AMD Instinct MI300A vs NVIDIA GeForce RTX 5090 SE Comparison

The Verdict

The AMD Instinct MI300A and NVIDIA GeForce RTX 5090 SE occupy entirely different positions in the database. The MI300A is an Instinct-series accelerator built for compute-heavy workloads, while the RTX 5090 SE is a GeForce 50-series consumer graphics card. The recorded data shows no benchmark scores for either part, and both sit at the 50th percentile among all GPUs with an average benchmark score of zero. The MI300A targets scenarios requiring massive memory capacity and bandwidth, whereas the RTX 5090 SE delivers a conventional graphics feature set with display outputs and a dual-slot form factor. The MI300A has no display outputs and uses an OAM module slot, so it cannot serve as a desktop graphics card. The RTX 5090 SE, with its 1x HDMI 2.1b and 3x DisplayPort 2.1b outputs, is the only one of the two that can drive a monitor.

The data indicates the RTX 5090 SE offers higher FP32 throughput at 66.94 TFLOPS versus 61.29 TFLOPS for the MI300A. The MI300A counters with a 5.32 TB/s memory bandwidth figure that is nearly four times the RTX 5090 SE's 1.34 TB/s, along with 128 GB of HBM3 memory versus 24 GB of GDDR7. The RTX 5090 SE includes 110 RT cores and 440 tensor cores; the MI300A lists no RT cores and no tensor cores in the database. The RTX 5090 SE also supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, while the MI300A reports N/A for all three APIs. The MI300A has a 750 W TDP and requires a 1150 W suggested PSU, while the RTX 5090 SE draws 500 W with a 900 W suggested PSU.

Architecture Differences

The MI300A uses the Aqua Vanjaram chip built on the CDNA 3.0 architecture, fabricated by TSMC on a 5 nm process. It contains 153,000 million transistors on a 1017 mm² die, yielding a transistor density of 150.4M per mm². The chip belongs to the Instinct (MIx) generation and was released on 2023-12-05. The RTX 5090 SE uses the GB202 chip on the Blackwell 2.0 architecture, also built by TSMC on a 5 nm process. Its transistor count is 92,200 million on a 750 mm² die, for a density of 122.9M per mm². It is part of the GeForce 50-series and carries a release date of 2025-12-31. The MI300A lists Radeon Instinct as its predecessor, while the RTX 5090 SE lists GeForce 40 as its predecessor and GeForce 60 as its successor.

The MI300A's memory subsystem uses HBM3 with a 8192 bit bus and 128 GB capacity. The RTX 5090 SE uses GDDR7 with a 384 bit bus and 24 GB capacity. The MI300A clocks memory at 1300 MHz with 5.2 Gbps effective speed, while the RTX 5090 SE runs memory at 1750 MHz with 28 Gbps effective speed. The base clock of the MI300A is 1000 MHz with a boost of 2100 MHz; the RTX 5090 SE starts at 1740 MHz base and boosts to 2377 MHz.

The MI300A has 0 ROPs and a pixel rate of 0 MPixel/s, which aligns with its lack of display outputs. The RTX 5090 SE has 160 ROPs and a pixel rate of 380.3 GPixel/s. Texture rates also differ substantially: the MI300A reaches 1,915.2 GTexel/s with 912 TMUs, while the RTX 5090 SE reaches 1,045.9 GTexel/s with 440 TMUs. Shading unit counts are close, with the MI300A at 14,592 and the RTX 5090 SE at 14,080.

Head-to-Head Benchmarks

The database records no head-to-head benchmark entries for this pair, and neither part has individual benchmark scores. The wins counters show zero for both sides. However, the specification data provides measurable comparisons. In FP32 compute, the RTX 5090 SE leads with 66.94 TFLOPS, which is approximately 9% higher than the MI300A's 61.29 TFLOPS. The MI300A holds a decisive advantage in memory bandwidth at 5.32 TB/s versus 1.34 TB/s, a margin of roughly 297%. Texture fill rate favors the MI300A at 1,915.2 GTexel/s compared to 1,045.9 GTexel/s for the RTX 5090 SE, an advantage of about 83%. Pixel fill rate belongs exclusively to the RTX 5090 SE at 380.3 GPixel/s, since the MI300A reports 0 MPixel/s.

