AMD Instinct MI455X vs NVIDIA GeForce RTX 5090 SE Comparison

Head-to-Head Benchmarks

The database contains no recorded head-to-head benchmark results for the AMD Instinct MI455X and the NVIDIA GeForce RTX 5090 SE. Both entries show an average benchmark score of zero, with no benchmark entries listed and no nearest rival comparisons available. Consequently, direct performance deltas cannot be computed from recorded measurements. The absence of data means that any comparative performance statements must rely strictly on the architectural and specification differences captured in the database, rather than observed test outcomes.

The raw compute specifications, however, provide a basis for theoretical comparison. The AMD Instinct MI455X delivers 157.3 TFLOPS of FP32 throughput, which is more than double the 66.94 TFLOPS of the RTX 5090 SE. Similarly, FP16 performance stands at 157.3 TFLOPS for the AMD part versus 66.94 TFLOPS for the NVIDIA part, with both implementations running at a 1:1 ratio to their FP32 figures. Texture rate follows the same pattern: the MI455X posts 2,457.6 GTexel/s, while the RTX 5090 SE reaches 1,045.9 GTexel/s. These figures indicate that the AMD accelerator holds a substantial raw compute advantage on paper.

Memory bandwidth similarly favors the AMD part. The Instinct MI455X has a memory bandwidth of 23.3 TB/s, compared to 1.34 TB/s for the RTX 5090 SE. The AMD card also carries 432 GB of HBM4 memory across a 24576-bit bus, whereas the NVIDIA card uses 24 GB of GDDR7 across a 384-bit bus. The pixel rate measurement is the only specification where the NVIDIA card reports a meaningful value: 380.3 GPixel/s, while the MI455X lists 0 MPixel/s, reflecting its lack of a conventional raster output stage.

Clock speeds show a mixed picture. The RTX 5090 SE has a higher base clock at 1740 MHz versus 1000 MHz, and a slightly higher boost clock at 2377 MHz versus 2400 MHz for the AMD part. Memory clocks also differ: the NVIDIA card runs at 1750 MHz with 28 Gbps effective, while the AMD card runs at 1900 MHz with 7.6 Gbps effective. The effective data rate difference is explained by the very different memory architectures, with HBM4 on the AMD side and GDDR7 on the NVIDIA side.

Power and physical characteristics diverge sharply. The MI455X has a TDP of 2300 W and a suggested PSU of 2700 W, while the RTX 5090 SE has a TDP of 500 W and a suggested PSU of 900 W. The AMD module uses no power connectors and is specified as an EAM Module slot width, whereas the NVIDIA card is a dual-slot design with a single 16-pin power connector. Dimensions are recorded only for the NVIDIA card: 267 mm in length, 111 mm in height, and 40 mm in width.

The Verdict

Given the absence of benchmark results, the verdict must be drawn from the specification differentials recorded in the database. The AMD Instinct MI455X is positioned as a compute-oriented accelerator with a 2 nm process node, a 320,000 million transistor count on a 2990 mm² die, and a transistor density of 107.0M per mm². The NVIDIA GeForce RTX 5090 SE is a consumer graphics card built on a 5 nm process with 92,200 million transistors on a 750 mm² die, yielding a higher transistor density of 122.9M per mm².

The MI455X targets workloads where FP32 and FP16 throughput, memory capacity, and memory bandwidth are the dominant factors. Its 432 GB memory pool and 23.3 TB/s bandwidth exceed the RTX 5090 SE by a wide margin. The RTX 5090 SE, by contrast, offers a conventional graphics feature set: 160 ROPs, 110 RT cores, 440 tensor cores, and support for DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The MI455X lists no API support in the database, no display outputs, and no RT or tensor core counts.

