a flagship mobile GPU, the data suggests that performance headroom is limited at this power envelope.

Memory Subsystem

The memory configuration is where the RTX 5090 Mobile makes its most compelling case. It ships with 24 GB of GDDR7 memory on a 256-bit bus, yielding a bandwidth of 896.0 GB/s. This is a substantial amount of VRAM for a laptop GPU, and the GDDR7 type represents the latest memory standard. The effective memory clock is 1,750 MHz, translating to 28 Gbps effective data rate.

For high-resolution gaming, the 24 GB capacity is the standout feature. At 4K, texture-heavy scenes and high-detail assets can consume 12-16 GB alone, and the extra headroom beyond that provides future-proofing for the next wave of game releases. The 896.0 GB/s bandwidth is also critical for feeding the 10,496 shading units and 328 texture mapping units. This bandwidth figure is what allows the GPU to maintain high fill rates—169.7 GPixel/s pixel rate and 496.9 GTexel/s texture rate—without stalling on memory fetches.

The 256-bit bus width is a balanced choice; it is narrower than flagship desktop parts but paired with faster GDDR7 memory to compensate. The data indicates that this configuration delivers enough bandwidth to keep the GPU busy in most scenarios. However, the benchmark scores suggest that the memory subsystem does not fully differentiate the RTX 5090 Mobile from rivals with smaller memory pools, as the overall score gap to the RTX 4070 Ti is only 0.6%. This implies that in many workloads, the GPU cores are the limiting factor rather than memory capacity or bandwidth.

Ray Tracing and Feature Set

The RTX 5090 Mobile is equipped with 82 ray tracing cores and 328 tensor cores, built on the Blackwell 2.0 architecture. The tensor core count matches the texture mapping unit count at 328, which is a notable design choice that aligns AI compute resources with texture processing throughput. The FP32 compute rating is 31.80 TFLOPS, with FP16 also rated at 31.80 TFLOPS at a 1:1 ratio, indicating that half-precision workloads run at the same throughput as full precision.

API support is comprehensive for modern applications. The GPU supports DirectX 12 Ultimate with feature level 12_2, which is the current flagship API tier. OpenGL 4.6 and Vulkan 1.4 round out the API coverage. The ray tracing hardware, combined with DirectX 12 Ultimate support, means the GPU is fully compatible with hardware-accelerated ray tracing in supported titles. The tensor cores provide the compute foundation for DLSS-style upscaling and AI-based rendering features, though the fact pack does not specify which AI features are enabled.

The 82 RT cores are a meaningful count for a mobile part, though the benchmark data does not isolate ray tracing performance specifically. The 3DMark Steel Nomad DX12 score of 5,871 likely includes some ray-traced elements, but the overall score is dominated by rasterization. For users prioritizing ray tracing, the hardware is present and capable, but the data does not provide a direct comparison to rivals in RT-specific workloads.

Power and Cooling

The RTX 5090 Mobile carries a TDP of 95 W, which is a modest power envelope for a GPU with this compute capability. The slot width is listed as IGP, meaning the GPU is integrated into the system board rather than a replaceable module. Power delivery is handled through the system, with no dedicated power connectors listed. The suggested PSU field is null, indicating that NVIDIA does not specify a separate power supply requirement for this mobile part.

The 95 W TDP is critical context for the benchmark results. Running at this power level, the GPU achieves performance within 1% of desktop parts like the RTX 4070 Ti, which typically consume significantly more power. This efficiency is a direct result of the 5 nm TSMC process node and the Blackwell architecture's power management. The lack of power connectors simplifies system integration, as the GPU draws power through the motherboard's standard power delivery circuitry.

Cooling requirements are not specified in the fact pack, but the 95 W TDP suggests that a capable cooling solution is necessary to sustain boost clocks. The base clock is 990 MHz with a boost clock of 1,515 MHz, a 53% increase under load. Sustaining that boost clock requires adequate thermal headroom, which is the responsibility of the laptop manufacturer's cooling design. The data does not provide thermal throttling information, so sustained performance is unknown, but the benchmark scores likely reflect typical sustained conditions.

Who Should Consider It

The RTX 5090 Mobile is positioned for users who demand maximum performance in a laptop form factor. The 85th percentile ranking across all GPUs indicates that this is a top-tier part, outperforming the vast majority of discrete GPUs in the database. The 24 GB VRAM capacity makes it suitable for 4K gaming with high-detail textures, as well as for content creation workloads that require large memory buffers for rendering and AI tasks.

For gamers, the data suggests this GPU can handle 4K gaming at high settings, though the narrow margin over the RTX 4070 Ti (0.6%) means that users upgrading from a high-end desktop GPU will not see a dramatic improvement. The performance is more compelling for users coming from mid-range laptops, where the step up to this level of compute and memory will be substantial. The DirectX 12 Ultimate support ensures compatibility with the latest game APIs, and the ray tracing cores provide future-proofing as RT adoption grows.

For professionals, the 24 GB VRAM and 31.80 TFLOPS FP32 compute make this a capable workstation GPU for 3D rendering, video editing, and machine learning inference. The Vulkan 1.4 and OpenGL 4.6 support cover the major professional APIs. However, the compute score of 13,401 in PassMark is not class-leading, and users with heavy compute workloads may want to verify that the 95 W power envelope does not limit sustained performance.

Users who should avoid this GPU are those targeting 1080p or 1440p gaming on a budget; the performance is overkill for those resolutions, and the 24 GB VRAM offers no benefit at lower settings. Similarly, users who primarily play esports titles at high frame rates will not see proportional gains from this hardware, as the benchmark data shows lower scores in older API workloads. The RTX 5090 Mobile is a halo product for flagship laptops, and the data confirms it delivers flagship-level performance—but the competition is closer than the generational name suggests.