The RTX 5090 SE also leads in clock speeds. Its boost clock of 2377 MHz exceeds the MI300A's 2100 MHz by 277 MHz, or about 13%. The base clock gap is larger in relative terms: 1740 MHz versus 1000 MHz, a 74% difference. The RTX 5090 SE's memory clock of 1750 MHz with 28 Gbps effective speed contrasts with the MI300A's 1300 MHz and 5.2 Gbps effective speed, though the MI300A's far wider 8192 bit bus compensates in total bandwidth.

The RTX 5090 SE is the only one with ray tracing and tensor hardware: 110 RT cores and 440 tensor cores. It also supports FP16 at 66.94 TFLOPS with a 1:1 ratio, while the MI300A lists no FP16 figure. The MI300A's transistor count is higher by 60,800 million, and its die is larger by 267 mm², but the RTX 5090 SE achieves higher transistor density per unit of FP32 throughput in the recorded data.

Specification Differences

The two parts differ across nearly every recorded field. Process node and foundry are identical: both use 5 nm TSMC. Bus interface is also the same at PCIe 5.0 x16. Everything else diverges.

• Transistors: 153,000 million (MI300A) vs 92,200 million (RTX 5090 SE)
• Die size: 1017 mm² vs 750 mm²
• Transistor density: 150.4M / mm² vs 122.9M / mm²
• Base clock: 1000 MHz vs 1740 MHz
• Boost clock: 2100 MHz vs 2377 MHz
• Memory clock: 1300 MHz 5.2 Gbps effective vs 1750 MHz 28 Gbps effective
• Memory size: 128 GB vs 24 GB
• Memory type: HBM3 vs GDDR7
• Memory bus: 8192 bit vs 384 bit
• Memory bandwidth: 5.32 TB/s vs 1.34 TB/s
• Shading units: 14592 vs 14080
• TMUs: 912 vs 440
• ROPs: 0 vs 160
• RT cores: none listed vs 110
• Tensor cores: none listed vs 440
• Pixel rate: 0 MPixel/s vs 380.3 GPixel/s
• Texture rate: 1,915.2 GTexel/s vs 1,045.9 GTexel/s
• FP32: 61.29 TFLOPS vs 66.94 TFLOPS
• FP16: not listed vs 66.94 TFLOPS (1:1)
• TDP: 750 W vs 500 W
• Slot width: OAM Module vs Dual-slot
• Power connectors: None vs 1x 16-pin
• Suggested PSU: 1150 W vs 900 W
• Display outputs: No outputs vs 1x HDMI 2.1b, 3x DisplayPort 2.1b
• DirectX: N/A vs 12 Ultimate (12_2)
• OpenGL: N/A vs 4.6
• Vulkan: N/A vs 1.4
• Length: not listed vs 267 mm (10.5 inches)
• Height: not listed vs 111 mm (4.4 inches)
• Width: not listed vs 40 mm (1.6 inches)
• Production status: not listed vs Active
• Release date: 2023-12-05 vs 2025-12-31
• Launch MSRP: not listed vs 1,499 USD (stated once, per database rules)

The MI300A uses no power connectors because it is an OAM module, while the RTX 5090 SE uses a single 16-pin connector. The MI300A's TDP is 250 W higher, and its suggested PSU is 250 W higher. The RTX 5090 SE is the only part with a physical dimension record and an active production status.

FAQ

Q: Which part has higher FP32 compute performance?

A: The NVIDIA GeForce RTX 5090 SE records 66.94 TFLOPS FP32, which is about 9% higher than the AMD Instinct MI300A's 61.29 TFLOPS.

Q: How do the memory configurations compare?

A: The MI300A has 128 GB of HBM3 on a 8192 bit bus with 5.32 TB/s bandwidth. The RTX 5090 SE has 24 GB of GDDR7 on a 384 bit bus with 1.34 TB/s bandwidth.

Q: Does the MI300A support ray tracing?

A: The database lists no RT cores for the MI300A. The RTX 5090 SE lists 110 RT cores and 440 tensor cores.

Q: Can the MI300A output video to a display?

A: No. The MI300A has no display outputs and reports a pixel rate of 0 MPixel/s. The RTX 5090 SE has 1x HDMI 2.1b and 3x DisplayPort 2.1b outputs.

Q: What are the power requirements for each part?

A: The MI300A has a 750 W TDP and a 1150 W suggested PSU. The RTX 5090 SE has a 500 W TDP and a 900 W suggested PSU.

Q: Which part has higher texture fill rate?

A: The MI300A reaches 1,915.2 GTexel/s with 912 TMUs, which is about 83% higher than the RTX 5090 SE's 1,045.9 GTexel/s with 440 TMUs.