The data indicates that these are not competing products in the same segment. The MI455X is a datacenter-oriented module with a 2300 W power envelope and EAM Module form factor, while the RTX 5090 SE is a dual-slot graphics card with a 500 W TDP, a 16-pin power connector, and display outputs including 1x HDMI 2.1b and 3x DisplayPort 2.1b. The launch MSRP for the RTX 5090 SE is 1,499 USD. The MI455X has no recorded launch MSRP.

Where Each One Wins

The AMD Instinct MI455X wins on raw compute throughput. Its FP32 and FP16 figures of 157.3 TFLOPS are 2.35 times the NVIDIA card's 66.94 TFLOPS. Its texture rate of 2,457.6 GTexel/s is more than double the 1,045.9 GTexel/s of the RTX 5090 SE. Memory capacity and bandwidth also favor AMD: 432 GB versus 24 GB, and 23.3 TB/s versus 1.34 TB/s. The MI455X uses HBM4 memory, which is a generational step beyond the GDDR7 used in the NVIDIA card.

The NVIDIA GeForce RTX 5090 SE wins on conventional graphics capabilities. It has 160 ROPs, 110 RT cores, and 440 tensor cores, none of which are reported for the MI455X. Its pixel rate of 380.3 GPixel/s is recorded, while the MI455X lists 0 MPixel/s. The NVIDIA card also supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, whereas the MI455X has no API support listed. The RTX 5090 SE has display outputs; the MI455X has none.

Clock speeds favor the NVIDIA card in absolute terms at both base and boost. The RTX 5090 SE operates at 1740 MHz base and 2377 MHz boost, compared to 1000 MHz base and 2400 MHz boost for the MI455X. The effective memory data rate also favors NVIDIA: 28 Gbps effective versus 7.6 Gbps effective. However, the AMD card's much wider bus and larger memory pool reverse the practical bandwidth comparison.

Power efficiency, measured as performance per watt, cannot be computed precisely without benchmark data, but the specification sheet shows the RTX 5090 SE requires 500 W versus 2300 W for the MI455X. The suggested PSU of 900 W for the NVIDIA card versus 2700 W for the AMD card reflects the same gap. The RTX 5090 SE is an active production product with dimensions of 267 mm by 111 mm by 40 mm; the MI455X has no recorded dimensions or production status.

FAQ

Q: Which card has higher FP32 compute performance?

A: The AMD Instinct MI455X reports 157.3 TFLOPS of FP32, compared to 66.94 TFLOPS for the NVIDIA GeForce RTX 5090 SE.

Q: What memory configurations do the two cards use?

A: The MI455X uses 432 GB of HBM4 on a 24576-bit bus with 23.3 TB/s bandwidth. The RTX 5090 SE uses 24 GB of GDDR7 on a 384-bit bus with 1.34 TB/s bandwidth.

Q: Does the AMD card support display outputs?

A: No. The MI455X lists no display outputs, while the RTX 5090 SE has 1x HDMI 2.1b and 3x DisplayPort 2.1b.

Q: What is the power requirement for each card?

A: The MI455X has a TDP of 2300 W and a suggested PSU of 2700 W. The RTX 5090 SE has a TDP of 500 W and a suggested PSU of 900 W.

Q: Which card has RT and tensor cores?

A: Only the NVIDIA GeForce RTX 5090 SE reports RT cores (110) and tensor cores (440). The AMD Instinct MI455X lists no values for either.

Q: What API support is recorded for each card?

A: The RTX 5090 SE supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The MI455X has no API support listed in the database.

Architecture Differences

The two accelerators are built on different architectures from different generations. The AMD Instinct MI455X uses CDNA 5.0 architecture with a chip designated as MI450 256CU, belonging to the Instinct (MIx) generation. The NVIDIA GeForce RTX 5090 SE uses Blackwell 2.0 architecture with a GB202 chip, belonging to the GeForce 50 generation. The MI455X has a predecessor listed as Radeon Instinct, while the RTX 5090 SE has a predecessor of GeForce 40 and a successor of GeForce 60.

