AMD Instinct MI300X vs NVIDIA GeForce RTX 5090 Mobile Comparison

Where Each One Wins

The recorded benchmark data splits these two accelerators into entirely different performance domains. The AMD Instinct MI300X holds the only direct head-to-head victory in the database, winning the Geekbench OpenCL test with a score of 317,994 against 201,834 for the NVIDIA GeForce RTX 5090 Mobile. That is a 57.6% advantage for the AMD part, a decisive margin that reflects its data-center orientation.

The NVIDIA GeForce RTX 5090 Mobile, however, does not go winless in the broader database. Its benchmark suite spans multiple DirectX and compute workloads, where it posts scores of 5871 in 3DMark Steel Nomad DX12, 198,405 in Geekbench Vulkan, 30,034 in Passmark G3D, and 13,401 in Passmark GPU Compute. The AMD Instinct MI300X has no recorded results in those tests, so the NVIDIA part wins by default in any graphics-oriented or Vulkan-specific workload. The MI300X also lists zero pixel rate and no display outputs, meaning it cannot participate in rasterization or graphics rendering at all. The RTX 5090 Mobile, with 112 ROPs and a 169.7 GPixel/s pixel rate, is the only one of the two with any rendering capability.

The use-case split is therefore stark. The MI300X is a compute-only accelerator for large-scale parallel workloads, while the RTX 5090 Mobile is a mobile graphics processor that also handles compute. In OpenCL, the AMD part dominates. In every graphics API and in Vulkan, the NVIDIA part is the sole participant and therefore the only option.

Architecture Differences

The two chips share a 5 nm TSMC process node, but diverge sharply in scale and design philosophy. The AMD Instinct MI300X uses the Aqua Vanjaram chip with CDNA 3.0 architecture, while the NVIDIA GeForce RTX 5090 Mobile uses the GB203 chip with Blackwell 2.0 architecture. The MI300X packs 153,000 million transistors on a 1017 mm² die, yielding a transistor density of 150.4 million per mm². The RTX 5090 Mobile has 45,600 million transistors on a 378 mm² die, for a density of 120.6 million per mm². The AMD part is more than three times larger in transistor count and nearly three times larger in die area.

Memory configurations differ fundamentally. The MI300X carries 192 GB of HBM3 on an 8192-bit bus, delivering 5.32 TB/s of bandwidth. The RTX 5090 Mobile has 24 GB of GDDR7 on a 256-bit bus, providing 896.0 GB/s. That is a 4.9x bandwidth advantage for the AMD accelerator. Clock speeds tell a different story: the MI300X runs at a 1000 MHz base and 2100 MHz boost, while the RTX 5090 Mobile runs at 990 MHz base and 1515 MHz boost. The AMD part boosts higher, but the NVIDIA part operates in a far lower power envelope.

The compute resources are heavily skewed toward the MI300X. It has 19,456 shading units and 1,216 TMUs, versus 10,496 shading units and 328 TMUs for the RTX 5090 Mobile. The MI300X has zero ROPs, while the RTX 5090 Mobile has 112. The RTX 5090 Mobile also includes 82 ray tracing cores and 328 tensor cores, features entirely absent from the MI300X. Pixel rate for the AMD part is 0 MPixel/s, versus 169.7 GPixel/s for the NVIDIA part. Texture rate is 2,553.6 GTexel/s for the MI300X versus 496.9 GTexel/s for the RTX 5090 Mobile.

Power and form factor also separate them. The MI300X has a 750 W TDP and uses an OAM Module slot with no power connectors, requiring a suggested PSU of 1150 W. The RTX 5090 Mobile has a 95 W TDP, uses an IGP slot, and requires no separate power connectors or suggested PSU. The MI300X has no display outputs, while the RTX 5090 Mobile outputs are portable device dependent. API support is absent for the MI300X (DirectX, OpenGL, and Vulkan all listed as N/A), while the RTX 5090 Mobile supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Head-to-Head Benchmarks

The sole recorded head-to-head benchmark is Geekbench OpenCL. The AMD Instinct MI300X scores 317,994, while the NVIDIA GeForce RTX 5090 Mobile scores 201,834. The delta is 57.6% in favor of AMD. That margin is substantial and consistent with the hardware specifications: the MI300X has nearly twice the shading units, over 5.9x the memory bandwidth, and a much larger FP32 throughput of 81.72 TFLOPS versus 31.80 TFLOPS for the RTX 5090 Mobile.