The manufacturing processes differ: the MI455X is fabricated on a 2 nm node at TSMC, while the RTX 5090 SE uses a 5 nm node, also at TSMC. Transistor counts diverge substantially: the MI455X contains 320,000 million transistors on a 2990 mm² die, giving a density of 107.0M per mm². The RTX 5090 SE contains 92,200 million transistors on a 750 mm² die, giving a higher density of 122.9M per mm².

The memory subsystems are entirely different in type and scale. The MI455X uses HBM4 with a 24576-bit bus and 23.3 TB/s bandwidth. The RTX 5090 SE uses GDDR7 with a 384-bit bus and 1.34 TB/s bandwidth. The shading units also differ: 32,768 for the AMD card versus 14,080 for the NVIDIA card. Texture mapping units are 1024 versus 440, and the NVIDIA card has 160 ROPs while the AMD card reports zero ROPs. The MI455X lists no RT cores or tensor cores, whereas the RTX 5090 SE has 110 and 440 respectively.

Form factor and power delivery reflect the different design goals. The MI455X is an EAM Module with no power connectors, while the RTX 5090 SE is a dual-slot card with one 16-pin power connector. The bus interface also differs: PCIe 6.0 x16 for the AMD card versus PCIe 5.0 x16 for the NVIDIA card. The MI455X has no display outputs; the RTX 5090 SE includes 1x HDMI 2.1b and 3x DisplayPort 2.1b.

Specification Differences

The recorded specifications that differ between the two products are extensive. The manufacturing node is 2 nm for the MI455X and 5 nm for the RTX 5090 SE. Transistor count is 320,000 million versus 92,200 million. Die size is 2990 mm² versus 750 mm². Transistor density is 107.0M per mm² versus 122.9M per mm².

Base clock is 1000 MHz for the AMD card and 1740 MHz for the NVIDIA card. Boost clock is 2400 MHz versus 2377 MHz. Memory clock is 1900 MHz with 7.6 Gbps effective for the AMD card, and 1750 MHz with 28 Gbps effective for the NVIDIA card. Memory size is 432 GB versus 24 GB. Memory type is HBM4 versus GDDR7. Bus width is 24576 bit versus 384 bit. Memory bandwidth is 23.3 TB/s versus 1.34 TB/s.

Shading units are 32,768 versus 14,080. TMUs are 1024 versus 440. ROPs are 0 versus 160. RT cores are not listed for the AMD card and are 110 for the NVIDIA card. Tensor cores are not listed for the AMD card and are 440 for the NVIDIA card. Pixel rate is 0 MPixel/s versus 380.3 GPixel/s. Texture rate is 2,457.6 GTexel/s versus 1,045.9 GTexel/s. FP32 is 157.3 TFLOPS versus 66.94 TFLOPS. FP16 is 157.3 TFLOPS (1:1) versus 66.94 TFLOPS (1:1).

TDP is 2300 W versus 500 W. Slot width is EAM Module versus dual-slot. Power connectors are none versus 1x 16-pin. Suggested PSU is 2700 W versus 900 W. Bus interface is PCIe 6.0 x16 versus PCIe 5.0 x16. Display outputs are none versus 1x HDMI 2.1b and 3x DisplayPort 2.1b. API support is N/A for DirectX, OpenGL, and Vulkan on the AMD side, versus DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 on the NVIDIA side.

Dimensions are recorded only for the NVIDIA card: 267 mm in length, 111 mm in height, and 40 mm in width. Production status is listed as Active for the RTX 5090 SE and is not recorded for the MI455X. Release dates differ: the MI455X is dated 2026-07-22, while the RTX 5090 SE is dated 2025-12-31. The RTX 5090 SE has a launch MSRP of 1,499 USD; the MI455X has no launch MSRP recorded. Both cards hold a percentile rank of 50 against all GPUs in the database, with an average benchmark score of zero for each.