Context from the nearest rivals in the database reinforces the MI300X's standing. Its score of 317,994 places it 5% behind the NVIDIA H200 NVL (334,891), 7.5% ahead of the NVIDIA L40S (295,763), 8% behind the NVIDIA B200 (345,482), and 10.7% ahead of the NVIDIA RTX 6000 Ada Generation (287,237). The MI300X sits in the 100th percentile of all GPUs, meaning no recorded GPU scores higher in the database's aggregate ranking.

The RTX 5090 Mobile's OpenCL score of 201,834 is far lower, but its nearest rivals show a different competitive picture. It sits 0.5% behind the AMD Radeon Pro 5500 XT (45,384), 0.8% ahead of the NVIDIA GeForce RTX 4070 Ti (44,795), 1% behind the Intel Arc A730M (45,592), and 1.8% behind the NVIDIA RTX 5880 Ada Generation (45,972). Its percentile is 84, which is high for a mobile part but nowhere near the MI300X's perfect percentile.

The RTX 5090 Mobile's other benchmarks show where it excels in the absence of AMD competition. Its 3DMark Steel Nomad DX12 score of 5871, Passmark G3D score of 30,034, and Passmark GPU Compute score of 13,401 are all unopposed in the database. The Passmark DirectX tests show a range from 138 in DX12 to 324 in DX9, with 183 in DX10 and 269 in DX11. The Passmark G2D score of 1057 and Geekbench Vulkan score of 198,405 round out its recorded results.

The Verdict

The data points to a clear division of roles. The AMD Instinct MI300X is the compute leader, delivering a 57.6% higher OpenCL score than the RTX 5090 Mobile and ranking in the 100th percentile of all GPUs. Its 192 GB of HBM3 memory, 5.32 TB/s bandwidth, and 81.72 TFLOPS FP32 make it a large-scale compute accelerator with no graphics capability. Any workload that relies on OpenCL, massive memory capacity, or extreme bandwidth will favor the MI300X by a wide margin.

The NVIDIA GeForce RTX 5090 Mobile is the graphics-capable option. It is the only one of the two with display outputs, ROPs, ray tracing cores, tensor cores, and DirectX/Vulkan support. Its 84th percentile ranking reflects strong performance for a mobile part, and its 95 W TDP makes it suitable for portable systems. The MI300X cannot run any graphics workload, so for any DirectX, Vulkan, or OpenGL application, the RTX 5090 Mobile is the only choice.

Users requiring pure compute throughput and memory capacity should select the MI300X. Users needing graphics rendering, ray tracing, or mobile deployment should select the RTX 5090 Mobile. The two are not substitutes; they serve mutually exclusive purposes.

FAQ

Q: Which GPU has the higher Geekbench OpenCL score?

A: The AMD Instinct MI300X scores 317,994, which is 57.6% higher than the NVIDIA GeForce RTX 5090 Mobile's 201,834.

Q: Does the AMD Instinct MI300X support DirectX or Vulkan?

A: No. The MI300X lists DirectX, OpenGL, and Vulkan as N/A, and it has no display outputs. The RTX 5090 Mobile supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4.

Q: How do the memory configurations compare?

A: The MI300X has 192 GB of HBM3 on an 8192-bit bus with 5.32 TB/s bandwidth. The RTX 5090 Mobile has 24 GB of GDDR7 on a 256-bit bus with 896.0 GB/s bandwidth.

Q: What is the TDP difference between the two?

A: The MI300X has a 750 W TDP and requires a suggested PSU of 1150 W. The RTX 5090 Mobile has a 95 W TDP and requires no suggested PSU.

Q: Which GPU has ray tracing cores?

A: Only the NVIDIA GeForce RTX 5090 Mobile has ray tracing cores, with 82 of them. The MI300X has no ray tracing cores or tensor cores.

Q: How does the MI300X rank against its nearest rivals?

A: The MI300X scores 5% behind the NVIDIA H200 NVL, 7.5% ahead of the NVIDIA L40S, 8% behind the NVIDIA B200, and 10.7% ahead of the NVIDIA RTX 6000 Ada Generation. It is in the 100th percentile of all GPUs.